JUDGMENT OF THE GENERAL COURT (Seventh Chamber, Extended Composition)

24 September 2019 (*)

(State aid – Aid implemented by the Netherlands – Decision declaring the aid to be incompatible with the internal market and unlawful and ordering its recovery – Tax ruling – Transfer pricing – Calculation of the tax base – Arm’s length principle – Advantage – Reference system – Fiscal and procedural autonomy of the Member States)

In Cases T‑760/15 and T‑636/16,

Kingdom of the Netherlands, represented initially by M. Bulterman, B. Koopman, M. de Ree and M. Noort, subsequently by M. Bulterman, M. de Ree and M. Noort, acting as Agents,

applicant in Case T‑760/15,

supported by

Ireland, represented initially by E. Creedon, G. Hodge, A. Joyce and K. Duggan, subsequently by G. Hodge, A. Joyce, K. Duggan, M. Browne and J. Quaney, acting as Agents, and by M. Collins, P. Gallagher, Senior Counsel, B. Doherty and S. Kingston, Barristers,

intervener,

Starbucks Corp., established in Seattle, Washington (United States),

Starbucks Manufacturing Emea BV, established in Amsterdam (Netherlands),

represented by S. Verschuur, M. Petite and M. Stroungi, lawyers,

applicants in Case T‑636/16,

v

European Commission, represented, in Case T‑760/15, initially by P.‑J. Loewenthal and B. Stromsky, subsequently by P.‑J. Loewenthal and F. Tomat, acting as Agents, and in Case T‑636/16, by P.‑J. Loewenthal and F. Tomat, acting as Agents,

defendant,

APPLICATIONS based on Article 263 TFEU seeking annulment of Commission Decision (EU) 2017/502 of 21 October 2015 on State aid SA.38374 (2014/C ex 2014/NN) implemented by the Netherlands to Starbucks (OJ 2017 L 83, p. 38),

THE GENERAL COURT (Seventh Chamber, Extended Composition)

composed of M. van der Woude, President, V. Tomljenović (Rapporteur), E. Bieliūnas, A. Marcoulli and A. Kornezov, Judges,

Registrar: S. Spyropoulos, Administrator,

having regard to the written procedure and further to the hearing on 2 July 2018,

gives the following

Judgment

I.      Background to the dispute and legal framework

1        Starbucks Manufacturing Emea BV (‘SMBV’) is a subsidiary of the Starbucks group (‘the Starbucks group’), established in the Netherlands. The Starbucks group is composed of Starbucks Corp. and all the companies controlled by that corporation. Starbucks Corp. is headquartered in Seattle, Washington (United States). Alki LP (‘Alki’) is a subsidiary of the Starbucks group, established in the United Kingdom, which indirectly controls SMBV. Alki and SMBV concluded a roasting agreement (‘the roasting agreement’), which provides inter alia that SMBV is to pay Alki a royalty for the use of Alki’s intellectual property (IP) rights, including roasting methods and other roasting know-how (‘the royalty’).

2        Commission Decision (EU) 2017/502 of 21 October 2015 on State aid SA.38374 (2014/C ex 2014/NN) implemented by the Netherlands to Starbucks (OJ 2017 L 83, p. 38; ‘the contested decision’) concerns a measure relating to the application of the Netherlands corporate tax system to the specific case of SMBV.

A.      National legislative framework

3        According to the general Netherlands corporate income tax system, corporate income tax must be paid by companies established in the Netherlands – which are resident taxpayers – and by non-established companies – which are non-resident taxpayers – that have an economic activity in the Netherlands. According to Article 2 of the Wet op vennootschapsbelasting (Law on corporation tax; ‘CIT’) of 1969, established companies – which necessarily includes companies incorporated under Netherlands law – are subject to corporate income tax on their worldwide income. According to Article 3 of the CIT, non-established companies, for their part, are taxed on their income from Netherlands sources.

4        In that context, the corporate tax base is made up of the profits made by the corporate taxpayer. It follows from Article 8 of the CIT, read in conjunction with Section 3.8 of the Wet inkomstenbelasting (Income Tax Act) of 2001, that all taxpayers must be taxed according to the total profit concept. According to that concept, all corporate profits are taxed, provided that they derive from an economic or commercial activity. Section 3.8 of the Income Tax Act provides that ‘the profit from a business enterprise is the amount of the aggregate benefits that, under whatever name and in whichever forms, are derived from a business enterprise’. According to Section 3.25 of the Income Tax Act 2001, which applies also to corporate taxpayers by virtue of Article 8 of the CIT, the taxable yearly profits must be determined on the principles of sound business practice and in a consistent manner independently of the likely outcome.

5        In general, the taxable profits correspond to the accounting profits as reflected in the company’s profit and loss accounts. However, adjustments can be made based on specific tax provisions, such as applicable tax incentives, the participation exemption, corrections to the tax result from transactions not executed at arm’s length and the application of different depreciation rules under tax and accounting rules.

6        Article 8b(1) of the CIT provides that, ‘where an entity participates, directly or indirectly, in the management, control or capital of another entity, and conditions are made or imposed between these entities in their commercial and financial relations (transfer prices) which differ from conditions which would be made between independent parties, the profit of these entities will be determined as if the last mentioned conditions were made’.

7        Decree IFZ2001/295M of the Netherlands State Secretary for Finance of 30 March 2001, entitled ‘Transfer pricing method, application of the arm’s length principle and the OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations’ (‘the transfer pricing decree’), describes the manner in which the Netherlands tax administration interprets the arm’s length principle on the basis of Article 8b(1) of the CIT. The preamble of the transfer pricing decree provides the following: 

‘The policy of the Netherlands on the arm’s length principle in the field of international tax law is that this principle forms an integral part of the Netherlands system of tax law as a result of its incorporation in the broad definition of income recorded in Section 3.8 of the Income Tax Act. In principle, this means that the [Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, adopted by the Committee on Fiscal Affairs of the Organisation for Economic Co-operation and Development (OECD) on 27 June 1995] apply directly to the Netherlands under Section 3.8 of the Income Tax Act. There are a number of areas in which the OECD Guidelines provide scope for individual interpretation by the member countries. In a number of other areas, practical experience has shown that the OECD Guidelines are in need of clarification. This decree explains the Netherlands’ position in relation to these particular points and seeks, where possible, to remove any confusion.’ 

8        The transfer pricing decree is divided into twelve parts on the arm’s length principle, on transfer pricing methods, on administrative approaches aimed at preventing and resolving transfer pricing disputes, on secondary adjustments, on determination of the arm’s length price where the assessment is highly uncertain at the time of the transaction, on the provision of services within a group, on contributions to a cost-sharing agreement with a profit mark-up, on arm’s length remuneration for financial services, on subsidies, fiscal advantages and partially deductible costs, on the allocation of profits to a parent company and to a permanent establishment, on the entry into force of that decree and on the application of the current policy.

9        More specifically, the transfer pricing decree states in point 1 thereof, inter alia, that the arm’s length principle under Netherlands law is based, in general, on comparing the conditions of a transaction between affiliated companies with the conditions of a transaction between independent companies. The administration is entitled to expect a taxpayer to demonstrate that the transfer prices it applies comply with the arm’s length principle. In that context, the premiss must be that each of the companies concerned receives a remuneration which reflects the functions performed, taking into account the assets mobilised and the risks run. In addition, arm’s length remuneration must in theory be determined on the basis of transactions. In the event of difficulty, transactions may be assessed together in order to determine compliance with the arm’s length principle. Moreover, in examining multiannual data, the tax administration may not use hindsight.

10      In point 2 thereof, the transfer pricing decree refers to five methods, set out in the Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, adopted by the Committee on Fiscal Affairs of the Organisation for Economic Co-operation and Development on 27 June 1995 and revised on 22 July 2010 (‘the OECD Guidelines’), concerning the determination of transfer prices. Those methods include the comparable uncontrolled price method (‘the CUP method’) and the transactional net margin method (the ‘TNMM’). According to the transfer pricing decree, the OECD Guidelines assume a certain hierarchy between the methods, preference being given to the traditional transactional methods. The Netherlands tax administration must always conduct a transfer pricing audit from the perspective of the method chosen by the taxpayer at the date of the transaction. The decree states that that rule complies with paragraph 1.68 of the 1995 version of the OECD Guidelines. It follows that the taxpayer is in principle free to choose a transfer pricing method as long as the selected method leads to a result in conformity with the arm’s length principle. Even though, where it chooses a transfer pricing method, the taxpayer is deemed to take into account the reliability of that method in the situation in question, that approach is not specifically aimed at encouraging it to assess all the methods and justifying, thereafter, the reasons why the method it has chosen gives the best result under the conditions present.

11      Point 5 of the transfer pricing decree provides inter alia that it may be difficult, during the transfer of intangible goods such as patents, for example, to assess the value of those assets at the time of the transfer on account of insufficient knowledge about future advantages and risks. In the event that independent companies have opted for a price revision clause, in comparable circumstances, the tax administration must have the possibility of determining the price on the basis of such a clause. The objective is to have a regime in which the remuneration is in conformity with the advantages the intangible good will generate in the future.

B.      Advance pricing arrangement

12      On 28 April 2008, the Netherlands tax authorities concluded an advance pricing arrangement with SMBV (‘the APA’), the objective of it being to determine SMBV’s remuneration for its production and distribution activities, as described in the APA, within the Starbucks group (‘SMBV’s remuneration’). Thereafter, SMBV’s remuneration served to determine annually its taxable profit on the basis of Netherlands corporate income tax. According to its preamble, the APA is a tax agreement relating to the conformity with the arm’s length principle of the transfer prices to be used in the calculation of profit within an international group. It is apparent from [confidential](1) the APA that the agreement was intended inter alia to be used for the purposes of the annual corporate income tax declarations in the Netherlands. The APA was valid for the period between 1 October 2007 and 31 December 2017.

13      The APA provided for a method for determining SMBV’s remuneration in accordance with the arm’s length principle for its production and distribution activities within the Starbucks group. In addition, the APA endorsed the amount of the royalty, paid by SMBV to Alki for the use of roasting technologies, coffee blends and coffee roasting curves (‘the roasting IP’), in the context of the process of producing and providing coffee to shop operators.

14      More specifically, as regards the scope of the APA, [confidential]. Regarding SMBV’s functions, [confidential] it was mainly responsible for the production of roasted coffee beans and for the provision of roasted coffee beans and related products to Starbucks shops in the Europe, Middle East and Africa region (‘the EMEA region’) and that it was the owner of a roasting facility in the Netherlands. Moreover, the APA indicated that SMBV was licensing certain intellectual property rights belonging to Alki and that those rights were necessary to the process of producing and providing coffee to store operators. It was stated that, to that end, SMBV paid the royalty to Alki. [confidential] Furthermore, according to [confidential] the APA, SMBV carried out a distributor function for a number of other coffee-related products and, besides the logistics functions relating to its own production activities, also provided logistics support for other products on certain markets.

15      Regarding the transfer pricing method for SMBV’s production and distribution activities, [confidential] the APA stated inter alia that SMBV’s remuneration had to be determined on the basis of the cost plus method (see paragraph 187 below in relation to the meaning of that expression), and that it was at arm’s length if the ‘operating margin’ reached [confidential]% of the relevant cost base (‘SMBV’s cost base’). In addition, according to the APA, SMBV’s cost base did not include:

–        the costs of the Starbucks cups, paper napkins, etc.;

–        the costs of green coffee beans;

–        the logistics and distribution cost relating to services provided by third parties and to the remuneration for activities provided by third parties under consignment manufacturing contracts;

–        the royalty payment.

16      So far as concerns the royalty that SMBV had to pay annually to Alki, [confidential] the APA provided that it was fixed on the basis of the difference between the operational profit made in respect of the production and distribution function, before royalty expenses, and SMBV’s remuneration. The royalty payment was deductible for corporate income tax purposes and was not subject to Netherlands withholding tax.

C.      Background to the dispute

1.      Administrative procedure before the Commission

17      On 30 July 2013, the European Commission sent the Kingdom of the Netherlands an initial request for detailed information on its national tax ruling practice in the area of corporate taxation. To that effect, it requested that all rulings concerning SMBV and Starbucks Coffee Emea BV, two subsidiaries of the Starbucks group established in the Netherlands, be sent to it. In response to that request, the Kingdom of the Netherlands inter alia sent the APA.

18      On 11 June 2014, the Commission initiated the formal investigation procedure under Article 108(2) TFEU (‘the opening decision’) in respect of the APA, on the ground that that agreement might constitute State aid within the meaning of Article 107(1) TFEU.

19      Following the adoption of the opening decision, the Commission exchanged, on numerous occasions, correspondence with the Kingdom of the Netherlands and with those entities of the Starbucks group that were the correspondents of the Commission during the administrative procedure (‘the Starbucks correspondents’) in relation inter alia to the APA.

2.      Contested decision

20      On 21 October 2015, the Commission adopted the contested decision. In that decision, the Commission, first, found that the APA constituted aid incompatible with the internal market and, second, ordered the recovery of that aid. The contested decision is divided into 11 sections.

(a)    Description of the contested measure

21      In Section 2 of the contested decision, entitled ‘Description of the contested measure’, the Commission identified the APA as the measure at issue. It stated that the APA had been concluded on the basis of the transfer pricing report prepared by the Starbucks group’s tax advisor (‘the transfer pricing report’) and found that that document formed an integral part of the APA (recitals 40 and 46 of the contested decision).

22      First, the Commission noted that, in adopting the APA, the Netherlands tax authorities had accepted that the remuneration for SMBV’s activities in the Netherlands, as determined by the Starbucks group’s tax advisor, constituted an arm’s length remuneration. The Commission then noted that the Netherlands authorities had also accepted that the level of the royalty payment from SMBV to Alki corresponded to the difference between the realised operating profit before royalty payments and SMBV’s remuneration as provided for by the APA. It found that the APA provided that the royalty payment would be deductible from SMBV’s taxable profit and would not be subject to withholding tax in the Netherlands (recitals 40 to 44 of the contested decision).

23      Second, the Commission set out the content of the transfer pricing report. First of all, the Commission stated that the transfer pricing report presented Starbucks Coffee Emea as the principal establishment of the Starbucks group in the EMEA region. It was described as sublicensing the Starbucks group’s intellectual property rights (trade mark, technology and know-how) – for which it itself was paying a royalty to Alki – to third parties operating Starbucks branches. Those then paid Starbucks Coffee Emea for IP rights corresponding to a percentage of their turnover. In that regard, the Commission noted that the Starbucks group’s tax advisor considered that the CUP method could be used to determine the arm’s length price of payments of intra-group IP rights to Starbucks Coffee Emea.

24      Next, as regards SMBV, first, the Commission noted that the transfer pricing report was limited to describing it as an entity with the primary function of roasting green coffee beans and selling the roasted coffee on to affiliated and non-affiliated entities. As part of those activities, SMBV was required to observe the specifications provided by the companies of the Starbucks group that were established in the territory of the United States (‘Starbucks US’) and had the responsibility of ensuring that its production was in conformity with the quality standards of Starbucks US. SMBV also acted as distribution intermediary for various products not derived from café and provided ‘support to the supply chain’. Second, the Commission stated that, to carry out that activity, SMBV supplied itself with green coffee beans from a subsidiary of the Starbucks group established in Switzerland, Starbucks Coffee Trading SARL (‘SCTC’). SMBV also paid a royalty to Alki for the use of roasting processes and for the right to provide Starbucks branches with coffee. In that regard, the Commission observed that the transfer pricing report did not list the licensing arrangement under which SMBV paid a royalty to Alki among the most important transactions. The relationships between SMBV and the various entities of the Starbucks group were restated in Figure 1 of the contested decision, reproduced below:

25      Last, the Commission stated that, as regards the choice of transfer pricing method, the transfer pricing report selected the TNMM, a method whereby account was taken of the net margins that had been received in comparable transactions by non-affiliated companies. According to the transfer pricing report, that method was appropriate in the case at hand, since the differences between the transactions and the functions of the entities to be compared to determine the net margin were less the sources of error than in traditional methods (recital 55 of the contested decision).

26      The Commission stated that, in order to apply the TNMM, the tax advisor had chosen as profit level indicator the operating costs for the activities where SMBV provided added value. After a comparability search, the tax advisor had considered that the net profit of the entities comparable to SMBV corresponded to a mark-up on total costs. Thereafter, the Starbucks group’s tax advisor had made two adjustments in order to take account of the differences between the compared entities and SMBV, such as the risks assumed or the functions performed. The first adjustment sought to take into account the fact that SMBV’s cost base, to which the mark-up was applied, did not include the cost of green coffee beans. The second adjustment purported to take into account the fact that the comparable undertakings bore the cost of raw materials and that their return was calculated on a cost base including raw materials. Applying those two adjustments, the mark-up had thus been set at [confidential]% of SMBV’s cost base (recitals 56 to 61 of the contested decision).

27      Third, the Commission set out the content of the OECD Guidelines in their 1995 and 2010 versions. According to the Commission, the OECD Guidelines described five methods to approximate an arm’s length pricing of transactions and profit allocation between companies of the same group. In its view, they classified those five methods either as traditional transaction methods or as transactional profit methods. According to the contested decision, the traditional transaction methods were to be given priority. The CUP method and the TNMM were included among the five methods listed in the OECD Guidelines (recitals 67 to 70 of the contested decision).

28      The first method, the CUP method, is, as the Commission described, a traditional transaction method, consisting in comparing the price charged for the transfer of property or services in a transaction between two associated undertakings with the price charged for the transfer of property or services in a comparable transaction carried out in comparable circumstances between two independent undertakings (recitals 67 and 71 of the contested decision).

29      The second method, the TNMM, is, as the Commission describes, a transactional profit method, which involves estimating the potential amount of arm’s length profit for an entire activity, rather than for specific transactions. In that context, a profit level indicator would be selected, such as costs, turnover or fixed investment, to which would be applied a profit ratio reflecting that observed in comparable transactions on the market (recitals 67 and 72 to 74 of the contested decision).

(b)    Assessment of the contested measure

30      In Section 9 of the contested decision, entitled ‘Assessment of the contested measure’, the Commission concluded that State aid had been granted. It found that the four conditions for the existence of State aid had been satisfied.

31      After recalling the conditions for the existence of State aid, set out in Article 107(1) TFEU, the Commission found that the first condition for the existence of State aid, the requirement of an intervention by the State or through State resources, had been satisfied. It noted, to that end, first, that the APA contained the acceptance by the Netherlands tax administration of a method for allocating profits to SMBV within the Starbucks group, as proposed by the Starbucks group’s tax advisor. SMBV then calculated on that basis the annual amount of corporate income tax it was required to pay in the Netherlands. According to the Commission, the APA was therefore imputable to the Kingdom of the Netherlands. Second, the Commission found that the APA resulted in a lowering of SMBV’s tax liability in the Netherlands by deviating from the tax that SMBV would otherwise – in the absence of the APA – have been obliged to pay under the general Netherlands corporate income tax system. The Commission thus took the view that the APA had given rise to a reduction of the Kingdom of the Netherlands’ tax revenue (recitals 223 to 226 of the contested decision).

32      As regards the second and fourth conditions for a finding of State aid, first, the Commission found that the APA was liable to affect intra-Union trade, since SMBV was part of the Starbucks group, a globally active entity operating in all Member States of the European Union. Second, it maintained that, given that the APA reduced the tax burden that SMBV would otherwise have had to bear under the general corporate income tax system, it distorted or threatened to distort competition by strengthening SMBV’s financial position (recital 227 of the contested decision).

33      As regards the third condition for a finding of State aid, the Commission considered that the APA conferred a selective advantage on SMBV, in so far as it had resulted in a lowering of SMBV’s tax liability in the Netherlands as compared with what it would have had to pay under the general corporate income tax system and as compared with stand-alone undertakings (recital 228 of the contested decision).

34      As a preliminary point, the Commission observed that the case-law had established a three-step analysis to be used in determining whether a particular tax measure was selective. The first step is to identify the ‘reference system’, namely, the tax regime normally applicable to the beneficiary of the tax measure. In the second step, it is necessary to determine whether the tax measure derogates from that reference system, in so far as it differentiates between economic operators who, in the light of the objectives intrinsic to the reference system, are in a comparable legal and factual situation. In the third and final step, if the measure constitutes a derogation from the reference system, the Member State must establish that the derogation is justified by the nature or the general scheme of the reference system (recital 230 of the contested decision).

35      As regards the first step, identification of the reference system, the Commission considered the reference system to be the general Netherlands corporate income tax system, the objective of which is to tax the profits of all companies subject to tax in the Netherlands. It stated in that regard that companies established in the Netherlands are resident taxpayers and are subject to corporate income tax on their worldwide income. Companies not established in the Netherlands are non-resident companies and are subject to tax with regard to income from Netherlands sources. According to the Commission, integrated companies and stand-alone companies were in a comparable legal and factual situation in the light of that objective and were therefore subject to corporate income tax without distinction. In that regard, the difference in the modalities of calculating the taxable profits of integrated companies had no bearing on the objective of the reference system, taxation of the profits of all companies subject to tax in the Netherlands (recitals 231 to 244 of the contested decision).

36      As regards the second step set out in paragraph 34 above, namely, demonstrating a derogation from the reference system, the Commission indicated, first of all, that the question of whether a tax measure constituted a derogation from the reference system would generally coincide with identification of the advantage granted to the beneficiary under that measure. In its view, where a tax measure results in a reduction of the tax liability of a beneficiary compared to what it would otherwise have to pay were it not for that measure, that reduction constitutes both the advantage granted by the tax measure and the derogation from the reference system (recital 253 of the contested decision).

37      Next, the Commission recalled the case-law according to which, in the case of an individual measure, the identification of the economic advantage is, in principle, sufficient to support the presumption that that measure is selective. It stated that, in the case at hand, the APA granted to SMBV was an individual aid measure (recital 254 of the contested decision).

38      Last, the Commission asserted that, in the judgment of 22 June 2006, Belgium and Forum 187 v Commission (C-182/03 and C-217/03, EU:C:2006:416), the Court of Justice had found that a tax measure which results in an integrated company charging transfer prices that did not reflect those which would have been charged in conditions of free competition, that is prices negotiated by independent undertakings negotiating under comparable circumstances at arm’s length, conferred an advantage on that group company in so far as it resulted in a reduction of its taxable base and thus its tax liability under the ordinary corporate income tax system. The Commission recalled that the arm’s length principle consisted in the notion that transactions between intra-group companies were to be remunerated as if they had been agreed to by stand-alone companies negotiating under conditions of free competition. Accordingly, the Commission stated that it had to verify whether the methodology accepted by the Netherlands tax administration via the APA for the purposes of determining SMBV’s taxable profits in the Netherlands departed from a methodology that resulted in a reliable approximation of a market-based outcome and, therefore, from the arm’s length principle. In that case, the APA would be regarded as conferring a selective advantage on SMBV within the meaning of Article 107(1) TFEU (recitals 259 to 263 of the contested decision).

39      Consequently, the Commission found that the arm’s length principle necessarily formed an integral part of its assessment, under Article 107(1) TFEU, of the tax measures granted to integrated companies, independently of whether a Member State had incorporated that principle into its national legal system. The Commission then stated that, in response to the arguments of the Kingdom of the Netherlands raised during the administrative procedure, it had not examined whether the APA observed the arm’s length principle, as defined in Article 8b of the CIT and by the transfer pricing decree, but that it had sought to determine whether the Netherlands tax administration had conferred a selective advantage on SMBV within the meaning of Article 107(1) TFEU (recitals 264 and 265 of the contested decision).

40      In the light of those elements, the Commission set out a series of lines of reasoning to prove that the APA conferred a selective advantage on SMBV. In a primary reasoning, the Commission elaborated several lines of reasoning, some of which were subsidiary to each other, to demonstrate that the APA derogated from the general Netherlands corporate income tax system. In a subsidiary reasoning, the Commission maintained that the APA derogated from Article 8b(1) of the CIT and from the transfer pricing decree.

41      In particular, in the first place, in its primary reasoning aimed at demonstrating that the APA derogated from the general Netherlands corporate income tax system, the Commission noted that the Starbucks group’s tax advisor had chosen the TNMM for the purposes of determining SMBV’s taxable profits. It considered that different methodological choices, proposed by the Starbucks group’s tax advisor and endorsed by the Kingdom of the Netherlands, led to a reduction in the corporate income tax paid by SMBV as compared to stand-alone companies whose taxable profits were determined under market conditions (recitals 268 to 274 of the contested decision).

42      First, the Commission considered that the transfer pricing report had failed to examine whether the intra-group transfer for which the APA had actually been requested and granted – namely, the royalty SMBV paid to Alki for the licence to use the roasting IP – was in conformity with the arm’s length principle. Consequently, the transfer pricing report failed to identify or analyse controlled and uncontrolled transactions, which was a necessary first step in assessing the arm’s length nature of commercial conditions applicable between related parties for transfer pricing purposes (recitals 275 to 285 of the contested decision).

43      Second, regarding the level of the royalty, the Commission considered that, had the transfer pricing report correctly identified and examined the royalty, it should have resulted in an arm’s length value of zero. It noted inter alia that SMBV generated no profit from the use of the IP rights that were the subject of the royalty, in so far as it did not exploit them on the market. The Commission then considered that the profits paid to Alki by means of the royalty should have been fully taxed in the Netherlands (recitals 286 to 341 of the contested decision).

44      In examining the arm’s length nature of the royalty, the Commission applied the CUP method and identified a number of manufacturing agreements between the Starbucks group and third entities, or between third entities to the Starbucks group, as being comparable transactions.

45      In addition, the Commission rejected the arguments raised by the Kingdom of the Netherlands and Starbucks seeking to justify the amount of the royalty. First, it took the view that the consideration for the royalty payment could not be Alki’s taking over of SMBV’s entrepreneurial risks; otherwise, integrated companies could contractually reallocate risk and thereby prevent any application of the arm’s length principle. Second, the Commission added that the royalty payment could not be justified by size of the amounts paid by Alki to Starbucks US.

46      Third, regarding the level of the purchase price of green coffee beans, the Commission noted that that transaction had not been examined in the transfer pricing report, when that report identified it as one of the main transactions effected by SMBV. Using SCTC’s financial data, the Commission calculated the average gross margin on the costs of green coffee beans for the APA’s validity period. The Commission considered that the gross margin between 2011 and 2014, which resulted in a significant increase in the price of green coffee beans to be borne by SMBV compared to the costs borne by SCTC, did not reflect a reliable approximation of a market-based outcome. It concluded that the price premium paid by SMBV, in lowering the latter’s recorded profits and therefore taxable base, constituted a selective advantage (recitals 342 to 361 of the contested decision).

47      In the second place, still in its primary line of reasoning to demonstrate that the APA derogated from the general Netherlands corporate income tax system, but subsidiary to the criticisms set out in paragraphs 42 to 46 above, the Commission considered that, in any event and even assuming that the TNMM were appropriate for identifying the profits made by SMBV, the transfer pricing report had incorrectly applied the TNMM. It concluded that, since that method did not result in an arm’s length outcome, the Netherlands tax authorities could not approve it in the APA (recitals 362 to 408 of the contested decision).

48      First, the Commission considered that the transfer pricing report had erroneously identified SMBV as being the least complex entity, and therefore as being the ‘tested party’, for the purposes of the application of the TNMM. It added that, on the contrary, SMBV should have been identified as the most complex entity, in so far as, first, Alki performed only limited functions and, second, apart from the fact that SMBV performed functions other than roasting, that function was not a routine activity but an essential one (recitals 362 to 377 of the contested decision).

49      Second, the Commission considered that the profit level indicator used in the transfer pricing report, namely, operating costs, was inappropriate. According to the Commission, the Starbucks group’s tax advisor had erroneously found that roasting was SMBV’s principal function rather than reselling and distribution. The Commission then concluded that the use of the sales recorded by SMBV as profit level indicator was more appropriate and would have led to a higher remuneration of SMBV’s activity. In support of that conclusion, the Commission calculated a profitability ratio from a group of independent entities exercising the same reselling and roasting activities as SMBV. It concluded, after comparison with Starbucks Manufacturing Corporation (‘SMC’) – the only other entity of the group to exercise roasting activities for the group – that SMC was [confidential] times more profitable than SMBV on the basis of the APA (recitals 379 to 400 of the contested decision).

50      Third, the Commission considered that, in any event and even assuming that operating costs were a profit level indicator appropriate for calculating SMBV’s transfer prices, the two adjustments made by the tax advisor in the transfer pricing report did not enable a reliable approximation of a market-based outcome to be reached. The Commission criticised, first, the use of a ‘working capital adjustment’ and, second, the exclusion of the costs of the undertakings designated in recital 300 of the contested decision as published in the Official Journal of the European Union by the term ‘unaffiliated manufacturing company 1’ (‘unaffiliated manufacturing company 1’) from the cost base used as profit level indicator (recitals 401 to 408 of the contested decision).

51      The Commission then concluded that the methodology accepted by the Netherlands tax authorities, whereby the profits generated by SMBV in excess of the [confidential]% margin of operating costs had to be paid as a royalty to Alki, was not in line with the arm’s length principle and led to a reduction in SMBV’s tax burden.

52      It is thus apparent from the foregoing that, in the examination in respect of the general Netherlands corporate income tax system, the Commission raised six errors justifying the conclusion that there was a selective advantage in the case at hand. In that examination, the first three errors come under  a principal position, whereas the other three errors come under a subsidiary position and are subsidiary to each other.

53      Specifically, as regards the principal position, the Commission considered that the method accepted by the APA departed from a methodology that led to a reliable approximation of a market-based outcome in line with the arm’s length principle, in so far as:

–        the choice of the TNMM was erroneous and the transfer pricing report did not examine the intra-group transaction for which the APA had effectively been requested and granted (‘the first line of reasoning’);

–        first, the APA did not establish a methodology to ensure that the royalty paid by SMBV to Alki was in line with the arm’s length principle; however, the CUP method should have been applied in order to determine the amount of the royalty paid by SMBV to Alki; under that method, the royalty should have been zero (‘the second line of reasoning’);

–        second, the APA did not examine whether the level of purchase price of green coffee beans was in line with the arm’s length principle; that level was overvalued (‘the third line of reasoning’).

54      With regard to the subsidiary position, the Commission considered that, even assuming that the TNMM were the appropriate method for determining the profits made by SMBV, the transfer pricing report had incorrectly applied the TNMM. In that regard, the Commission took the view that:

–        the method accepted by the APA derogated from a method that gave a reliable approximation of a market-based outcome in line with the arm’s length principle, in so far as SMBV had been incorrectly identified as the least complex entity and thus as the tested entity for the purposes of the application of the TNMM (‘the fourth line of reasoning’);

–        in the alternative, the method accepted by the APA did not give a reliable approximation of a market-based outcome in line with the arm’s length principle, in so far as SMBV’s functions had been incorrectly analysed and the choice of operating costs as profit level indicator was incorrect (‘the fifth line of reasoning’);

–        in the alternative, the method accepted by the APA did not give a reliable approximation of a market-based outcome in line with the arm’s length principle, in so far as the adjustments made to the mark-up were inappropriate (‘the sixth line of reasoning’).

55      In the third place, in its subsidiary reasoning aimed at demonstrating that the APA derogated from Article 8b(1) of the CIT and from the transfer pricing decree, the Commission considered that, even supposing that the relevant reference system were not composed of the general rules on corporate income tax, but, as the Netherlands authorities maintained, only of the provisions enshrining the arm’s length principle in Netherlands law, namely Article 8b(1) of the CIT and the transfer pricing decree, the APA, in approving a method of determining SMBV’s profits that did not result in an arm’s length outcome, also derogated from that reference system. To that end, the Commission referred to its analysis conducted in respect of the general Netherlands corporate income tax system and the six lines of reasoning set out in paragraphs 52 to 54 above (the ‘reasoning in respect of the limited reference system’) (recitals 409 to 412 of the contested decision).

56      So far as concerns the third step of the analysis of the selectivity of tax measures, as identified in paragraph 34 above, the Commission found that the derogation from the reference system was not justified. In that regard, it noted that neither the Netherlands authorities nor Starbucks had put forward any possible justifications for SMBV’s selective treatment, when the burden of proof in that regard lay with them. The Commission moreover added that it had not been able to identify any possible justification (recitals 413 and 414 of the contested decision).

57      The Commission concluded that SMBV’s APA conferred on SMBV a selective advantage, within the meaning of Article 107(1) TFEU, in so far as it had validated a method for allocating profits to SMBV which could not be regarded as resulting in a reliable approximation of a market-based outcome in line with the arm’s length principle. According to the Commission, that method had led to a reduction in SMBV’s tax burden, primarily, under the general Netherlands corporate income tax system, by comparison with stand-alone companies, and, in the alternative, under Article 8b(1) of the CIT and the transfer pricing decree, by comparison with other integrated companies (recitals 415 and 416 of the contested decision).

58      Consequently, the Commission concluded that the APA constituted State aid (recitals 422 and 423 of the contested decision).

59      The Commission then considered that the aid granted to SMBV was incompatible with the internal market. In its view, the Kingdom of the Netherlands had not invoked any of the grounds for a finding of compatibility provided for in Article 107(2) and (3) TFEU. The aid in question, which was to be considered operating aid, could not normally be considered compatible with the internal market (recitals 431 to 434 of the contested decision).

60      Moreover, the Commission found that the Kingdom of the Netherlands had not notified it, in accordance with Article 108(3) TFEU, of any plan corresponding to the APA and had not complied with the standstill obligation incumbent on it under that article. It could therefore only be unlawful State aid, put in effect in contravention of that provision (recitals 435 and 436 of the contested decision).

61      Furthermore, the Commission specified that the information on which it had based its decision was available for the Netherlands tax administration at the time the APA was adopted. It added, in respect of the cost of green coffee beans, that the transfer pricing report did not examine the prices charged by SCTC to SMBV under their green coffee bean sourcing agreement and that, had that transaction been examined in the APA in 2008 in order to determine its arm’s length price, the APA would not have been able to leave room for the price increases observed in 2011 (recitals 424 to 427 of the contested decision).

62      Last, the Commission identified SMBV and the Starbucks group as a whole as aid beneficiaries, on the ground that they formed a single economic entity (recitals 417 to 419 of the contested decision).

(c)    Recovery of the State aid

63      In Section 10 of the contested decision, entitled ‘Recovery’, first, the Commission inter alia considered that it was not required to quantify the exact amount of the aid to be recovered, but rather it was sufficient for its decision to include information enabling the addressee of the decision to work out that amount itself without overmuch difficulty. In the present case, the Commission considered that, since the royalty amount had to be zero, SMBV’s accounting profits should have been used to calculate its taxable profits. Moreover, those profits should have been increased by the difference between the price paid for the green coffee beans and the price that ought to have been paid. In that regard, the Commission considered that a gross margin of [confidential]% for SCTC constituted an arm’s length price for the purchase of the coffee beans. It then specified that the sum to be recovered corresponded to the difference between the tax that should have been paid on the basis of that price and the amount actually paid under the APA (recitals 442 to 448 of the contested decision).

64      Second, the Commission found that, in the first instance, the Kingdom of the Netherlands was required to recover from SMBV the aid and that, should SMBV not have been in a position to make the repayment, the Kingdom of the Netherlands was to recover the balance from Starbucks Corp., since it was the entity which controlled the Starbucks group (recital 449 of the contested decision).

(d)    Conclusion

65      In conclusion, the Commission found that the Kingdom of the Netherlands, by concluding the APA, had unlawfully granted State aid to SMBV and to the Starbucks group, in breach of Article 108(3) TFEU, that the Kingdom of the Netherlands was required to recover it, by virtue of Article 16 of Regulation (EU) 2015/1589 of 13 July 2015 laying down detailed rules for the application of Article 108 TFEU (OJ 2015 L 248, p. 9), from SMBV and, if SMBV failed to repay the full amount of the aid, from Starbucks Corp. for the amount of aid outstanding (recital 450 of the contested decision).

66      The operative part of the contested decision reads as follows:

‘Article 1

The advanced pricing arrangement entered into by the [Kingdom of the] Netherlands on 28 April 2008 with [SMBV], which enables the latter to determine its corporate income tax liability in the Netherlands on a yearly basis for a period of 10 years, constitutes aid within the meaning of Article 107(1) of the Treaty on the Functioning of the European Union that is incompatible with the internal market and that was unlawfully put into effect by the [Kingdom of the] Netherlands in breach of Article 108(3) of the Treaty.

Article 2

The [Kingdom of the] Netherlands shall recover the incompatible and unlawful aid referred to in Article 1 from [SMBV].

Any sums that remain unrecoverable from [SMBV], following the recovery described in the preceding paragraph, shall be recovered from Starbucks [Corp.].

The sums to be recovered shall bear interest from the date on which they were put at the disposal of the beneficiaries until their actual recovery.

The interest shall be calculated on a compound basis in accordance with Chapter V of Regulation (EC) No 794/2004.

Article 3

Recovery of the aid granted referred to in Article 1 shall be immediate and effective.

The [Kingdom of the] Netherlands shall ensure that this Decision is implemented within four months following the date of notification of this Decision.

Article 4

Within two months following notification of this decision, the [Kingdom of the] Netherlands shall submit information regarding the methodology used to calculate the exact amount of aid.

The [Kingdom of the] Netherlands shall keep the Commission informed of the progress of the national measures taken to implement this Decision until recovery of the aid granted referred to in Article 1 has been completed. It shall immediately submit, on simple request by the Commission, information on the measures already taken and planned to comply with this Decision.

...’

II.    Procedure and forms of order sought

A.      Written part of the procedure in Case T‑760/15

67      By application lodged at the Registry of the General Court on 23 December 2015, the Kingdom of the Netherlands brought the action in Case T‑760/15. The Commission lodged a defence on 30 March 2016. The reply and the rejoinder were lodged on 14 June 2016 and 9 September 2016, respectively.

1.      Composition of the formation of the Court and priority treatment

68      By document lodged at the Court Registry on 20 June 2016, the Kingdom of the Netherlands requested that Case T‑760/15 be heard and determined by a Chamber sitting in extended composition. The Court took formal note, pursuant to Article 28(5) of its Rules of Procedure, of the fact that Case T‑760/15 had been referred to the Fifth Chamber, Extended Composition.

69      Following a change in the composition of the Chambers of the Court on 26 September 2016, the Judge-Rapporteur was assigned, pursuant to Article 27(5) of the Rules of Procedure, to the Seventh Chamber, Extended Composition, of the General Court to which Case T‑760/15 was accordingly allocated.

70      Since a member of the Seventh Chamber, Extended Composition, of the General Court was unable to sit, the President of the General Court, by decision of 26 April 2017, designated the Vice-President of the General Court to complete the chamber.

71      By decision of 12 December 2017, the President of the Seventh Chamber, Extended Composition, of the General Court approved the proposal of the Judge-Rapporteur that Case T‑760/15 be given priority under Article 67(2) of the Rules of Procedure.

2.      Interventions

72      By document lodged at the Court Registry on 6 April 2016, the United Kingdom of Great Britain and Northern Ireland applied for leave to intervene in Case T‑760/15 in support of the form of order sought by the Commission.

73      By document lodged at the Court Registry on 7 April 2016, Ireland applied for leave to intervene in Case T‑760/15 in support of the form of order sought by the Kingdom of the Netherlands.

74      By order of 13 June 2016, the President of the Fifth Chamber of the General Court granted the applications for leave to intervene of the United Kingdom and Ireland.

75      By document lodged at the Court Registry on 9 November 2016, the United Kingdom withdrew its intervention. By order of 12 December 2016, the President of the Seventh Chamber, Extended Composition, of the General Court removed the United Kingdom as intervener from Case T‑760/15.

3.      Applications for confidential treatment

76      By document lodged at the Court Registry on 26 February 2016, the Kingdom of the Netherlands applied for confidential treatment, vis-à-vis the public, of part of the application and certain documents annexed to it.

77      By document lodged at the Court Registry on 17 May 2016, the Kingdom of the Netherlands applied for confidential treatment, vis-à-vis Ireland, of part of the application and certain documents annexed to it, as well as the contested decision and part of the defence.

78      By document lodged at the Court Registry on 17 May 2016, the Commission applied for confidential treatment, vis-à-vis Ireland, of part of the defence.

79      By document lodged at the Court Registry on 1 July 2016, the Kingdom of the Netherlands applied for confidential treatment, vis-à-vis Ireland, of part of the reply and certain documents annexed to it.

80      By document lodged at the Court Registry on 21 July 2016, the Kingdom of the Netherlands informed the Court that it had reached an agreement with the Commission concerning the non-confidential version of the contested decision for publication and that it was amending the applications for confidential treatment it had made in Case T‑760/15 with regard to the contested decision in accordance with that agreement.

81      By document lodged at the Court Registry on 11 October 2016, the Kingdom of the Netherlands applied for confidential treatment, vis-à-vis Ireland, of part of the rejoinder and certain documents annexed to it.

82      Following its admission as intervener, Ireland received only non-confidential versions of the procedural documents and raised no objection to the applications for confidential treatment made with regard to it.

83      On the proposal of the Judge-Rapporteur, the Seventh Chamber, Extended Composition, of the General Court adopted a measure of organisation of procedure provided for in Article 89 of the Rules of Procedure by which the Kingdom of the Netherlands was invited to revise its applications for confidential treatment of the APA, the roasting agreement between SMBV and Alki, mentioned in recital 142 of the contested decision, and the transfer pricing report, in order to remove certain inconsistencies in those applications. The Kingdom of the Netherlands provided new non-confidential versions of those documents within the prescribed period.

4.      Forms of order sought

84      The Kingdom of the Netherlands claims that the Court should:

–        annul the contested decision;

–        order the Commission to pay the costs in Case T‑760/15.

85      The Commission contends that the Court should:

–        dismiss the application in Case T‑760/15 as unfounded;

–        order the Kingdom of the Netherlands to pay the costs in Case T‑760/15.

86      Ireland claims that the Court should annul the contested decision in accordance with the form of order sought by the Kingdom of the Netherlands.

B.      Written part of the procedure in Case T‑636/16

87      By application lodged at the Court Registry on 5 September 2016, Starbucks Corp. and Starbucks Manufacturing Emea (together, ‘Starbucks’) brought the action in Case T‑636/16. The Commission lodged a defence on 16 March 2017. The reply and the rejoinder were lodged on 26 June and 20 October 2017, respectively.

1.      Composition of the formation of the Court and priority treatment

88      On the proposal of the Seventh Chamber of the General Court, the Court decided, on 12 July 2017, pursuant to Article 28 of the Rules of Procedure, to refer the case to a Chamber sitting in extended composition.

89      Since a member of the Seventh Chamber, Extended Composition, of the General Court was unable to sit, the President of the General Court, by decision of 1 August 2017, designated the Vice-President of the General Court to complete the Chamber.

90      By decision of 12 December 2017, the President of the Seventh Chamber, Extended Composition, of the General Court approved the proposal of the Judge-Rapporteur that Case T‑636/16 be given priority under Article 67(2) of the Rules of Procedure.

2.      Applications for confidential treatment

91      By document lodged at the Court Registry on 7 April 2017 and regularised by documents lodged on 23 April 2018, Starbucks applied for confidential treatment, vis-à-vis Ireland, of certain information in the application, in the defence, in the reply, in the rejoinder as well as in certain of the annexes to those submissions.

3.      Forms of order sought

92      Starbucks claims that the Court should:

–        annul Articles 1 to 4 of the contested decision;

–        in the alternative, annul Article 2(1) of the contested decision;

–        order the Commission to pay the costs in Case T‑636/16.

93      The Commission contends that the Court should:

–        dismiss the action in Case T‑636/16 as unfounded;

–        order Starbucks to pay the costs in Case T‑636/16.

C.      Joinder for the purposes of the oral part of the procedure, and the oral part of the procedure

94      By document lodged at the Court Registry on 23 February 2017, Starbucks applied for Cases T‑760/15 and T‑636/16 to be joined for the purposes of the oral part of the procedure.

95      By decision of 7 June 2017, the President of the Seventh Chamber, Extended Composition, of the General Court decided not to join, at that stage of the procedure, Cases T‑760/15 and T‑636/16.

96      By decision of the President of the Seventh Chamber, Extended Composition, of the General Court of 8 May 2018, Cases T‑760/15 and T‑636/16 were joined for the purposes of the oral part of the procedure, in accordance with Article 68 of the Rules of Procedure.

97      Acting on a report from the Judge-Rapporteur, the Court decided to open the oral part of the procedure and, by way of the measures of organisation of procedure under Article 89 of the Rules of Procedure, asked the parties to answer questions in writing. The parties responded to that measure of organisation of procedure within the prescribed period.

98      By documents lodged at the Court Registry on 7 and 15 June 2018, Starbucks requested confidential treatment of certain information contained in its response to the measure of organisation of procedure and in that of the Commission.

99      By document lodged at the Court Registry on 8 June 2018, Starbucks submitted observations on the report for the hearing.

100    By document lodged at the Court Registry on 14 June 2018, the Commission requested that Starbucks’ observations on the report for the hearing be withdrawn from the case file.

101    After having received the only non-confidential versions of the documents mentioned in paragraphs 91, 98 and 99 above, Ireland raised no objection to the applications for confidential treatment made with regard to it.

102    By document lodged at the Court Registry on 26 June 2018, Starbucks requested authorisation to use technical means during the hearing and proposed having recourse, during the hearing, to an expert. At the hearing, the Commission was requested to submit orally its observations on that request and in turn requested authorisation to use technical means during the hearing.

103    The parties presented oral argument, with the use of the technical means requested, and answered the questions put by the Court at the hearing on 2 July 2018.

104    The parties were heard during the hearing on a possible joinder of Cases T‑760/15 and T‑636/16 for the purposes of the decision closing the proceedings, and the Court took formal note of it in the minutes of the hearing.

III. Law

105    For the purposes of the present actions, it is necessary, at the outset, to address certain procedural questions raised by the parties, before analysing the substantive pleas the parties raise.

A.      Procedural matters

106    Regarding the procedural matters that are at issue in the case at hand, first of all, it is necessary to examine the possible joinder of the present case for the purposes of the decision closing the proceedings. Next, it is appropriate to examine the Commission’s request that Starbucks’ observations of 8 June 2018 on the report for the hearing be withdrawn from the case file. Last, it is appropriate to examine the issue of the admissibility of Annex A.7 to the application in Case T‑760/15, which has been disputed by the Commission.

1.      Joinder of the present cases for the purposes of the decision closing the proceedings 

107    In accordance with Article 19(2) of Rules of Procedure, the President of the Seventh Chamber, Extended Composition, of the General Court referred the decision as to whether Cases T‑760/15 and T‑636/16 should be joined for the purposes of the decision closing the proceedings, which fell within his  remit, to the Seventh Chamber, Extended Composition, of the General Court.

108    The parties having been heard at the hearing with respect to a possible joinder of the cases, it is appropriate for Cases T‑760/15 and T‑636/16 to be joined for the purposes of the decision which closes the proceedings, on account of the connection between them.

2.      Request that Starbucks’ observations on the report for the hearing be withdrawn from the case file 

109    By letter of 14 June 2018, the Commission requested the General Court to withdraw from the case file of Cases T‑760/15 and T‑636/16 Starbucks’ letter of 8 June 2018 (see paragraph 100 above), to the extent that it contained observations on the report for the hearing, on the ground that such observations are provided for neither in the Rules of Procedure nor in the practice rules for the implementation of the latter.

110    First, it is appropriate to recall that, by decision of 13 June 2018, the President of the Seventh Chamber, Extended Composition, of the General Court decided to place on the case file Starbucks’ letter of 8 June 2018. Second, it must be recalled that the General Court is the sole judge of whether documents not provided for in the Rules of Procedure need to be placed on the case file. Accordingly, the Commission’s request that the letter of 8 June 2018 be withdrawn from the case file must be rejected.

111    However, according to Article 84(1) of the Rules of Procedure, no new plea in law may be introduced in the course of proceedings unless it is based on matters of law or fact which have come to light in the course of the procedure.

112    Given that Starbucks has not provided any justification for the late submission of the arguments raised in its letter of 8 June 2018, it is necessary, as the Commission has argued, to reject them as inadmissible to the extent that they go beyond mere observations on the confidentiality and accuracy of the report for the hearing by amending the pleas raised in the application. 

3.      Admissibility of Annex A.7 to the application in Case T‑760/15 

113    The Commission disputes the admissibility of Annex A.7 to the application in Case T‑760/15, which contains a schematic comparison of certain aspects of the functioning of the contracts concluded between the Starbucks group and certain third parties. According to the Commission, the essential factual and legal elements on which the action is based must be set out – or otherwise be deemed inadmissible – at least in summary form, but coherently and intelligibly, from the actual text of the application. That condition is not, in its view, satisfied in this case.

114    It must be noted that, under Article 21 of the Statute of the Court of Justice of the European Union and Article 76(d) of the Rules of Procedure, each application is required to state the subject matter of the proceedings, the pleas in law and arguments on which the application is based and a summary of those pleas. That statement must be sufficiently clear and precise to enable the defendant to prepare his defence and the Court to rule on the application, if necessary, without any further information. It is necessary, for an action to be admissible, that the basic matters of law and fact relied on be indicated, at least in summary form, coherently and intelligibly in the application itself. Whilst the body of the application may be supported and supplemented on specific points by references to extracts from documents annexed thereto, a general reference to other documents, even those annexed to the application, cannot make up for the absence of the essential arguments in law which, in accordance with the abovementioned provisions, must appear in the application. The annexes may be taken into consideration only in so far as they support or supplement pleas or arguments expressly set out by applicants in the body of their pleadings and in so far as it is possible to determine precisely what are the matters they contain that support or supplement those pleas or arguments. Furthermore, it is not for the Court to seek and identify in the annexes the pleas and arguments on which it may consider the action to be based, since the annexes have a purely evidential and instrumental function (see judgment of 14 March 2013, Fresh Del Monte Produce v Commission, T‑587/08, EU:T:2013:129, paragraphs 268 to 271 and the case-law cited).

115    In the case at hand, it is appropriate to observe that, as regards the contracts concluded between the Starbucks group and external roasters and manufacturers of coffee-derived products, the Kingdom of the Netherlands submits that Annex A.7 contains a ‘schematic overview of the points of comparison of the contracts produced by the Commission ... which includes the three differences cited’ in the application in Case T‑760/15. However, the Kingdom of the Netherlands sets out, in paragraphs 140 to 155 of the application in Case T‑760/15, the reasons which demonstrate, in its view, that the contracts that the Commission invokes in the contested decision are not comparable with the contractual relationship between Alki and SMBV.

116    In that regard, it must be held that all the arguments contained in Annex A.7 to the application in Case T‑760/15 are set out in a sufficiently clear and precise manner in paragraphs 140 to 155 of the application in Case T‑760/15. Thus, even without Annex A.7 to the application in Case T‑760/15, the Commission would have been capable of preparing its defence and the Court of ruling on the action. The only added value of Annex A.7 to the application in Case T‑760/15 consists, therefore, in indicating which specific contracts are concerned by the respective arguments of the Kingdom of the Netherlands when it refers, in paragraphs 140 to 155 of the application, to the ‘majority’ or ‘most’ of those contracts.

117    It is therefore appropriate to reject the Commission’s argument that Annex A.7 to the application in Case T‑760/15 should be rejected as inadmissible.

B.      Pleas raised and structure of the examination of the present actions

118    The actions brought in Cases T‑760/15 and T‑636/16 seek the annulment of the contested decision to the extent that it classifies the APA as State aid for the purposes of Article 107(1) TFEU and to the extent that it orders the recovery of sums that were not collected by the Kingdom of the Netherlands from SMBV by way of corporate income tax.

119    In support of their actions, the Kingdom of the Netherlands and Starbucks put forward five and two pleas, respectively, which overlap for the most part.

120    By the first plea in Case T‑760/15 as well as by the first part of the first plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks call into question the Commission’s examination of the selective nature of the APA. Specifically, they argue that the Commission used an erroneous reference system for the examination of the selectivity of the APA.

121    By the second, third and fourth pleas in Case T‑760/15 as well as by the second part of the first plea and the second plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks argue that the Commission’s analysis according to which the APA conferred an advantage on SMBV is erroneous.

122    Specifically, by the second plea in Case T‑760/15 as well as by the second part of the first plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks invoke, in essence, an infringement of Article 107 TFEU, in that the Commission erroneously examined whether there was an advantage in relation to the arm’s length principle particular to EU law and thereby violated the Member States’ fiscal autonomy.

123    By the third plea in Case T‑760/15 as well as by the third part of the first plea and the first, second, fourth and fifth parts of the second plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks claim, in essence, infringement of Article 107 TFEU in that the Commission erroneously considered the choice of the TNMM to set transfer pricing to constitute an advantage. The Kingdom of the Netherlands and Starbucks dispute, in essence, the Commission’s principal line of reasoning regarding the existence of a tax advantage in favour of SMBV, set out in recitals 255 to 361 of the contested decision. Those pleas concern the first to third lines of reasoning mentioned in paragraph 53 above.

124    By the fourth plea in case T‑760/15 as well as by the third part of the second plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks claim infringement of Article 107 TFEU to the extent that the Commission erroneously considered the detailed rules for the application of the TNMM as validated in the APA to confer an advantage on SMBV. Those pleas concern the fourth to sixth lines of reasoning mentioned in paragraph 54 above.

125    By the fifth plea in Case T‑760/15, the Kingdom of the Netherlands claims breach of the principle of due diligence.

126    With regard to the analysis of the pleas raised by the Kingdom of the Netherlands and Starbucks, first of all, it is necessary to examine, the plea disputing the existence of the arm’s length principle as the Commission describes it in the contested decision. Next, it is appropriate to examine the pleas disputing that, in its first six lines of reasoning set out in paragraphs 53 and 54 above, the Commission demonstrated that the APA derogated from the general Netherlands corporate income tax system and conferred an advantage, within the meaning of Article 107 TFEU, on SMBV. Additionally, it is necessary to examine the plea disputing that, in its reasoning in respect of the limited reference system, set out in paragraph 55 above, the Commission demonstrated that the APA derogated from the limited reference system composed of Article 8b of the CIT and from the transfer pricing decree and conferred an advantage, within the meaning of Article 107 TFEU, on SMBV. Last, should the examination carried out on the existence of an advantage lead to the rejection of those pleas, it will be appropriate to examine the pleas alleging an absence of selectivity of the contested measure and breach of the obligation of due diligence.

127    In that regard, it must moreover be borne in mind that, according to the case-law, classification as State aid requires all the conditions referred to in Article 107(1) TFEU to be fulfilled. It is thus established that, for a measure to be categorised as State aid within the meaning of that provision, there must, first, be an intervention by the State or through State resources; second, the intervention must be liable to affect trade between Member States; third, it must confer a selective advantage on the recipient; and, fourth, it must distort or threaten to distort competition (see judgment of 21 December 2016, Commission v Hansestadt Lübeck, C‑524/14 P, EU:C:2016:971, paragraph 40 and the case-law cited).

128    In the case at hand, however, as is apparent from the presentation of the pleas, which is set out in paragraphs 118 to 125 above, the Kingdom of the Netherlands and Starbucks do not dispute the assessment made by the Commission regarding the first two conditions and the fourth condition to be satisfied in order for a measure to be classified as State aid. They do not dispute that, supposing that the Commission demonstrated that the APA conferred a tax advantage, that advantage constituted intervention by the State or through State resources, that it was liable to affect trade between Member States and that it distorted or threatened to distort competition. The four first pleas raised in Case T‑760/15 seek, in essence, to call into question the Commission’s finding that the APA conferred a selective advantage on SMBV.

129    Moreover, as regards demonstrating the selective advantage, it must be noted that the Commission’s approach of examining the criteria of advantage and selectivity concurrently is not itself incorrect, given that both the advantage and the selective nature of that advantage are examined. Nevertheless, the Court considers it appropriate to consider, first of all, whether the Commission was entitled to conclude that there was an advantage, before going on, if necessary, to examine whether that advantage had to be considered to be selective.

130    Thereafter, it will be appropriate to analyse the arguments of the Kingdom of the Netherlands and of Starbucks on the absence of an advantage, within the meaning of Article 107 TFEU, conferred on SMBV by the APA.

C.      Existence of an arm’s length principle in the field of State aid control and compliance with the principle of Member States’ fiscal autonomy

131    By its second plea, the Kingdom of the Netherlands argues that the Commission erred in identifying an arm’s length principle particular to EU law and in identifying it as a criterion for assessing the existence of State aid. Starbucks raises, in essence, the same complaints, in the second part of its first plea.

132    First, the Kingdom of the Netherlands argues that the judgment of 22 June 2006, Belgium and Forum 187 v Commission (C‑182/03 and C‑217/03, EU:C:2006:416), on which the Commission relies to identify an arm’s length principle particular to EU law, is not authority for the existence of such a principle. In addition, the Commission neither stated the basis from which it had identified the existence of an arm’s length principle in EU law nor explained the content of that principle. Starbucks adds that, notwithstanding the fact that observance of Article 107 TFEU effectively constitutes a limit on Member States’ fiscal autonomy, the Commission exceeded the powers conferred on it by Article 107 TFEU. Starbucks criticises the Commission for having replaced, purportedly under the principle of equal treatment, Netherlands rules of tax law with a transfer pricing principle developed autonomously and, thus, for having imposed substantive rules of tax law.

133    Second, the Kingdom of the Netherlands argues that the Commission could not examine the APA under an arm’s length principle particular to EU law, since only the legislation and national rules of the Member State concerned are relevant for the purposes of State aid control. Specifically, the Kingdom of the Netherlands contends that the existence of an advantage can be examined only by reference to the charges which are normally included in the budget of the undertaking under national law and not by reference to an arm’s length principle particular to EU law. Starbucks adds, furthermore, that the Commission did not take Netherlands law into account and that its reasoning deviates from – and indeed conflicts with – Netherlands transfer pricing rules.

134    First of all, Ireland adds that the Commission, which was required to identify a derogation, did not compare Starbucks’ situation to that of any other taxpayer but merely sought to apply the arm’s length principle. Next, it argues that the Commission cannot impose rules that have never been incorporated into the national system. Accepting a principle of equal treatment in tax matters would then encroach on the sovereignty and the autonomy of the Member States. Last, Ireland submits that the judgment of 22 June 2006, Belgium and Forum 187 v Commission (C‑182/03 and C‑217/03, EU:C:2006:416), did not identify an arm’s length principle particular to EU law, in so far as, in that case, first, the arm’s length principle had been incorporated into Belgian national law and, second, the judgment referred to the OECD Guidelines, which had also been incorporated into Belgian national law.

135    The Commission disputes those arguments. It contends inter alia that it examined the existence of a selective advantage by reference to the reference framework deriving from national law and not by reference to the arm’s length principle. It maintains that it is clear from the contested decision that the existence of an advantage was assessed by comparison with the tax burden that would normally have been borne by SMBV under the general Netherlands system of corporate income tax.

136    Under the present plea, in essence, the Kingdom of the Netherlands and Starbucks therefore criticise the Commission for having identified an arm’s length principle specific to EU law in breach of the fiscal autonomy of the Member States and for having examined the APA only by reference to that principle without taking Netherlands law into account.

137    First and foremost, it must be noted that, as is apparent in particular from recitals 252, 267 and 408 of the contested decision, the examination in the light of the arm’s length principle as described by the Commission in the contested decision forms part of its principal analysis of the selective advantage. As has been set out in paragraph 35 above, that analysis entails examining whether the APA derogates from the general Netherlands corporate income tax system. It must be noted in that regard that the Commission had previously indicated, in recitals 232 to 244 of the contested decision, that the objective of the general Netherlands corporate income tax system was to tax the profits of all companies resident in the Netherlands, whether or not integrated, and that both types of company are in a similar factual and legal situation in the light of that objective.

138    As regards the definition of the arm’s length principle, the Commission asserted, in recitals 258 and 261 of the contested decision, that, according to that principle, intra-group transactions should be remunerated as if they had been agreed to by independent companies. It added, in recital 262 of the contested decision, that the purpose of that principle was to ensure that intra-group transactions were treated for tax purposes by reference to the amount of profit that would have arisen if the same transactions had been executed by independent companies. The Commission moreover argued during the hearing that the arm’s length principle was, in its view, a tool for assessing the price level of intra-group transactions, and the Court took formal note of that in the minutes of the hearing.

139    With regard to the legal nature of the arm’s length principle, the Commission considered, in recital 264 of the contested decision, that the arm’s length principle necessarily formed part of its assessment, under Article 107 TFEU, of tax measures granted to group companies, irrespective of whether the Member State had incorporated that principle into its national legal system. It stated that the arm’s length principle which it was applying was a general principle of equal treatment in taxation, which fell within the application of Article 107 TFEU. The French-language version of the contested decision mentions in that connection a ‘principe de traitement équitable’ (principle of equitable treatment) which is a translation error of the expression ‘principle of equal treatment’. The Commission based that statement on the judgment of 22 June 2006, Belgium and Forum 187 v Commission (C‑182/03 and C‑217/03, EU:C:2006:416), concerning the tax regime for coordination centres in Belgium, in which the Court of Justice had held that the method for determining taxable income under that regime conferred a selective advantage on those centres. Specifically, the Commission referred to paragraph 96 of that judgment, in which the Court of Justice held that the method for determining the taxable income of the centres ‘[meant] that the transfer prices [did] not resemble those which [were] charged in conditions of free competition’.

140    As regards the application of the arm’s length principle, in recital 263 of the contested decision, the Commission indicated that, to assess whether the Kingdom of the Netherlands had granted a selective advantage to SMBV, it accordingly had to verify whether the methodology accepted by the Netherlands tax administration through the APA for the purposes of determining SMBV’s taxable profits in the Netherlands departed from a methodology that resulted in a reliable approximation of a market-based outcome and, therefore, from the arm’s length principle. It added, in recital 264 of the contested decision, that the arm’s length principle was used to establish whether the taxable profit of a group company for corporate income tax purposes had been determined on the basis of a methodology that approximated market conditions, so that that company was not treated favourably under the general corporate income tax system as compared to non-integrated companies whose taxable profit was determined by the market.

141    It must therefore be examined whether the Commission was entitled to analyse the measure at issue in the light of the arm’s length principle as described in the contested decision, summarised in recitals 138 to 140 above, which consists in verifying whether intra-group transactions are remunerated as if they had been negotiated under market conditions.

142    According to settled case-law, while direct taxation, as EU law currently stands, falls within the competence of the Member States, they must nonetheless exercise that competence consistently with EU law (see judgment of 12 July 2012, Commission v Spain, C‑269/09, EU:C:2012:439, paragraph 47 and the case-law cited). Thus, intervention by the Member States in matters of direct taxation, even if it relates to issues that have not been harmonised in the European Union, is not excluded from the scope of the rules on the monitoring of State aid.

143    It follows that the Commission can classify a tax measure as State aid provided that the conditions for classification are met (see, to that effect, judgments of 2 July 1974, Italy v Commission, 173/73, EU:C:1974:71, paragraph 28, and of 22 June 2006, Belgium and Forum 187 v Commission (C‑182/03 and C‑217/03, EU:C:2006:416, paragraph 81). Member States must exercise their competence in respect of taxation in accordance with EU law (judgment of 3 June 2010, Commission v Spain, C‑487/08, EU:C:2010:310, paragraph 37). Consequently, they must refrain from taking, in that context, any measure likely to constitute State aid that is incompatible with the internal market.

144    As regards the condition that the measure at issue must grant an economic advantage, it should be borne in mind that, according to settled case-law, measures which, whatever their form, are likely directly or indirectly to favour certain undertakings or are to be regarded as an economic advantage which the recipient undertaking would not have obtained under normal market conditions are regarded as State aid (see judgment of 2 September 2010, Commission v Deutsche Post, C‑399/08 P, EU:C:2010:481, paragraph 40 and the case-law cited; judgment of 9 October 2014, Ministerio de Defensa et Navantia, C‑522/13, EU:C:2014:2262, paragraph 21).

145    Specifically, a measure by which the public authorities grant certain undertakings favourable tax treatment which, although not involving the transfer of State resources, places the recipients in a more favourable financial position than that of other taxpayers amounts to State aid within the meaning of Article 107(1) TFEU (judgment of 15 March 1994, Banco Exterior de España, C‑387/92, EU:C:1994:100, paragraph 14; see, also, judgment of 8 September 2011, Paint Graphos and Others, C‑78/08 to C‑80/08, EU:C:2011:550, paragraph 46 and the case-law cited).

146    In the case of tax measures, the very existence of an advantage may be established only when compared with ‘normal’ taxation (judgment of 6 September 2006, Portugal v Commission, C‑88/03, EU:C:2006:511, paragraph 56). Accordingly, such a measure confers an economic advantage on its recipient if it mitigates the burdens normally included in the budget of an undertaking and which, accordingly, without being subsidies in the strict meaning of the word, are similar in character and have the same effect (judgment of 9 October 2014, Ministerio de Defensa and Navantia, C‑522/13, EU:C:2014:2262, paragraph 22).

147    Consequently, in order to determine whether there is a tax advantage, the position of the recipient as a result of the application of the measure at issue must be compared with his position in the absence of the measure at issue (see, to that effect, judgment of 26 April 2018, Cellnex Telecom and Telecom Castilla-La Mancha v Commission, C‑91/17 P and C‑92/17 P, not published, EU:C:2018:284, paragraph 114), and under the normal rules of taxation.

148    In the context of determining the fiscal position of an integrated company which is part of a group of undertakings, it must be noted at the outset that the pricing of intra-group transactions carried out by that company is not determined under market conditions. That pricing is agreed to by companies belonging to the same group, and is therefore not subject to market forces.

149    Where national tax law does not make a distinction between integrated undertakings and stand-alone undertakings for the purposes of their liability to corporate income tax, that law is intended to tax the profit arising from the economic activity of such an integrated undertaking as though it had arisen from transactions carried out at market prices. In those circumstances, it must be held that, when examining, pursuant to the power conferred on it by Article 107(1) TFEU, a fiscal measure granted to such an integrated company, the Commission may compare the fiscal burden of such an integrated undertaking resulting from the application of that fiscal measure with the fiscal burden resulting from the application of the normal rules of taxation under national law of an undertaking, placed in a comparable factual situation, carrying on its activities under market conditions.

150    Furthermore, and as the Commission correctly stated in the contested decision, those findings are supported by the judgment of 22 June 2006, Belgium and Forum 187 v Commission (C‑182/03 and C‑217/03, EU:C:2006:416), concerning Belgian tax law, which provided for integrated companies and stand-alone companies to be treated on equal terms. The Court of Justice recognised in paragraph 95 of that judgment the need to compare a regime of derogating aid with the ‘ordinary tax system, based on the difference between profits and outgoings of an undertaking carrying on its activities in conditions of free competition’.

151    In that context, although, through that fiscal measure granted to an integrated company, national authorities have accepted a certain level of pricing for an intra-group transaction, Article 107(1) TFEU allows the Commission to check whether that pricing corresponds to pricing under market conditions, in order to determine whether there is, as a result, any mitigation of the burdens normally included in the budget of the undertaking concerned, thus conferring on that undertaking an advantage within the meaning of that article. The arm’s length principle, as described by the Commission in the contested decision, is thus a tool for making that determination in the exercise of the Commission’s powers under Article 107(1) TFEU. The Commission also stated, correctly, in recital 261 of the contested decision, that the arm’s length principle was a ‘benchmark’ for establishing whether an integrated company was receiving, pursuant to a tax measure determining its transfer pricing, an advantage within the meaning of Article 107(1) TFEU.

152    It should also be stated that when the Commission uses that tool to check whether the taxable profit of an integrated undertaking pursuant to a tax measure corresponds to a reliable approximation of a taxable profit generated under market conditions, the Commission can identify an advantage within the meaning of Article 107(1) TFEU only if the variation between the two comparables goes beyond the inaccuracies inherent in the methodology used to obtain that approximation.

153    In the present case, the APA concerns the determination of SMBV’s taxable profits under the CIT the objective of which, irrespective of whether the normal rules of taxation are to be broadly or narrowly defined, is to tax integrated and stand-alone undertakings in the Netherlands in the same way with regard to corporate income tax. The Commission was therefore in a position to verify whether SMBV’s taxable profit pursuant to the APA was lower than its tax burden in the absence of the APA and under the normal rules of taxation in Netherlands law. Given that SMBV is an integrated undertaking and that the CIT is intended to tax the profit resulting from the economic activity of such an integrated undertaking as if it had resulted from transactions carried out at market prices, it is necessary, in examining the APA, to compare SMBV’s taxable profit as a result of the application of the APA with the position, as it would be if the normal tax rules under Netherlands law were applied, of an undertaking in a factually comparable situation, carrying on its activities in conditions of free competition. Against that background, although the APA accepted a certain level of pricing for intra-group transactions, it is necessary to check whether that pricing corresponds to prices that would have been charged under market conditions.

154    In that context, it must be stated that, with regard to the examination as to whether an integrated undertaking has obtained an advantage within the meaning of Article 107(1) TFEU, the Commission cannot be criticised for having used a methodology for determining transfer pricing that it considers appropriate in this instance in order to examine the level of transfer pricing for a transaction or for several closely connected transactions that is part of the contested measure. The Commission is nevertheless required to justify its choice of methodology.

155    Even though the Commission correctly observed that it cannot be formally bound by the OECD Guidelines, the fact remains that those guidelines are based on important work carried out by groups of renowned experts, that they reflect the international consensus achieved with regard to transfer pricing and that they thus have a certain practical significance in the interpretation of issues relating to transfer pricing, as the Commission acknowledged in recital 66 of the contested decision.

156    Consequently, the Commission correctly concluded that it was entitled to examine, in the context of its analysis under Article 107(1) TFEU, whether intra-group transactions were remunerated as though they had been negotiated under market conditions. That finding is not called into question by the other arguments of the Kingdom of the Netherlands and of Starbucks.

157    First, as regards the argument of the Kingdom of the Netherlands that the Commission failed to explain the content of the arm’s length principle as defined in the contested decision, it is sufficient to recall that the contested decision highlights it as a useful tool that can be used to verify that intra-group transactions are remunerated as if they had been negotiated between stand-alone undertakings (see paragraph 138 above). That argument must therefore be rejected.

158    Second, to the extent that the Kingdom of the Netherlands and Starbucks maintain that the arm’s length principle as described by the Commission in the contested decision would permit it alone to prescribe the taxable profit of an undertaking and that it would have the effect of disguised direct tax harmonisation in contravention with Member States’ fiscal autonomy, that argument must be rejected.  

159    While, in the absence of EU rules governing the matter, it falls within the competence of the Member States to designate bases of assessment and to spread the tax burden across the different factors of production and economic sectors (see, to that effect, judgment of 15 November 2011, Commission and Spain v Government of Gibraltar and United Kingdom, C‑106/09 P and C‑107/09 P, EU:C:2011:732, paragraph 97), it does not mean that every tax measure, which affects, inter alia, the basis of assessment taken into account by the tax authorities, will escape the application of Article 107 TFEU (see, to that effect, judgment of 15 November 2011, Commission and Spain v Government of Gibraltar and United Kingdom, C‑106/09 P and C‑107/09 P, EU:C:2011:732, paragraph 104). It follows that the Commission does not have, at this stage of the development of EU law, competence to allow it to define in an autonomous manner the ‘normal’ taxation of an integrated undertaking, by disregarding national tax rules. However, although ‘normal’ taxation is defined by national tax rules and although the very existence of an advantage must be established by reference to them, the fact remains that, if those national rules provide that stand-alone companies and integrated companies must be dealt with under the same conditions, Article 107(1) TFEU allows the Commission to verify whether the price level of intra-group transactions, accepted by the national authorities for determining the tax base of an integrated undertaking, corresponds to a price level of a transaction which has been negotiated in market conditions.

160    Consequently, when the Commission examines whether the method validated in a national tax measure leads to an outcome established in conformity with the arm’s length principle as has been defined in paragraph 137 above, it is not exceeding its competences.

161    Third, as regards the argument of the Kingdom of the Netherlands that the Commission has failed to provide any legal basis for its arm’s length principle, it must be pointed out that, in recitals 264 and 265 of the contested decision, the Commission stated that the arm’s length principle as described in the contested decision existed independently of the incorporation of that principle into the national legal system. It also made clear that it had not examined whether the APA complied with the arm’s length principle laid down in Article 8b of the CIT or in the transfer pricing decree, which incorporate the arm’s length principle into Netherlands law. The Commission also asserted that the arm’s length principle which it applied was distinct from that enshrined in Article 9 of the OECD Model Tax Convention on Income and on Capital.

162    However, the Commission also made clear, in recital 264 of the contested decision, that the arm’s length principle necessarily formed an integral part of the examination, under Article 107(1) TFEU, of tax measures granted to group companies and that the arm’s length principle was a general principle of equal treatment in taxation, which fell within the application of Article 107 TFEU.

163    It is therefore apparent from the contested decision that the arm’s length principle, as described by the Commission, is a tool which it used, correctly, in the context of the examination carried out under Article 107(1) TFEU.

164    It is true that, at the hearing, the Commission inter alia stated that the arm’s length principle as described in the contested decision did not fall within EU law or international law, but that it was inherent in the ordinary system of taxation as provided for by national law. Thus, according to the Commission, if a Member State chooses, in the context of its national tax system, the approach of the separate legal entity, according to which tax law is concerned with legal entities, and not with economic entities, the arm’s length principle is necessarily a corollary of that approach, which is binding in the Member State concerned, independently of whether the arm’s length principle has, expressly or impliedly, been incorporated into national law.

165    In that regard, the Kingdom of the Netherlands and Starbucks indicated at the hearing that, by those assertions, the Commission seemed to be changing its stance on the arm’s length principle as described in the contested decision. However, on the assumption that the interpretation put forward by the Kingdom of the Netherlands and Starbucks is found to be correct, it must be stated, in any event, that the Commission cannot change the legal basis of the arm’s length principle, as set out in the contested decision, at the hearing stage (see, to that effect, judgment of 25 June 1998, British Airways and Others v Commission, T‑371/94 and T‑394/94, EU:T:1998:140, paragraph 116).

166    In all events, it must be noted that the clarification provided at the hearing does not call into question the finding in paragraph 156 above that it is apparent from the contested decision that the arm’s length principle is being applied in the context of the examination under Article 107(1) TFEU. It is, moreover, apparent from all of the written submissions of the Kingdom of the Netherlands and Starbucks that they understood the contested decision to mean that the arm’s length principle as described by the Commission in the contested decision was being applied in the context of the examination of a national tax measure under Article 107(1) TFEU.

167    Fourth, the Kingdom of the Netherlands and Ireland submit, in essence, that the Commission wrongly asserted, in the contested decision, that there was a general principle of equal treatment in taxation  on the profits of integrated and non-integrated undertakings.  

168    It is true that the Commission indicated, in recital 264 of the contested decision, that the arm’s length principle was a general principle of equal treatment in taxation, which fell within the scope of Article 107(1) TFEU. However, that wording must not be taken out of context and cannot be interpreted as meaning that the Commission asserted that there was a general principle of equal treatment in relation to tax inherent in Article 107(1) TFEU, which would give that article too broad a scope.

169    In any event, it is implicitly but necessarily evident from recitals 258 to 267 of the contested decision, and in particular from recitals 262 and 265 of that decision, that the arm’s length principle as described by the Commission in the contested decision was perceived by the Commission only as a tool enabling it to check that intra-group transactions are remunerated as though they had been negotiated between independent companies. The argument of the Kingdom of the Netherlands and of Ireland does not alter the finding in paragraphs 147 to 156 above that the Commission was entitled to examine, in its analysis under Article 107(1) TFEU, whether intra-group transactions were remunerated as though they had been negotiated under market conditions.

170    Accordingly, the Court must reject the argument of the Kingdom of the Netherlands and of Ireland in that respect.

171    Fifth, the Kingdom of the Netherlands and Starbucks argue that the Commission made an assessment in the light of the arm’s length principle, but that it did not examine whether there was an advantage by reference to national tax law. In that regard, it must be noted that it is clear from recitals 267, 341, 415 and 416 of the contested decision that the Commission carried out its examination of whether there was an advantage by reference to the general Netherlands corporate income tax system. It will be necessary to verify whether that examination was vitiated by error in respect of the specific examination of the six lines of reasoning and, if necessary, in respect of the limited reference system.

172    On the basis of the foregoing, it is necessary to reject the second plea in Case T‑760/15 and the second part of the first plea in Case T‑636/16, according to which the Commission committed an error in identifying an arm’s length principle as a criterion for assessing the existence of State aid. It is therefore in the light of the considerations set out in paragraphs 137 to 170 above that the merits of each line of reasoning set out in the contested decision should be analysed (see paragraphs 53 and 54 above).

D.      Dispute as to the principal reasoning regarding the existence of a tax advantage in favour of SMBV (recitals 275 to 361 of the contested decision)

1.      Choice of the TNMM in the case at hand and the lack of examination of the intra-group transaction for which the APA had in reality been requested (first line of reasoning)

173    The first part of the third plea in Case T‑760/15 as well as the third part of the first plea and the first and second parts of the second plea in Case T‑636/16 concern the Commission’s analysis, conducted in the contested decision, according to which, first, the transfer pricing report had not identified or analysed the transaction for which a price was effectively determined in the APA, namely, the royalty, and, second, the CUP method should have been given priority, in order to determine the level of the royalty, over the TNMM, in order to determine the net profit of SMBV’s production and distribution activities. Those two complaints made against the APA, as a question of principle, precede the Commission’s concrete analysis, according to which the level of the royalty paid by SMBV to Alki should have been zero and the level of the prices of green coffee beans from 2011 onwards was too high, questions which will be examined in paragraphs 217 to 404 below.

174    By the first part of the third plea in Case T‑760/15, the Kingdom of the Netherlands disputes the Commission’s argument that the TNMM does not allow the conformity of the royalty with the arm’s length principal to be examined and assessed distinctly. It maintains that that argument is erroneous and is not such as to cast into doubt the relevance of the choice of the TNMM in the case at hand.

175    First, the Kingdom of the Netherlands submits that it is apparent from the contested decision that the Commission made the transfer pricing method an end in itself, when it was merely a means of determining the conformity of the conditions of intra-group transactions with the arm’s length principle. If the method chosen led to an arm’s length outcome, the Commission cannot cast it into doubt on the ground that the royalty and the mark-up applied to the resale price of green coffee beans were not examined individually. In addition, according to the Kingdom of the Netherlands, the Commission could not take the view that the OECD Guidelines prioritised the use of traditional methods, such as the CUP method, over transactional methods, such as the TNMM. However, it is apparent from point 2 of the transfer pricing decree and from paragraph 4.9 of the 1995 version of the OECD Guidelines that the taxpayer is free to choose a transfer pricing method, provided that the method chosen leads to an arm’s length outcome.

176    Second, the Kingdom of the Netherlands considers that, unlike what the Commission maintains in the contested decision, the only transactions concerned by the APA are the roasting of coffee beans and the provision of logistics and administrative services on behalf of Alki. The purpose of the APA is not to determine whether the royalty is in conformity with the arm’s length principle. Moreover, the Kingdom of the Netherlands observes that, in the contested decision, the Commission did not explain the reasons which led it to assume that the APA had been requested and concluded for a licensing arrangement and for the royalty.

177    Third, the Kingdom of the Netherlands submits that the TNMM was the method the best suited to the case at hand. According to the Kingdom of the Netherlands, the main reason for choosing that method was the lack of similar unrelated external or internal transactions, necessary for the purposes of applying the CUP method, with which it would be possible to compare the transactions between Alki and SMBV and, therefore, the remuneration that was associated with them. However, according to the Kingdom of the Netherlands, the TNMM could be applied in SMBV’s case since information was indeed available on the operational profit of the undertakings that were comparable to it in terms of function, namely, coffee bean roasting.

178    By the third part of the first plea and by the second part of the second plea in Case T‑636/16, Starbucks claims that the TNMM was the most appropriate method for calculating transfer pricing in the case at hand and that the Commission could not reject that method for the reasons it provided in the contested decision. According to Starbucks, as the TNMM was correctly applied to calculate an arm’s length remuneration for SMBV, there is no need to assess separately the royalty payments by SMBV, given that those payments could not have impacted SMBV’s remuneration, as calculated on the basis of the TNMM.

179    More specifically, first, Starbucks argues that Commission’s assertion that there is a strict rule favouring use of the CUP method has no basis in Netherlands tax law or in the OECD Guidelines. In addition, Starbucks is of the view that the use of a different transfer pricing method does not in itself result in a lower tax liability, as all methods attempt to achieve a profit allocation reflecting arm’s length transfer prices. Merely alleging a methodological error does not equate to proving that an advantage exists.

180    Second, according to Starbucks, the Commission compared the price of green coffee beans and the royalty with ‘controlled’ transactions (intra-group) in disregard of Netherlands tax law. Starbucks chose the TNMM since the roasting contract combined various intra-group transactions by which routine, low-risk activities were entrusted to SMBV, namely, coffee roasting and conditioning activities as well as administrative and logistical support activities.

181    Third, Starbucks submits that the contested decision contains no argument asserting that the mere absence of identification and analysis of SMBV’s intra-group transactions is sufficient to prove the existence of an advantage and that that argument was raised for the first time in the defence in Case T‑636/16 and is thus inadmissible.

182    The Commission disputes those arguments.

183    First, the Commission explains that nowhere in the contested decision does it impose a strict rule concerning the application of the CUP method rather than another method for determining transfer pricing, but that the most reliable method should be chosen based on the circumstances of the case. It first of all established that the APA had been requested and granted for setting the price of the IP licensing arrangement between SMBV and Alki and then concluded that, since a comparable price for the price of that transaction could be determined, the use of the CUP method was preferable, in the case at hand, to the TNMM. The Commission contends that, in doing so, it relied on the guidance set out in the OECD Guidelines.

184    Second, the Commission maintains that the method approved in the APA for determining the royalty amount, by which SMBV pays Alki the residual profit of the sale of roasted beans and of non-coffee products, cannot result in an arm’s length outcome. According to the Commission, as there were comparable transactions allowing the value of the royalty to be assessed, the tax advisor should have used the CUP method to define the royalty price owed by SMBV to Alki, which was the transaction for which the APA was actually requested and granted. In addition, the prices invoiced by SCTC to SMBV for green coffee beans should also have been subject to a transfer pricing analysis. The Commission argues that, contrary to what the Kingdom of the Netherlands and Starbucks maintain, setting the price of individual transactions is the very essence of that principle. Thus, the establishment and analysis of controlled and uncontrolled transactions is a necessary first step of the evaluation of the arm’s length nature of transfer pricing.  

185    Third, the Commission contends that the Kingdom of the Netherlands has not proved that the TNMM was more appropriate, in the case at hand, than the CUP method. The Commission contends, first of all, that the 1995 version of the OECD Guidelines, which was in force at the time the APA was concluded, and the 2010 version, give preference to traditional transaction methods, such as the CUP method, over transactional profit methods. According to the Commission, the particular circumstances justifying the preference of the TNMM over the CUP method are not present in the case at hand.  

(a)    Preliminary observations

186    As a preliminary point, it must be noted that the APA, as set out in paragraphs 12 to 16 above, calls for two important clarifications.

187    First, it is undisputed among the parties that the method applied in the APA is indeed the TNMM. In that regard, the Kingdom of the Netherlands has specified in the application in Case T‑760/15 and stated at the hearing that the reference to the cost plus method in the APA constituted a non-technical use of that expression.

188    Second, in their answers to measures of organisation of procedure and during the hearing, the parties specified that, in reality and contrary to what is set out in the APA, the royalty to be paid to Alki was not set on the basis of the difference between the operating profit made in connection with the production and distribution function, before royalty-related expenses, and SMBV’s remuneration, but on the basis of the difference between SMBV’s total revenue, on the one hand, and SMBV’s cost base, increased by SMBV’s remuneration, on the other.

189    In addition, it should be recalled that the Commission set out its first line of reasoning regarding the existence of a selective advantage in recitals 272 and 275 to 285 of the contested decision, primarily in Section 9.2.3.2, entitled ‘The transfer pricing report fails to examine the intra-group transaction for which the … APA was effectively requested and granted’.

190    First, in recitals 272, 276 to 279 and 285 of the contested decision, the Commission considered, in essence, that the transfer pricing report, accepted by the Netherlands tax authorities upon the conclusion of the APA with SMBV, failed to identify or analyse SMBV’s controlled and uncontrolled transactions, which was a necessary first step of the evaluation of the arm’s length nature of transfer pricing. More specifically, it considered that the royalty payment for the roasting IP licence between Alki and SMBV was the transaction for which the APA had effectively been requested.

191    Second, in recitals 280 to 284 of the contested decision, the Commission asserted, in essence, that an approach consisting in determining transfer prices for each transaction taken individually was to be prioritised over an approach consisting in determining transfer prices for a function as a whole. In other words, the Commission took the view that the CUP method was to be given priority over transactional profit methods, such as the TNMM. In recital 285 of the contested decision, the Commission maintained that, since the analysis of an arm’s length remuneration for SMBV had been conducted in the transfer pricing report starting from an incorrect point of departure, that remuneration was, by necessity, improperly estimated by using the TNMM. Moreover, it took the view that the transfer pricing report, in order to establish transfer pricing in the case at hand, should have made use of reliable comparisons with available information on transactions between unrelated parties which owned Starbucks at the time the request for the APA was made.

192    The Commission moreover confirmed in its submissions that its first line of reasoning consisted in criticising the use of the TNMM to determine the net profit of SMBV’s production and distribution activities rather than the CUP method to determine the level of the royalty. It maintained that the validity of its first line of reasoning was not dependent on the conclusion that the arm’s length value of the royalty was zero. The fact that the transfer pricing report failed to identify or analyse SMBV’s controlled and uncontrolled transactions means that a necessary first step in assessing the arm’s length nature of commercial conditions applicable between related parties for transfer pricing purposes was not taken.  

193    Without it being necessary, at this stage, to analyse Starbucks’ complaint that the contested decision contains no argument asserting that the mere absence of identification and analysis of SMBV’s intra-group transactions is sufficient to prove the existence of an advantage, an argument raised for the first time in the defence in Case T‑636/16 and thus inadmissible, it is appropriate to examine whether the criticisms formulated by the Commission as part of its first line of reasoning justified the finding that the APA conferred an advantage on SMBV on the ground that the very choice of the transfer pricing method, proposed in transfer pricing report, did not result in a reliable approximation of a market-based outcome, in line with the arm’s length principle.

(b)    Burden of proof

194    It must be borne in mind that, in its review of State aid, the Commission must, in principle, provide proof in the contested decision of the existence of the aid (see, to that effect, judgments of 12 September 2007, Olympiaki Aeroporia Ypiresies v Commission, T‑68/03, EU:T:2007:253, paragraph 34, and of 25 June 2015, SACE and Sace BT v Commission, T‑305/13, EU:T:2015:435, paragraph 95). In that context, the Commission is required to conduct a diligent and impartial examination of the measures at issue, so that it has at its disposal, when adopting a final decision establishing the existence and, as the case may be, the incompatibility or unlawfulness of the aid, the most complete and reliable information possible (see, to that effect, judgments of 2 September 2010, Commission v Scott, C‑290/07 P, EU:C:2010:480, paragraph 90, and of 3 April 2014, France v Commission, C‑559/12 P, EU:C:2014:217, paragraph 63).

195    By contrast, it is for the Member State which has made a distinction between undertakings to show that it is actually justified by the nature and the general scheme of the system in question. The concept of State aid does not refer to State measures which differentiate between undertakings and which are, therefore, prima facie selective where that differentiation arises from the nature or the general scheme of the system of which they form part (see, to that effect, judgment of 21 June 2012, BNP Paribas and BNL v Commission, C‑452/10 P, EU:C:2012:366, paragraphs 120 and 121 and the case-law cited).

196    It follows that it was for the Commission to show, in the contested decision, that the requirements for a finding of State aid, within the meaning of Article 107(1) TFEU, were met. In that regard, it must be held that, while it is common ground that the Member State has a margin of appreciation in the approval of transfer pricing, that margin of appreciation cannot lead to the Commission being deprived of its power to check that the transfer pricing in question does not lead to the grant of a selective advantage within the meaning of Article 107(1) TFEU. In that context, the Commission must take into account the fact that the arm’s length principle allows it to verify whether the transfer pricing accepted by a Member State corresponds to a reliable approximation of a market-based outcome and whether any variation that may be identified in the course of that examination does not go beyond the inaccuracies inherent in the methodology used to obtain that approximation.

(c)    Intensity of review to be conducted by the Court

197    With regard to the intensity of the review to be conducted by the Court in the present case, it should be noted that, as is clear from Article 263 TFEU, the object of an action for annulment is to review the legality of the acts adopted by the EU institutions named therein. Consequently, the analysis of the pleas in law raised in such an action has neither the object nor the effect of replacing a full investigation of the case in the context of an administrative procedure (see, to that effect, judgment of 2 September 2010, Commission v Deutsche Post, C‑399/08 P, EU:C:2010:481, paragraph 84).

198    In the field of State aid, it must be recalled that State aid, as defined in the FEU Treaty, is a legal concept which must be interpreted on the basis of objective factors. For that reason, the Courts of the European Union must, in principle, having regard both to the specific features of the case before them and to the technical or complex nature of the Commission’s assessments, carry out a comprehensive review as to whether a measure falls within the scope of Article 107(1) TFEU (judgments of 4 September 2014, SNCM and France v Corsica Ferries France, C‑533/12 P and C‑536/12 P, EU:C:2014:2142, paragraph 15, and of 30 November 2016, Commission v France and Orange, C‑486/15 P, EU:C:2016:912, paragraph 87).

199    As to whether a method for determining transfer pricing of an integrated company complies with the arm’s length principle, it should be borne in mind that, as has already been indicated above, when using that tool in carrying out its assessment under Article 107(1) TFEU, the Commission must take into account its approximate nature. The purpose of the Court’s review is therefore to verify whether the errors identified in the contested decision, and on the basis of which the Commission found there to be an advantage, go beyond the inaccuracies inherent in the application of a method designed to obtain a reliable approximation of a market-based outcome.

(d)    Failure to identify and analyse the royalty paid by SMBV to Alki in the APA

200    With regard to the finding made by the Commission, according to which the transfer pricing report did not identify or analyse the transaction for which a price was actually determined in the APA, it must be noted that, in recital 276 of the contested decision, the Commission explained that SMBV’s taxable profit was lower than that actually recorded on account of the acceptance by the Netherlands tax authorities because the actual level of profits generated by SMBV in the Netherlands was to be reduced, for corporate tax purposes, by the amount of the royalty for the roasting IP. In recitals 277 and 278 of the contested decision, the Commission inferred that the royalty for the roasting IP was the transaction for which the APA had effectively been requested and that the methodology for determining the level of that royalty as an adjustment variable was the transaction effectively being transfer priced by the APA. It also considered that the price of coffee beans should have undergone an analysis.

201    In that regard, first, it is sufficient to hold that mere non-compliance with methodological requirements does not necessarily lead to a reduction of the tax burden. It is further necessary for the Commission to demonstrate that the methodological errors that it identified in the APA do not allow a reliable approximation of an arm’s length outcome to be reached and that they led to a reduction in the taxable profit compared with the tax burden resulting from the application of normal taxation rules under national law on an undertaking placed in a comparable factual situation to SMBV and carrying out its activities under market conditions. Thus, the mere finding of a methodological error does not in itself suffice, in principle, to demonstrate that the APA conferred an advantage on SMBV and, thus, to establish that there is State aid within the meaning of Article 107 TFEU.

202    Second, it should be recalled that the various methods for setting transfer prices – whether it be the CUP method or the TNMM – endeavour to attain profit levels reflecting arm’s length transfer prices and that it cannot be concluded, as a rule, that one method does not allow a reliable approximation of an arm’s length outcome to be reached.

203    It follows that the mere fact that, according to the Commission, neither the transfer pricing report nor the APA identified the royalty as the transaction for which a transfer price was in reality determined in the APA and that they did not analyse whether the royalty was in conformity with the arm’s length principle does not suffice to demonstrate that that royalty was not actually in conformity with the arm’s length principle. That finding alone therefore did not prove that the APA conferred an advantage on SMBV.

204    In addition, it must be pointed out that the Commission’s argument that the transfer pricing report did not identify or analyse the royalty as the transaction for which a price was actually determined in the APA is based on the assertion that, in the transfer pricing report, the payment of a royalty is not regarded as the adjustment variable of the structure of that suggested remuneration. In that regard, it should be noted that the transfer pricing report in no way disregards the licensing arrangement concluded between SMBV and Alki. That arrangement is mentioned both in the description of the Starbucks group’s activities in the EMEA region and the Netherlands and in the graphic representation of the EMEA region’s transactions. Therefore, the Starbucks group’s tax advisor did indeed take into account those transactions when he proposed SMBV’s remuneration.

205    Accordingly, it is necessary to uphold the complaint of the Kingdom of the Netherlands and of Starbucks according to which the Commission wrongly found that the absence of separate analysis of the royalty in transfer pricing report and in the APA conferred an advantage on SMBV.

(e)    Necessity of prioritising the CUP method over the TNMM

206    So far as concerns the Commission’s position that the CUP method should have been given priority over the TNMM, since that first method was applicable in the case at hand, first, it should be noted that, in the present case, the APA accepted the use of the TNMM in order to determine the operating margin for SMBV’s production and distribution activities. The APA, however, accepted that the royalty was determined as, in essence, the difference between the operating profit generated in connection with the production and distribution function and the operating margin. It follows that the APA does not directly provide for the use of a transfer pricing method to calculate the level of the royalty, which is defined as a purely residual value.

207    It is true that it follows from the considerations set out in paragraphs 148 to 156 above that, since the royalty was an intra-group transaction the level of which was determined in the APA, the Commission was entitled to examine as part of its analysis under Article 107(1) TFEU, using a transfer pricing method which it regarded as appropriate in the case at hand, whether the royalty amount was determined as if it had been negotiated in market conditions.

208    However, with regard to the contested decision, while the Commission contends that the CUP method should have been given priority over the TNMM in order to be able to determine the arm’s length level of the royalty, it fails to consider that, actually, in the APA, the level of the royalty was not calculated using a transfer pricing method, in particular the TNMM. On the contrary, the TNMM was used in the APA to determine SMBV’s remuneration for the production and distribution activities. Thus, the Commission’s position essentially amounts to criticising the fact that the TNMM was used to determine SMBV’s remuneration for the production and distribution activities instead of the CUP method, which should in the Commission’s view have been used to calculate the level of the royalty.

209    In that connection, it must be held that the two methods are applied to calculate the price level of different intra-group transactions. However, while the Commission argues that the OECD Guidelines expressed a certain preference towards using traditional methods, such as the CUP method, it cannot mandate the examination of another transaction than that for which the APA determined a transfer price on the basis of the TNMM solely because, for that other transaction, a transfer price should have been determined on the basis of the CUP method. The rule invoked by the Commission merely allows for a choice of the transfer pricing method appropriate for the same type of transaction or closely linked transactions. Choosing the transfer pricing method is not an end in itself, but is done with a view to the intra-group transaction for which the arm’s length level must be determined and not the other way around.

210    Second, it must be recalled that, as has been set out in paragraphs 146 and 147 above, a tax measure confers an economic advantage where it leads to a reduction of the tax burden compared to what it would normally have had to bear in the absence of that measure.

211    As has been held in paragraph 201 above, mere non-compliance with methodological requirements does not necessarily lead to a reduction of the tax burden. It follows that the mere finding by the Commission of errors in the choice or application of the transfer pricing method does not, in principle, suffice to demonstrate the existence of an advantage.

212    However, in recitals 275 to 285 of the contested decision, the Commission invokes no element grounding the conclusion – without a comparison being carried out with the result that would have been obtained using the CUP method – that the choice of the TNMM necessarily leads to a result that is too low. In that context, the Commission merely contends, in recital 284 of the contested decision, that the taxpayer was under an obligation to verify whether the transfer pricing method chosen by the latter led to a reliable approximation of an arm’s length price, before the tax authorities can accept an APA request based on that method.

213    In addition, it must be held that the obligation stated by the Commission falls within the area of tax law and, while breach of it may have consequences in terms of taxation, in the field of State aid, such a breach does not necessarily mean that the method chosen by the taxpayer does not produce a reliable approximation of a market-based outcome, in accordance with the arm’s length principle.

214    For the sake of completeness, it should be recalled (see paragraph 10 above) that point 2 of the transfer pricing decree provides that the Netherlands tax administration must always conduct a transfer pricing audit from the perspective of the method adopted by the taxpayer at the date of the transaction. That rule complies with paragraph 1.68 of the 1995 OECD Guidelines. It follows that the taxpayer is in principle free to choose a transfer pricing method as long as the method selected leads to an arm’s length outcome for the transaction in question. Even though the taxpayer is expected to take into account, when choosing a transfer pricing method, the reliability of that method in the situation in question, that step does not specifically seek to incentivise the taxpayer to assess all methods and then justify how the method he has chosen produces the best result under the conditions present.

215    It follows that, in the case at hand, the Commission was not entitled to find that the CUP method had to be given priority, in principle, over the TNMM.

216    It is therefore necessary to uphold the complaint of the Kingdom of the Netherlands and of Starbucks according to which the Commission wrongly found that the mere choice of the TNMM in the case at hand conferred an advantage on SMBV, without it being necessary to examine Starbucks’ argument disputing the admissibility of certain arguments put forward by the Commission.

2.      Whether the royalty paid by SMBV to Alki should have been zero (second line of reasoning)

217    Under the second part of the third plea in Case T‑760/15, the Kingdom of the Netherlands claims that the Commission is wrong to argue that the remuneration paid by SMBV to Alki should have been zero and that it resulted in an advantage within the meaning of Article 107(1) TFEU. The contracts concluded between the Starbucks group and external roasters and manufacturers of coffee-derived products on which the comparison conducted by the Commission is based cannot be used for a comparison of the contractual arrangements between Alki and SMBV, based on the CUP method. The Kingdom of the Netherlands is of the view that the Commission did not demonstrate that the TNMM had not led to an arm’s length outcome.

218    Under the fourth part of the second plea in Case T‑636/16, Starbucks maintains, in essence, that the Commission’s analysis of the royalty relies almost exclusively on evidence that was not available in April 2008. In addition, contrary to Netherlands law and to the OECD Guidelines, the Commission did not establish an arm’s length range for the royalty, but concluded that it should equal zero. Like the Kingdom of the Netherlands, Starbucks is of the view that all third-party manufacturers, mentioned in the contested decision, which, similar to SMBV, supply Starbucks-branded coffee products to stores or retailers, pay substantial royalties for the use of Starbucks’ roasting IP. No royalties are paid by subcontractors which, unlike SMBV, do not supply such products to customers, but merely provide a roasting service to the Starbucks group. Contrary to what is stated in the contested decision, the value of the roasting IP is generated when Starbucks-branded coffee products are sold to stores and retailers, which are willing to pay premium prices for those products. Additionally, Starbucks contends that, contrary to what the contested decision finds, for the period examined, SMBV’s roasting activities have always been profitable.

219    The Commission disputes those arguments.

220    First, the Commission argues, in essence, that it compared the amount of the royalties under the seven contracts mentioned in recital 300 of the contested decision with that of the royalty in connection with the relationship between SMBV and Alki. In addition, the Commission explains, in Case T‑760/15, that it also relied on the contracts mentioned in recital 303 of the contested decision and, in Case T‑636/16, that that was not, in principle, the case. The Commission adds that it also relied on the agreements between competitors of the Starbucks group and third coffee roasting companies, mentioned in recitals 305 to 308 of the contested decision, to arrive at the conclusion that the arm’s length value of the royalty paid in the context of the relationship between SMBV and Alki had to be zero. It adds that, in recitals 292 to 298 of the contested decision, it explained the reasons why it had taken the view that those transactions constituted a direct point of comparison enabling the amount of the royalty owed by SMBV to Alki in exchange for the roasting IP to be determined.

221    Second, the Commission argues that it does not deny that the roasting IP may represent a value. However, the value of that IP is not captured before the Starbucks-branded coffee products are sold by Starbucks stores to final consumers. According to the Commission, the roasting IP therefore cannot be considered to represent a benefit for SMBV for which a royalty should be paid.

(a)    Preliminary observations

222    It must be recalled that the Commission set out its second line of reasoning in recitals 286 to 341 of the contested decision, under Section 9.2.3.3, entitled ‘The royalty payment to Alki ... resulting from the ... APA is not priced at arm’s length’.

223    As a preliminary point, two observations should be made.

224    First, it must be recalled that it is settled between the parties that the royalty constituted, in principle, a deductible charge under Netherlands tax law. It is moreover undisputed that, since it is a transaction within the Starbucks group, the royalty is an intra-group transaction. It is apparent from paragraphs 147 to 156 above that the level of such a transaction must, for the purposes of the determining SMBV’s corporate income tax, be assessed as if it had been determined in market conditions.

225    Second, it must be noted that, although the Commission considered that the level of the royalty paid by SMBV to Alki should have been zero, it acknowledged in recital 310 of the contested decision that coffee roasting know-how and coffee roasting curves could represent a value. Similarly, in paragraph 126 of its defence in Case T‑636/16, the Commission explains that it does not deny that the roasting IP might represent a value.

226    It follows that the only matter on which the parties are in disagreement is the level the royalty transfer price would have been if it had been set in market conditions.

227    In that regard, it should be recalled that, in recitals 286 to 341 of the contested decision, the Commission maintains that the royalty paid by SMBV to Alki should have been zero. According to the wording of the contested decision, the Commission does not contend that the level of that royalty should have been lower than the level of the royalty accepted by the APA, but that no royalty should have been paid. The Commission itself asserts that it did not estimate a range for the level of the royalty on the ground that it should have been exactly zero (recital 340 of the contested decision).

228    In the contested decision, the Commission based its demonstration according to which the royalty paid by SMBV should have been zero (recital 318 of the contested decision), in essence, on three elements.

229    So far as concerns the first element, the Commission stated that the variable nature of the royalty payment during the period from 2006 to 2014 gave a ‘first indication’ that the level of that payment bore no relation to the value of the roasting IP (recital 289 of the contested decision). As far as the second element is concerned, the Commission maintained that SMBV did not capture the value of the roasting IP in its relationship with Alki (recitals 310 to 313 of the contested decision). With regard to the third element, the Commission explained that the manufacturing agreements concluded by Starbucks with third parties did not require any royalty for the use of the roasting IP (recitals 291 to 309 of the contested decision).

230    In addition, in the contested decision, the Commission rejected the arguments raised by the Kingdom of the Netherlands and Starbucks during the administrative procedure. More specifically, the Commission took the view that the royalty did not represent remuneration for the transfer of business risks (recitals 319 to 332 of the contested decision) and that the amount of the royalty was not justified by the amounts paid by Alki to Starbucks US for the technology under the cost sharing agreement (recitals 333 to 338 of the contested decision).

231    Next, first of all, it is necessary briefly to set out the theory advanced by the Commission in the contested decision concerning SMBV’s royalty-related functions and concerning the relevant normal taxation rules. Those elements are the basis underpinning the analysis of the level of the royalty, conducted by the Commission in the contested decision. Second, it is appropriate to examine Starbucks’ argument that the Commission’s analysis of the royalties could not rely on evidence that was not available in April 2008. Third, it is appropriate to analyse the arguments of the Kingdom of the Netherlands and of Starbucks on the question of who exploited the roasting IP. Fourth, it is necessary to examine whether the Commission was justified in finding, based on a comparison with the royalties provided for in contracts with third parties, that the royalty should have been zero. Fifth, it is appropriate to examine the Commission’s argument, raised at the hearing, according to which, in reality, it maintained in the contested decision that the royalty should have been lower than the level endorsed by the APA.

(b)    SMBV’s royalty-related functions

232    With regard to the functions of SMBV that are relevant for the analysis of the royalty, first of all, it is settled that it roasts green coffee beans which it purchases from SCTC.

233    Next, the Commission maintains in the contested decision, in particular in recitals 49, 96, 137, 313 and 330, as well as in its submissions, that Starbucks stores, both affiliated and unaffiliated, are required to purchase roasted coffee from SMBV and that SMBV is therefore also the vendor of the roasted coffee.  

234    In addition, in the contested decision, the Commission takes the view that, in line with accounting standards, the stocks that SMBV purchases and sells need to appear on its balance sheet, since it is the entity responsible for contracting and invoicing with stores.

235    Last, it is apparent from the contested decision, read as a whole, that the Commission considers it incorrect that SMBV is presented in the transfer pricing report as a low-risk coffee producer. In that regard, in recitals 319 to 332 of the contested decision, the Commission inter alia rejected the arguments of the Kingdom of the Netherlands and of the Starbucks correspondents according to which the contractual arrangements between SMBV and Alki, on which the tax advisor’s report was based, effectively cause a transfer of business risks from SMBV to Alki. In addition, the Commission explained that SMBV assumed commercial risks in its relations with SCTC and Starbucks stores.

236    It follows that, according to the Commission, SMBV is not, with regard to its sales of roasted coffee to Starbucks stores, a toll manufacturer or a supplier, but roasts coffee for its own behalf and acts as vendor. According to the contested decision, ‘toll manufacturing’ is usually understood to mean an arrangement in which a company processes raw materials or semi-finished goods on behalf of another company.

(c)    Normal taxation rules under Netherlands law

237    As has been found in paragraph 146 above, the examination under Article 107(1) TFEU of a tax measure granted to an integrated undertaking means determining, first, the normal taxation rules applicable to the beneficiary of that measure.

238    In recital 232 of the contested decision, the Commission asserted that the Netherlands rules against which the APA must be examined are the rules of the general Netherlands corporate income tax system. Those rules are summarised in paragraphs 3 to 11 and 35 above.

239    In the case at hand, it is settled that the APA was concluded in order to enable SMBV to anticipate the application of corporate income tax rules by determining its taxable profit. It follows that the APA forms part of the general Netherlands corporate income tax system, with the objective of taxing undertakings – integrated or stand-alone – subject to corporate income tax.

240    Therefore, it is in the light of SMBV’s functions as identified in paragraphs 232 to 236 above and the normal taxation rules as identified above that it is appropriate to analyse the question of whether the level of the royalty corresponded to a level that would have been practised under market conditions.

(d)    Use of evidence by the Commission that was not available when the APA was concluded

241    Starbucks claims that, in the contested decision, the Commission relied predominantly on information that was not available in April 2008, when the APA was concluded. More specifically, Starbucks cites the case-law of the EU Courts on the private investor criterion according to which, in order to assess the economic rationality of a certain measure, it is necessary to place oneself in the context of the time at which the financial support measures were adopted, and thus to refrain from any assessment based on events that happened at a later date. According to Starbucks, the same principle is well established under Netherlands tax law and in the OECD Guidelines.

242    The Commission does not dispute that that principle is applicable in the case at hand and merely argues that a considerable number of arguments in support of its conclusion that the APA did not comply with the arm’s length principle relied on information and data available to the Netherlands tax administration at the time the APA was concluded.

243    At the outset, it must be stated that the fact that Netherlands tax law and the OECD Guidelines provide, according to Starbucks, that it is necessary to refrain from any assessment based on events that happened after the adoption of an advance pricing agreement to examine whether it complies with the arm’s length principle has no bearing on the examination in this case of the APA in the light of the conditions of Article 107 TFEU.

244    Starbucks bases its argument on an analogous application of the case-law of the EU Courts according to which, in order to examine whether or not the Member State or the public body concerned has adopted the conduct of a prudent private operator operating in a market economy, it is necessary to place oneself in the context of the period during which the measures at issue were taken in order to assess the economic rationality of the conduct of the Member State or of the public body, and thus to refrain from any assessment based on a later situation (judgment of 25 June 2015, SACE and Sace BT v Commission, T‑305/13, EU:T:2015:435, paragraph 93; see also, to that effect, judgments of 16 May 2002, France v Commission, C‑482/99, EU:C:2002:294, paragraphs 69 and 71, and of 5 June 2012, Commission v EDF, C‑124/10 P, EU:C:2012:318, paragraph 105).

245    In that regard, it is sufficient to note that the determination of a transfer price in line with market conditions does not find its basis in the principle of equal treatment between public and private undertakings, but, as the Commission recognises, in the legitimate objective of a prior tax agreement, such as the APA, which is to establish, for reasons of legal certainty, in advance, the application of a tax provision.  

246    It must be held that, to the extent that the Commission considers that the adoption of a prior tax agreement, such as the APA, gave rise to new aid, it should have been notified of that aid before its implementation, in accordance with Article 108(3) TFEU. Had the Commission stated a position on such a notification, however, it could not have taken into consideration information that was not known or reasonably foreseeable at the time of its decision. It therefore cannot criticise the Member State concerned for not having taken into account information that was not known or reasonably foreseeable at the time of the adoption of the agreement in question.

247    In that context, first, it must be recalled, that it is apparent from Article 1 and recital 40 of the contested decision that the measure contested by the Commission is the APA alone.

248    Second, while it is true that the APA could be revoked or amended during its validity period, from 2007 to 2017, it is important to note that, in the contested decision, the Commission did not consider that the fact that the Netherlands authorities had not revoked or modified the APA during its validity had conferred an advantage on SMBV. Point 6, second indent, of the APA, read in conjunction with point 4, first indent, stipulates that it comes to an end when a significant change of the facts and circumstances approved by the APA occurs, unless the parties have amicably agreed a revision arrangement. There was therefore nothing to prevent the Commission from finding that a substantial change of the facts and circumstances approved by the APA had taken place and that, consequently, a continued application of the APA conferred a selective advantage on SMBV.

249    Third, regarding the Commission’s argument that the APA was subject to a half-term check after the sixth accounting year, ending on 31 December 2013, and that the APA was not modified on that occasion, it is sufficient to point out that nowhere in the contested decision did the Commission argue that the absence of modification or revocation of the APA, following that half-term check, had conferred an advantage on SMBV under Article 107(1) TFEU.

250    It follows that, in those circumstances, the examination of the existence of an advantage conferred by a prior agreement, such as the APA, should be determined in view of the context of the time at which that agreement was concluded. The implication of that finding is that the Commission is required to refrain from assessments based on a situation subsequent to the adoption of the APA.

251    Accordingly, it is necessary to uphold Starbucks’ argument that, in the circumstances of the case at hand, the Commission could not base its analysis on information that was not available or reasonably foreseeable in April 2008, when the APA was concluded.

(e)    Whether the roasting IP represented a value for SMBV

252    By the second argument set out in recitals 310 to 332 of the contested decision (see paragraph 230 above), the Commission sought, in essence, to demonstrate that the payment of a royalty by SMBV to Alki was not justified, in principle, since SMBV did not, according to the Commission, benefit from the value of the roasting IP. That argument is divided into two parts. In essence, first, the Commission took the view that SMBV did not exploit the roasting IP directly on the market. Second, it found that the coffee roasting activity did not generate sufficient profit to allow for royalty payments.

(1)    Whether SMBV exploited the roasting IP directly on the market

253    As regards the argument that SMBV did not exploit the roasting IP directly on the market, the Commission explained, in recitals 310 to 313 of the contested decision that, first of all, in the specific relationship between Alki and SMBV, the value of the roasting IP was not ‘captured’ by the roaster, namely, SMBV. According to the Commission, the importance of the roasting know-how and curves lay in ensuring a consistent taste associated with the brand and its individual products. It inferred that the value of Starbucks’ roasting know-how and curves was ‘exploited’ only when Starbucks products were sold by stores under the Starbucks brand. In addition, the Commission maintained that, on their own, the roasting know-how and curves did not generate value for the roaster on an ongoing basis if they could not be exploited on the market. As far as it was concerned, in the case of SMBV, the roasting know-how and curves ‘appear[ed]’ to constitute a technical specification according to which the roasting ought to proceed due to a preference or a choice of the purchasing company. The fact that the specifications laid down by Alki regarding the roasting process and, in particular, the roasting curves allowed SMBV to roast coffee sold under the Starbucks brand did not, according to the Commission, bring any benefit to SMBV in terms of increased sales or sales price, given that SMBV did not, in principle, sell its production to final customers who valued the Starbucks brand. Last, the Commission added that SMBV sold virtually all its production to Starbucks-franchised stores and that it therefore did not exploit the roasting IP directly on the market.

254    In its submissions, the Commission adds that the value of the roasting IP is exploited only where the products are sold to final customers who valued the consistent taste associated with the brand in question. Economically, it would be irrational for the roaster/coffee producer to pay a royalty to use the roasting IP when it does not market the finished product directly. That lies in the fact that, in such a scenario, the roaster/coffee producer uses that IP to roast coffee beans at the request of the contractor.  

255    As a preliminary point, first, [confidential]. It follows that, under the roasting agreement, SMBV was obliged to pay the royalty in return for the use of the roasting IP.

256    Second, it must be pointed out that the Commission did not argue, in the contested decision, that its thesis according to which the roasting IP is exploited with end consumers constituted a test prescribed by Netherlands tax law. On the contrary, it is apparent from recitals 310 to 313 of the contested decision, read in combination with the introductory recitals setting out the Commission’s position upon the adoption of the opening decision, that the Commission conducted a purely economic examination that it based on the 1995 and 2010 versions of the OECD Guidelines.

257    In the light of those considerations, it is necessary to examine the merits of the proposition of the Commission, set out in recitals 298, 300 and 310 to 313 of the contested decision, according to which SMBV did not directly exploit the IP on the market on the ground that it did not sell products to final consumers.

258    In that regard, it must be held that the explanations given in recitals 310 to 313 of the contested decision lack plausibility. The reasoning followed by the Commission in recitals 310 to 313 of the contested decision as well as in its submissions before the Court is, in essence, based on the premiss that the value of the roasting IP is exploited only where the products are sold to final consumers who value the consistent taste associated with the mark in question and where, economically, it would not  be rational for the roaster/coffee producer to pay a royalty to use the roasting IP when it does not market the finished product directly. However, that premiss is not borne out by the facts established in the contested decision.

259    First, it is settled between the parties that the roasting IP was, in principle, capable of representing an economic value. Second, it is also settled between the parties that SMBV is a roaster that was obliged to use the roasting IP to roast its coffee. Third, the Commission maintains that Starbucks stores, both affiliated and unaffiliated, are required to purchase roasted coffee from SMBV and that SMBV is thus also the vendor of the roasted coffee.

260    In that context, it must be held that the Commission was wrong to focus its analysis on the premiss that the value of the roasting IP is exploited only where the products are sold to final consumers. The question of who ultimately bears the costs corresponding to the compensation of the value of the IP used for coffee production is clearly separate from the question of whether the roasting IP was necessary to allow SMBV to produce roasted coffee according to the criteria stipulated by Starbucks stores, to which it sells, on its own behalf, the coffee.

261    In the event that SMBV sells the coffee it has roasted to Starbucks stores which require coffee to have been roasted according to Starbucks’ specifications, it is plausible that, in the absence of the right to use, or – to use the terminology of the contested decision – exploit the roasting IP, SMBV would not have been in a position to produce and supply roasted coffee according to Starbucks’ specifications in stores of the same name.

262    From this it must be concluded that, contrary to what the Commission argues, SMBV’s payment of a royalty to use the roasting IP is not devoid of all economic rationality. The IP was, after all, necessary for exercising SMBV’s economic activity, namely, the production of roasted coffee according to Starbucks’ specifications. It follows that SMBV does indeed derive  added value from the use of the roasting IP, without which it could not then resell the roasted coffee to Starbucks stores.

263    Furthermore, it is necessary to reject the Commission’s argument that it is Starbucks stores which pay royalties to Starbucks Coffee Emea which also include a remuneration [confidential]. First, the submissions under the present line of reasoning in the contested decision do not contain any element such as to substantiate that proposition. Second, the circumstance that Starbucks stores pay a royalty to Starbucks Coffee Emea does not preclude SMBV from being able to pass on [confidential] in the prices invoiced to stores. In addition, the fact that, according to the Commission, Starbucks stores pay a second royalty [confidential] to Starbucks Coffee Emea, [confidential], is capable of conferring an advantage, at most, on the latter, but not on SMBV.

264    It follows from the foregoing that the Commission was wrong to consider, in recitals 298 and 300 of the contested decision, that an unaffiliated manufacturing company exploits a roasting IP only if it sells its products to end customers. The exploitation of roasting IP is not limited to situations in which a roaster sells its coffee on the retail market to end consumers, but also includes situations such as that of SMBV, in which a roaster is active as a seller on the wholesale market. On the contrary, merely processing coffee on behalf of a contractor that procures the technical specifications for manufacture does not suffice to demonstrate that such an IP is being exploited.

265    Accordingly, it must be held that the Commission erred in finding that SMBV, as described in the contested decision, did not have to pay a royalty since it did not exploit the roasting IP directly on the market.

(2)    Whether SMBV has been loss-making on its roasting activities

266    The Kingdom of the Netherlands and Starbucks dispute the argument of the Commission, set out in recitals 314 to 317 of the contested decision, that SMBV has generated a loss on its roasting activities since 2010, a situation which did not permit the payment of a royalty for the roasting IP. According to the Kingdom of the Netherlands, the Commission inter alia did not take sufficient account of the fact that the beans purchased by SMBV were also used for the production of coffee by third parties. The Commission accordingly took the view that this demonstrated that the method used to determine the royalty as an adjustment variable, as approved by the APA, was not in conformity with the arm’s length principle.

267    The Commission retorts that, according to the information it received from Starbucks during the administrative procedure, only a limited portion of roasted coffee was processed by external manufacturers. It was thus right to find that virtually all the beans purchased by SMBV were processed as part of its own coffee-production activities.

268    In the contested decision, the Commission found, in essence, that SMBV has generated a loss on its roasting activities since 2010 and that the royalty paid by SMBV to Alki was financed in part by SMBV’s other activities, without it having had any prospects of future profits resulting from the roasting. According to the Commission, the coffee roasting activity did not generate sufficient profits to enable the payment of the royalty. Moreover, the Commission contends that the royalty paid by SMBV to Alki for the roasting IP in an intra-group context ‘appear[ed]’ to serve structurally the sole purpose of shifting profits derived from SMBV’s reselling function to Alki.

269    At the outset, it must be pointed out that the Commission’s reasoning is based on the premiss according to which it is necessary to achieve profits on the roasting activities in order to be in a position to pay a royalty for the roasting IP. The Commission does not, however, demonstrate that the Netherlands tax rules provide that the obligation to pay a royalty is dependent on the profitability of the activity in question. In addition, the question of whether SMBV’s roasting activities were profitable is unrelated to whether an obligation to pay a royalty such as that at issue in the present case could be economically justified.

270    In that regard, first of all, it must be pointed out that the Commission argues that the roasting activity did not generate sufficient profit for the period commencing in 2010. That finding therefore does not concern the entirety of the validity period of the APA (commencing in 2007).

271    Next, it must be stated that, as has been indicated in paragraphs 243 to 251 above, in the circumstances of the case at hand, the Commission was required to refrain from any assessment based on a situation subsequent to the APA’s conclusion. The Commission does not explain, in the contested decision, however, how the losses it mentions in recitals 314 to 317 thereof would have been foreseeable when the APA was adopted, when they pertained to SMBV’s situation only from 2010. The Commission has therefore not demonstrated that it was entitled to rely on the fact that SMBV had sustained a loss on its roasting activities since 2010.

272    Last, in any event, to the extent that Starbucks claims that SMBV’s roasting activities have always been profitable, it is worth recalling that the Commission conducted its analysis based on a comparison of the revenue received from Starbucks stores with the value of SMBV’s purchase of green coffee beans from SCTC. However, under the third line of reasoning, the Commission specifically maintains that the increase of the price of green coffee beans, from 2010 onwards, was too high. It is thus already apparent from the Commission’s arguments in the contested decision that the costs of green coffee were considerably overvalued and that, accordingly, the losses it mentions in the contested decision did not exist, at least in the proportions found in recitals 314 to 317 of the contested decision.

273    Those findings are sufficient to reject the Commission’s argument that SMBV was not in a position to pay a royalty for the roasting IP due to its having been loss-making on its roasting activities.

274    In any event, Starbucks claims that the Commission’s calculation is erroneous since no account was taken of the fact that a considerable volume of the total purchased green coffee had not been roasted by SMBV. The Commission contends that that argument is inadmissible, on the ground that that information is new and moreover contradicts the information provided during the administrative procedure.

275    In that regard, it must be pointed out that the Commission acknowledges, both in footnote 155 of the contested decision and in its submissions, that the information provided by Starbucks during the administrative procedure led to the conclusion that practically all the green coffee purchased by SMBV, apart from ‘limited volumes’ that had been provided to third parties, had also been roasted by SMBV. In that context, the Commission makes reference to the letter of the Starbucks correspondents that it was sent on 23 September 2015. However, it is apparent from that letter that the third party in question had a custom production contract with the Starbucks group which ‘predominantly’ concerned the production of products other than roasted coffee, ‘but also the roasting of green coffee as such (be it in limited volumes)’. The reference to ‘limited volumes’ indicates that the third party in question produced a limited quantity of roasted coffee compared to its production of products other than coffee powder, but not that it produced negligible quantities of roasted coffee. The Commission was thus informed during the administrative procedure that a portion of the green coffee purchased by SMBV had not been roasted by SMBV. The objection made by the Commission as to the admissibility of Starbucks’ argument, based on the notion that that argument is based on information not brought to its attention during the administrative procedure, therefore has no basis in fact and must be rejected.

276    So far as concerns the merits of Starbucks’ argument that the Commission took into account the entirety of the sums corresponding to SMBV’s purchases of green coffee as costs for its calculation when a considerable volume of the total purchased green coffee had not been roasted by SMBV, it must be noted that the Commission contends that Starbucks did not indicate, in the documents communicated on 29 May 2015, that a meaningful portion of green coffee beans had been roasted by third parties. However, as Starbucks rightly argues, the response to question 2 in the letter of the Starbucks correspondents of 29 May 2015 that the Commission cites in its submissions pertained to the allocation of SMBV’s profits to its various functions, and not to the allocation of its costs to those functions. It follows that the responses of the Starbucks correspondents on which the Commission, according to its submissions, based its finding that SMBV’s roasting function had been loss-making from 2010 did not suffice to enable the Commission to arrive at that conclusion.

277    Furthermore, as has been set out in paragraph 275 above, when the contested decision was adopted, the Commission already had indication that its calculation, set out in recital 314 of the contested decision, consisting in subtracting the price paid by SMBV to SCTC for green coffee beans from the revenue generated from coffee roasting, was erroneous.

278    If follows that the Commission has not demonstrated that SMBV has generated a loss on its roasting activities since 2010, a situation which did not permit the payment of a royalty for the roasting IP.

(f)    Comparison with coffee roasting agreements concluded by Starbucks with third parties and against similar licensing arrangements ‘on the market’

279    By the third argument set out in the contested decision (see paragraph 229 above), the Commission sought to explain, in essence, that the manufacturing agreements concluded by Starbucks with third parties and certain agreements between Starbucks’ competitors and third roasters did not provide for any royalty for the use of the roasting IP (recitals 291 to 309 of the contested decision).

280    In that context, the Commission explained, in recital 309 of the contested decision, that a transfer pricing analysis of the arm’s length value of the royalty paid to Alki by SMBV for the roasting IP led to the conclusion that no royalty ought to be due for that IP in that specific relationship. It based that finding, first, on an analysis of the manufacturing agreements concluded by Starbucks with third parties and, second, on a comparison with agreements concluded between Starbucks’ competitors and third roasters. It is apparent inter alia from recitals 291 and 299 of the contested decision that the Commission sought to determine the level of an arm’s length royalty between SMBV and Alki.

281    The Kingdom of the Netherlands and Starbucks are at odds with the Commission, in essence, on whether the contracts concluded by Starbucks with external roasters and with manufacturers of coffee-derived products, on which the comparison conducted by the Commission is based, were relevant to carrying out a comparison with the contractual arrangements between Alki and SMBV, under the CUP method.

282    In essence, in respect of the question of whether the manufacturing agreements concluded by Starbucks with third parties imply that the royalty should have been zero, the Kingdom of the Netherlands and Starbucks claim that:

–        the contracts concluded between Starbucks and external roasters and manufacturers of coffee-derived products, on which the contested decision is based, could not be used for a comparison with the contractual arrangements between Alki and SMBV, based on the CUP method;

–        the Commission’s analysis of the royalties relies almost exclusively on evidence that was not available in April 2008;

–        the majority of the contracts used by the Commission for comparing transactions pertained to specific coffee-derived products other than roasted coffee beans;

–        Alki’s remuneration was inextricably linked to the purchase of green coffee beans from SCTC, but none of the transactions derived from the contracts used by the Commission for the comparison was inextricably linked to another transaction in that way;

–        all third-party manufacturers, mentioned in the contested decision, which – similar to SMBV – supplied Starbucks-branded coffee products to stores or retailers, paid substantial royalties for the use of the IP for the roasting of Starbucks’ coffee.

283    As for the manufacturing agreements concluded by Starbucks with third parties, the Commission examined, as a first step, in recitals 291 to 298 of the contested decision, whether the coffee roasting contracts concluded by the Starbucks group with 10 third companies offered a direct point of comparison enabling the amount of the royalty owed by SMBV to Alki to be determined. In that regard, the Commission based its examination on paragraph 1.36 of the 2010 version of the OECD Guidelines, which, for the purposes of the comparability analysis of the controlled transactions of the corporate taxpayer and the comparable transactions on the free market, lists five comparability factors, including the characteristics of the property or services transferred, the functions performed by the parties, the contractual terms, the economic circumstances of the parties and the business strategies pursued by the parties. The Commission also referred, in footnote 147 of the contested decision, to paragraph 1.17 of the 1995 version of the OECD Guidelines. According to the latter, for the purposes of the comparability analysis, the characteristics that may be important are those of the property or services transferred, the functions performed by the parties, the contractual terms, the respective economic circumstances of the parties and the business strategies that they pursue.

284    As a second step, in recitals 299 to 304 of the contested decision, the Commission found that, on the basis of those 10 uncontrolled transactions, the level of an arm’s length royalty between SMBV and Alki could be determined by using the CUP method.

285    More specifically, first, in order to determine the level of the royalty using the CUP method, it compared the payment of the royalty from SMBV to Alki with the payments due from third parties to other Starbucks group undertakings, in comparable transactions concluded under similar uncontrolled conditions. Second, the Commission analysed contracts concluded by the Starbucks group with unaffiliated manufacturing company 1 and with the companies designated, in recital 300 of the version of the contested decision published in the Official Journal of the European Union, by the terms ‘unaffiliated manufacturing companies 2, 3, 4, 8, 9 and 10’ (respectively, ‘unaffiliated manufacturing company 2’, ‘unaffiliated manufacturing company 3’, ‘unaffiliated manufacturing company 4’, ‘unaffiliated manufacturing company 8’, ‘unaffiliated manufacturing company 9’ and ‘unaffiliated manufacturing company 10’). It then found that those third parties did not pay royalties under their licensing agreements with the Starbucks group if they did not exploit the roasting IP directly on the market. Third, the Commission found, in respect of the relationships between the Starbucks group and the companies designated, in recital 303 of the version of the contested decision published in the Official Journal of the European Union, by the terms ‘unaffiliated manufacturing companies 5, 6 and 7’ (respectively, ‘unaffiliated manufacturing company 5’, ‘unaffiliated manufacturing company 6’ and ‘unaffiliated manufacturing company 7’), that only the trade mark and technology licence agreements concluded with those third parties by Starbucks contained a royalty payment.

286    As a third step, the Commission found, in recital 309 of the contested decision, that the coffee roasting contracts concluded by the Starbucks group with ten third companies did not require any royalty for the use of the roasting IP. The Commission thus concluded that no royalty ought to be due for that IP in the specific relationship between SMBV and Alki.

287    Without it being necessary, at this stage, to examine whether the choice by the Commission of the elements relevant to the comparability analysis, namely, the characteristics of the property or services transferred, the functions performed by the parties, the contractual terms, the respective economic circumstances of the parties and the business strategies that they pursued, was vitiated by error, it must be stated that there exist a number of elements in the context of that analysis that are at odds with the comparability between, on the one hand, the relationships between the Starbucks group and third parties and, on the other hand, the relationships between SMBV and Alki. Those elements are set out in paragraphs 288 to 345 below.

(1)    Contracts concluded subsequent to the APA

288    It must be noted that 7 of the 10 contracts examined by the Commission, namely, those concluded with unaffiliated manufacturing companies 1, 3, 4, 7, 8, 9 and 10, were examined after the APA’s conclusion. Given that the Commission does not explain how those contracts were available or reasonably foreseeable at the time of the APA’s conclusion, it was not in a position, for the reasons set out in paragraphs 243 to 251 above, to base its analysis of the APA on elements subsequent to its conclusion. It is therefore necessary to exclude those seven contracts from the comparison exercise.

(2)    Contracts concluded with undertakings which do not roast coffee

289    As has been set out in paragraphs 232 to 236 above, SMBV is a roaster of green coffee that pays Alki a royalty to use the roasting IP.

290    In recital 295 of the contested decision, the Commission acknowledged that, of the 10 third companies that concluded a contract with the Starbucks group, certain of them did not roast coffee. It is commonly known, however, that a company which does not roast coffee will not pay a royalty to the Starbucks group for the use of the roasting IP in order to produce roasted coffee.

291    In addition, in the contested decision, the Commission did not provide elements indicating that the contracts under which the third party did not produce roasted coffee would be comparable with the contract concluded between SMBV and Alki. It is true that that finding does not rule out that the Commission could have based its analysis on the transactions of an undertaking which did not exercise exactly the same functions as SMBV or which was in a different factual situation. In that case, the onus would have been on it to justify such a choice and to explain the adjustments that it would have made in its analysis so as to take into consideration the differences between the undertakings.

292    Therefore, a contract concluded with an undertaking that was not a roaster could not be used, in the case at hand, without adjustments or amendments, for the purposes of the comparison exercise to demonstrate that the level of the royalty paid by SMBV to Alki should have been zero.

293    In that regard, the contracts concluded with unaffiliated manufacturing companies 5, 6 and 7 did not concern, according to their description in the contested decision, the roasting of green coffee. Given that, under the contracts at issue, unaffiliated manufacturing companies 5, 6 and 7 did not exercise the function of coffee roaster, it must be concluded that the contracts concluded with the said undertakings could not be used, in the case at hand, for the purposes of the comparison exercise.

(3)    Contracts with undertakings that did not engage in the sale of roasted coffee to stores or to consumers

294    As has been set out in paragraph 235 above, the stock that SMBV purchased from SCTC and sold to stores appears on SMBV’s balance sheet because SMBV is the entity responsible for contracting and invoicing with stores. It follows that SMBV became the owner of the stock of green coffee that it roasted and sold to stores. It must be stated, however, that, if SMBV was a stand-alone company, it would not have been in a position to produce its coffee according to the Starbucks group’s specifications without having obtained the right to use the roasting IP. Therefore, it would not have been able to produce its roasted coffee without paying a royalty.

295    On the contrary, as has been set out in paragraph 236 above, a toll manufacturer or a supplier processes raw materials or semi-finished products on behalf of the contractor. Consequently, the roasting IP does not represent for it a technical specification for which it will not pay a royalty to the contractor.

296    In that regard, first, it must be noted that, in the defence in Case T‑636/16, the Commission maintains that, as regards their contractual relationship with the Starbucks group, unaffiliated manufacturing companies 1, 8 and 9 operated under toll-manufacturer agreements and that they mainly produced coffee products such as flavoured coffee, powder for a coffee-based product protected by a registered trade mark or soluble coffee. According to the Commission, unaffiliated manufacturing companies 1, 8 and 9 did not acquire title to the Starbucks components. In addition, the Commission recognised that the agreements with unaffiliated manufacturing companies 1, 8 and 9 differ from the coffee roasting agreement between SMBV and Alki.

297    Second, in terms of the contract concluded between the Starbucks group and unaffiliated manufacturing company 4, the Commission specified, in recital 148, third indent, of the contested decision, that that manufacturing company subcontracted coffee roasting. In that regard, Starbucks claims that unaffiliated manufacturing company 4 purchases green coffee from the Starbucks group and then roasts it in accordance with the brand curves and recipes for mixing beans provided. It then sells all of its roasted coffee to a subsidiary wholly owned by the Starbucks group, which sells the roasted coffee on to stores.

298    It follows from that description that unaffiliated manufacturing company 4 did not sell the coffee it roasted to stores. It merely provided roasted coffee, as a subcontractor, to a Starbucks group company engaged in the sale of coffee. In those conditions, the roasting IP constituted a mere technical specification for manufacture. Therefore, the fact that unaffiliated manufacturing company 4 did not pay a royalty to use the roasting IP to the Starbucks group does not mean that SMBV did not have to pay a royalty to Alki.

299    Third, with regard to unaffiliated manufacturing company 10, the Commission explains in its submissions in Case T‑636/16 that that company manufactured and roasted green coffee beans, purchased directly from green coffee suppliers, and sold all Starbucks-branded coffee products to a single entity of the Starbucks group engaged in their sale.

300    It follows from that description that unaffiliated manufacturing company 10 did not, therefore, sell its roasted coffee to stores, but to a Starbucks group company engaged in its sale. In those conditions, the roasting IP constituted a mere technical specification for manufacture. It is therefore unsurprising that that company did not pay a royalty to the Starbucks group to use the roasting IP.

301    The Commission retorts that both unaffiliated manufacturing companies 4 and 10 and SMBV manufacture coffee products of which they are not the independent supplier on the market and that, therefore, they are in comparable situations. However, that argument is unconvincing. It must be recalled that, in order to determine whether SMBV has benefited from an advantage within the meaning of Article 107(1) TFEU, it is appropriate to compare SMBV’s situation, under the measure in question, with the situation of a comparable undertaking exercising its activities autonomously in conditions of free competition (see paragraphs 148 and 149 above). The object of comparison in such an analysis is thus a stand-alone company in SMBV’s situation, namely, a company that roasts coffee and sells it to stores, on the market.

302    In view of those differences between the situation of SMBV and that of unaffiliated manufacturing companies 1, 4, 8, 9 and 10 and in the absence of additional elements indicating that there was nevertheless comparability between the contracts at issue, it was thus necessary to reject the analysis of the comparability of the contracts concluded between the Starbucks group and those companies.

(4)    Contracts concerning products other than roasted coffee

303    In recital 295 of the contested decision, the Commission recognised that, of the 10 third companies that concluded a contract with the Starbucks group, certain of them produced ready-to-drink beverages or other products and ingredients for drink preparation and that, therefore, not all of the 10 third companies produced roasted coffee. According to the same recital, the contracts which concerned the roasting of green coffee were those concluded with unaffiliated manufacturing companies 2, 3, 4 and 10.

304    As has been set out in paragraph 296 above, the Commission recognised that, as regards their contractual relationship with the Starbucks group, unaffiliated manufacturing companies 1, 8 and 9 mainly produced coffee products such as flavoured coffee, powder for a coffee-based product protected by a registered trade mark or soluble coffee. Moreover, the Commission recognised that the agreements with unaffiliated manufacturing companies 1, 8 and 9 differ, in that regard, from the coffee roasting agreement between SMBV and Alki.

305    In addition, it should be recalled that the Commission maintains, in the defence in Case T‑636/16, that its assessment of the third-party agreements does not, in principle, rely on the agreements concluded with unaffiliated manufacturing companies 5, 6 and 7, due to the differences in the licensed know-how – the roasting IP as opposed to the ready-to-drink production know-how – and the place of those companies in the supply chain – the fact that SMBV roasts coffee beans and them sells them on to distributors or third-party manufacturers, whereas unaffiliated manufacturing companies 5, 6 and 7 produce coffee-related products that they sell directly to their customers, in this case mainly supermarkets.

306    So far as concerns the contractual relationships between the Starbucks group and unaffiliated manufacturing companies 1, 5, 6, 7, 8 and 9, it must be pointed out that, in the contested decision, the Commission does not provide elements indicating that the contracts under which the third party does not produce roasted coffee to be sold to stores related or unrelated to the Starbucks group are comparable to that concluded between SMBV and Alki. It is apparent in particular from recitals 298 and 300 of the contested decision that, in the context of the comparison exercise between the royalty paid by SMBV to Alki and the royalties provided for, depending on the case, in the 10 contracts concluded between the Starbucks group and the third parties, the Commission considered that the element relevant for comparability was the question of whether the third party directly exploited the IP on the market by selling products to final consumers.

307    However, unaffiliated manufacturing companies 1, 5, 6, 7, 8 and 9 did not have, according to the Commission, a roasting function which involved the same product as SMBV’s coffee roasting function. The Commission therefore has not managed to demonstrate that those contracts were sufficiently comparable with the roasting contract concluded between SMBV and Alki.

308    Consequently, in the case at hand, for that reason, the contracts between the Starbucks group and unaffiliated manufacturing companies 1, 5, 6, 7, 8 and 9 must also be excluded from the comparison exercise.

(5)    Contract which provides for the payment of a royalty for the use of the roasting IP

309    In respect of the contract concluded between the Starbucks group and unaffiliated manufacturing company 3, the Commission maintained, in the second indent of recital 148 of the contested decision, that, under a roasting licence arrangement, unaffiliated manufacturing company 3 provided coffee roasting services. Coffee was sold to the Starbucks group and to a joint venture, held by unaffiliated manufacturing company 3 and the Starbucks group (‘the joint venture’), that was operating Starbucks stores in a country outside the European Union. Unaffiliated manufacturing company 3 paid the Starbucks group a royalty for the roasting of coffee, the amount of which was fixed, for a certain quantity of green coffee produced and sold to the joint venture.

310    In recital 301 of the contested decision, the Commission added that unaffiliated manufacturing company 3 paid a royalty to the Starbucks group only when it sold its production to the joint venture. In that case, according to the Commission, unaffiliated manufacturing company 3 ‘directly exploit[ed] the roasting IP on the market through a related party’, so that the royalty payment ‘appears’ to cover the distribution of Starbucks-branded products to third parties by the joint venture. According to that Commission, that conclusion is confirmed by the fact that, when unaffiliated manufacturing company 3 sold the roasted coffee on to the Starbucks group, rather than to the joint venture, and the distribution and exploitation on the market of the brand was ensured by the Starbucks group, no royalty was paid by unaffiliated manufacturing company 3 to Starbucks for the roasting IP.

311    In that regard, it must be found that it is settled between the Commission and Starbucks that, when unaffiliated manufacturing company 3 sells its roasted beans to the joint venture for a given territory, it pays the Starbucks group a roasting licence fee at a fixed amount per quantity of roasted and packaged coffee and that, when it sells its roasted beans to Starbucks [confidential] no roasting licence fee is paid.

312    That finding clearly contradicts the Commission’s theory that unaffiliated manufacturing company 3 did not pay a royalty under its licensing agreement concluded with the Starbucks group if it did not exploit the roasting IP directly in respect of end consumers on the market. As Starbucks argues – rightly – the obligation of unaffiliated manufacturing company 3 to pay a royalty depends solely on its sales of roasted coffee to stores in the territory concerned, irrespective of whether the stores distribute the roasted coffee to end consumers or not.

313    In that context, the Commission contends that there is a difference between the situation of unaffiliated manufacturing company 3 and that of SMBV, which lies in the fact that unaffiliated manufacturing company 3 and Starbucks stores in the territory concerned are controlled by the same party, namely, the parent company of unaffiliated manufacturing company 3. The Commission adds that the payment of a royalty by unaffiliated manufacturing company 3 ‘appears’ to be made on behalf of the joint venture, rather than remuneration for the use of the roasting IP by unaffiliated manufacturing company 3.  

314    First of all, it must be stated that, as has been set out in paragraphs 194 to 196 above, it is, in principle, for the Commission to provide proof, in the contested decision, of the existence of aid.

315    That obligation is not met if the Commission merely makes prima facie findings, such as, in the case at hand, when it is limited to finding that the payment of a royalty ‘appears’ to cover the distribution of Starbucks-branded products to third parties or it ‘appears’ to be made on behalf of the joint venture.

316    Next, it should be noted that the difference between SMBV’s situation and that of unaffiliated manufacturing company 3 noted by the Commission, namely, the fact that unaffiliated manufacturing company 3 sold its roasted coffee through the joint venture to Starbucks stores present in the territory concerned, does not call into question the fact that a roasting licence fee was paid, at a fixed amount per quantity of roasted and packaged coffee, by unaffiliated manufacturing company 3 to the Starbucks group. [confidential]

317    Last, the Commission itself maintains, in its submissions, that, since unaffiliated manufacturing company 3 and the joint venture are related parties, it is not possible to make a direct comparison to the relationship between SMBV and Starbucks stores in the EMEA region. That finding undermines all the more the Commission’s theory that the contractual relationships between unaffiliated manufacturing company 3 and the Starbucks group are comparable to those between SMBV and Alki and permit the conclusion that the royalty should be zero.

318    In summary, it follows from the foregoing that, contrary to what the Commission asserted in the contested decision, unaffiliated manufacturing company 3 was a roaster that paid a royalty to the Starbucks group for the use of a roasting IP.

319    Accordingly, for the reasons set out in paragraphs 289 to 318 above, it must be held that the Commission has not managed to demonstrate that a comparison between, on the one hand, the contractual relationships between Alki and SMBV and, on the other hand, the contractual relationships between the Starbucks group and unaffiliated manufacturing companies 1 and 3 to 10 permit the conclusion that the level of the royalty paid by SMBV to Alki should have been zero.

(6)    Contract concluded with unaffiliated manufacturing company 2

320    It is apparent from recital 148, first indent, of the contested decision that, in order to subcontract the roasting of coffee, the Starbucks group entered into two types of agreements with unaffiliated manufacturing company 2, which were amended at several instances. First, pursuant to a technology licence agreement, concluded before 2008, an affiliate of the Starbucks group granted a non-exclusive licence to unaffiliated manufacturing company 2 to use, inter alia, Starbucks’ technology and know-how to produce and sell roasted coffee to selected third parties with which Starbucks entered into supply agreements, namely, in essence, unaffiliated manufacturing company 5. In return, unaffiliated manufacturing company 2 was to perform services to ensure that the roasted coffee was of high quality. To that end, unaffiliated manufacturing company 2 was to comply, inter alia, with certain quality assurance standards established by Starbucks. The technology licence agreement stipulated that unaffiliated manufacturing company 2 did not have to pay any fees for the licence. Second, a green coffee supply agreement stipulated that unaffiliated manufacturing company 2 had the obligation to buy green coffee exclusively from the Starbucks group for a fixed fee per a certain quantity. The technology licence agreement and the supply agreement were concluded with two different entities within the Starbucks group.

321    In recitals 300 and 302 of the contested decision, the Commission added that unaffiliated manufacturing company 2 did not pay a royalty under its licensing agreement concluded with the Starbucks group if it did not exploit directly on the market the roasting IP by selling products to end consumers. However, it must be stated that it is apparent from the description in recital 148 of the contested decision that unaffiliated manufacturing company 2 did not sell its roasted coffee to end consumers.

322    As regards the question of whether unaffiliated manufacturing company 2 was in a situation comparable to that of SMBV, it must be held that the contractual arrangement between unaffiliated manufacturing company 2 and the Starbucks group is closely linked to that concluded between unaffiliated manufacturing company 5 and the Starbucks group. Several years before the APA’s conclusion, unaffiliated manufacturing company 5 and SMBV concluded a supply agreement in which the Starbucks group undertook to supply roasted coffee beans, concentrate and other coffee ingredients to unaffiliated manufacturing company 5.

323    At a later stage, but before the APA’s conclusion, unaffiliated manufacturing company 5 and SMBV concluded a [confidential] delegation agreement [confidential] to which unaffiliated manufacturing company 2 acceded on the same day. [confidential]

324    [confidential]

325    [confidential]

326    It follows from those provisions that the role of unaffiliated manufacturing company 2 was different from that of SMBV, which, according to the Commission, was a roaster that engaged also in the sale of roasted coffee to Starbucks stores. According to the delegation agreement, unaffiliated manufacturing company 2 supplied unaffiliated manufacturing company 5, for the purposes of enabling the Starbucks group to meet its contractual obligations towards the latter, as followed from the supply agreement.  

327    In that context, it must be recalled that, in the contested decision, the Commission categorised the contractual arrangement between the Starbucks group and unaffiliated manufacturing company 2 as a subcontract (see paragraph 320 above). However, as has been set out in paragraph 236 above, such a subcontractor is limited to carrying out roasting in accordance with the instructions of the contractor  in order to meet its contractual obligation to provide roasted coffee. In those conditions, the roaster is merely following the ordering party’s technical requirements.

328    It must be pointed out that the Commission does not provide, in the contested decision, sufficient evidence indicating that such a subcontract would be comparable to that concluded between SMBV and Alki for the purposes of determining the level of the royalty.

329    In any event, even supposing that, for the purposes of the determination of the level of the royalty, the contractual arrangements between the Starbucks group and unaffiliated manufacturing company 2 were comparable to those concluded between SMBV and Alki, the Commission is limited, in recital 302 of the contested decision, to rejecting Starbucks’ argument that the higher mark-up on green coffee beans that unaffiliated manufacturing company 2 paid to the Starbucks group represented a ‘disguised’ remuneration for the roasting IP. In that context, first, it asserted that the mark-up ‘appears’ to have been passed on entirely to unaffiliated manufacturing company 5. Second, it maintained that there ‘[we]re no indications that any mark-up to a purchase price would not be passed on directly to [unaffiliated manufacturing company 5] or otherwise affect the commercial conditions between [unaffiliated manufacturing company 5] and [unaffiliated manufacturing company 2], as this contractual arrangement was not concluded independently of the contractual arrangement between [the Starbucks group] and [unaffiliated manufacturing company 5]’.

330    However, the considerations set out in paragraph 302 of the contested decision have no bearing on the finding according to which the position of unaffiliated manufacturing company 2 as ‘subcontractor’ is not sufficient to conclude that SMBV, as a seller of its roasted coffee, should not have paid any royalty for the use of the roasting IP.

331    In addition, regarding the question of whether the higher mark-up on green coffee beans paid by unaffiliated manufacturing company 2 to the Starbucks group represented the remuneration for a coffee roasting IP, it must be noted that the Commission’s argument that the higher mark-up on coffee beans paid by unaffiliated manufacturing company 2 ‘appears’ to be transferred to unaffiliated manufacturing company 5 is speculative and does not rule out, as such, the possibility that a remuneration for the use of the roasting IP was actually paid to the Starbucks group by unaffiliated manufacturing company 2.

332    On the contrary, a number of elements raise doubts about the Commission’s argument that, in the case at hand, unaffiliated manufacturing company 2 paid no remuneration to the Starbucks group for the use of the roasting IP.

333    First, it must be stated that, at first sight, the level of the price of the green coffee beans provided by SMBV and paid by unaffiliated manufacturing company 2 to the Starbucks group appears to be high in the light of the figures Starbucks cites in footnote 189 of the application in Case T‑636/16. The Commission does not dispute those figures. Moreover, in recital 302 of the contested decision, the Commission did not dispute the Starbucks correspondents’ assertion that that price was high.

334    Second, the Commission submits that it found in the contested decision that the technology licence agreement stipulated that unaffiliated manufacturing company 2 ought not to pay any royalty for the use of the roasting IP. It considers that, consequently, it was for the Kingdom of the Netherlands and Starbucks to prove that the difference in the prices of green coffee represented a ‘disguised’ remuneration for the roasting IP, which they have not managed to do.

335    It must be recalled that the 1995 and 2010 versions of the OECD Guidelines, on which the Commission bases its comparability analysis, expressly provide, in paragraph 6.17 thereof, that the compensation for the use of intangible property may be included in the price charged for the sale of goods when, for example, one enterprise sells unfinished products to another and, at the same time, makes available its experience for further processing of those products. In that context, it must be noted that the Commission maintains – rightly – that a price difference is, in principle, different from a royalty, which potentially has different tax consequences, which is moreover restated, in essence, in paragraph 6.19 of the 2010 version of the OECD Guidelines.

336    In the case at hand, it is clear from the contested decision that Starbucks had argued during the administrative procedure that the higher mark-up on the costs of green coffee beans that unaffiliated manufacturing company 2 paid to the Starbucks group represented a remuneration for the roasting IP.

337    In those circumstances, the Starbucks correspondents’ arguments, raised during the administrative procedure, could not be rejected on the basis of the mere finding that the technology licence agreement stipulated that unaffiliated manufacturing company 2 ought not to pay any royalty for the use of the roasting IP.

338    Third, while the Commission is right to argue that the supply of green coffee beans and the sublicensing of IP are separate transactions on the basis of two contracts concluded with two different counterparties within the Starbucks group, the fact remains that the technology licence agreement between the Starbucks group and unaffiliated manufacturing company 2 indicates, [confidential].

339    Fourth, the Commission adds, in essence, that the difference in price between the green coffee beans unaffiliated manufacturing company 2 buys and those that SMBV buys can have various explanations, such as, first of all, the strong bargaining power of Starbucks [confidential]; then, the fact that unaffiliated manufacturing company 2 does not buy its green coffee beans directly from SCTC but from Starbucks [confidential] – which buys them from SCTC and sells them on to it – which could also lead to an extra mark-up on the cost to cover the value added by Starbucks [confidential], or, last, the differences in delivery terms.

340    First of all, it must be noted that the Commission’s argument that Starbucks [confidential] had a bargaining power that was so strong compared to unaffiliated manufacturing company 2 that it could demand a far higher price than that which it was capable of obtaining from [confidential] SMBV is unconvincing.

341    Next, while the Commission contends that the fact that unaffiliated manufacturing company 2 does not buy its green coffee beans directly from SCTC but from Starbucks [confidential] – which buys them from SCTC and sells them on to it – could also lead to an additional mark-up on the cost to cover [confidential], it does not explain [confidential]. Starbucks retorts, in that respect, that SCTC takes care of the entire procurement process, including transportation from the port of origin to the port of destination, whereby the coffee beans are delivered to unaffiliated manufacturing company 2 without any processing whatsoever. Moreover, according to Starbucks, for reasons of administrative efficiency, [confidential]. That argument of the Commission should therefore also be rejected.

342    Last, the Commission argues that there is a difference in the delivery terms of green coffee beans enjoyed by unaffiliated manufacturing company 2 and SMBV. It contends that Starbucks [confidential] sells green coffee beans to unaffiliated manufacturing company 2 CIF (cost, insurance and freight) at the port of entry in the territory in which that manufacturing company carries out its economic activity, while the green coffee beans that SMBV obtains from SCTC are delivered FOB (free on board) at the port of Amsterdam (Netherlands). However, it must be pointed out, first, that the Commission does not quantify its contention that the difference in costs for an FOB delivery as opposed to a CIF delivery may be considerable. Second, Starbucks, for its part, claims that the difference in costs between an FOB delivery and a CIF delivery is too low to explain the ‘higher mark-up’. The Commission has therefore not managed to substantiate its contention that the ‘higher mark-up’ could not represent, even in part, a remuneration for the use of the roasting IP, since it was due entirely to the difference in delivery terms between the contracts in question.

343    In those circumstances, neither the succinct reasoning set out in recital 302 of the contested decision nor the other explanations provided by the Commission – which Starbucks disputes – allowed the Commission to conclude that the contracts between the Starbucks group and unaffiliated manufacturing company 2 demonstrated, to a requisite legal standard, that that undertaking paid no remuneration to the Starbucks group for the use of the roasting IP.

344    It follows that, on the basis of what has been set out in the contested decision, the Commission has not managed to demonstrate to a requisite legal standard that the contract between the Starbucks group and unaffiliated manufacturing company 2 indicates that the royalty paid by SMBV to Alki should have been zero.

345    In summary, it follows that the Commission has not managed to demonstrate, on the basis of its comparison with the contracts concluded with the 10 unaffiliated manufacturing companies, that the royalty should have been zero. The contracts concluded with unaffiliated manufacturing companies 1, 3, 4, 7, 8, 9 and 10 were concluded subsequent to the adoption of the APA. The contracts concluded with unaffiliated manufacturing companies 5, 6 and 7, for their part, involve undertakings which do not roast coffee. The contracts concluded with unaffiliated manufacturing companies 1, 4, 8, 9 and 10 are not resale contracts. The contracts concluded with unaffiliated manufacturing companies 1, 5, 6, 7, 8 and 9 involve products other than roasted coffee and the contract concluded with unaffiliated manufacturing company 3 mentions the possibility of the payment of a royalty. Regarding the analysis of the contract concluded with unaffiliated manufacturing company 2, the Commission’s succinct and speculative arguments do not suffice to demonstrate that that company did not pay any remuneration to the Starbucks group for the use of the roasting IP.

346    Accordingly, it is apparent from the considerations set out in paragraphs 288 to 345 above that the Commission has not demonstrated that applying the CUP method on the basis of a comparison with the contracts concluded between the Starbucks group and the 10 unaffiliated manufacturing companies would have led to the conclusion that the royalty paid by SMBV to Alki for the roasting IP, had it been set in market conditions, ought to have been zero.

(g)    Arrangements between Starbucks’ competitors and third-party roasters

347    The Commission also compared the royalty paid by SMBV to Alki with what was provided for in a number of arrangements between Starbucks and third-party roasters. The Commission found that it was apparent from that comparability analysis that no royalty for the use of the roasting IP ought to have been paid by SMBV to Alki.

348    It is apparent from recital 309 of the contested decision that, for the purposes of assessing whether SMBV paid an arm’s length royalty to Alki for the roasting IP, the Commission compared the arrangement between Alki and SMBV to a number of arrangements between Starbucks’ competitors and third-party roasters. In that context, the Commission made reference to the responses of Melitta, Dalmayr and Company Y.

349    Starbucks disputes the analysis conducted by the Commission. It is of the view that the arrangements concerning Melitta and Company X are ‘toll or contract manufacturer agreements, which – unlike SMBV – supply the finished products back to their principal and not directly to the principal’s customers’. That, in its view, makes those arrangements inherently different from the roasting agreement and, therefore, examining them is irrelevant to the present case. It is thus appropriate to examine whether those three contractual relationships were comparable to the roasting agreement between SMBV and Alki.

350    First, so far as concerns Melitta, the Commission indicated, in recital 306 of the contested decision, that that competitor company of Starbucks explained to the Commission that when outsourcing the roasting of coffee it did not perceive royalties from third parties, even though it put its roasting curves at their disposal.

351    In that regard, it must be noted that it is apparent from recitals 207 and 208 of the contested decision that, in certain conditions where roasting capacities were exhausted, Melitta outsourced the roasting of coffee. However, it does not follow from that description that the third party roaster did in fact sell roasted coffee to stores or to other customers.

352    It must thus be held that, according to the findings set out in the contested decision, Melitta’s situation is not comparable to that of SMBV.

353    Second, regarding Company Y, which belongs to a group of companies, the Commission indicated, in recitals 211 and 307 of the contested decision, that that company’s roasting was ensured by a group company designated as a toll manufacturer and that that company did not pay any royalty to the group.

354    It must be stated that it is apparent from that description that the roaster of the group to which Company Y belonged operated as a toll manufacturer. The roaster processed green coffee on behalf of another company within the group to which Company Y belonged. This means that the roaster did not sell roasted coffee to stores or to other customers.

355    It thus follows from the findings made in the contested decision that Company Y’s situation is not comparable to that of SMBV.

356    Third, as regards Dallmayr, it is indicated in recital 308 of the contested decision that that competitor considered the payment of a royalty by a company providing roasting to be unusual, as it would have expected the customers to pay the roaster, not the other way around. It is apparent from recitals 204 and 205 of the contested decision that Dallmayr stated that coffee roasting was performed either as a stand-alone business or vertically integrated within a company. It stated that the sourcing function was ‘typically’ integrated with the roasting function. Dallmayr thus considered the payment of a royalty by a third party that provided the roasting services to be rather unusual. In fact, Dallmayr would have expected the customer to pay the roaster, not the other way around.

357    In that regard, it must be noted that Dallmayr merely states that it considers the payment of a royalty in the area of roasting to be ‘rather unusual’. At the same time, that statement does not preclude such a royalty from being provided for. Dallmayr’s declarations therefore do not rule out the existence of a royalty such as that paid by SMBV.

358    It is thus apparent from the considerations set out in paragraphs 347 to 357 above that the Commission has not demonstrated that the arrangements between Starbucks’ competitors and third-party roasters, identified in the contested decision, were relevant for the purposes of the analysis of SMBV’s situation. After all, the findings made in the contested decision in that regard do not support the conclusion that those arrangements were comparable to the roasting agreement. Accordingly, even supposing it were proved that no royalty was paid under the arrangements between Starbucks’ competitors and third-party roasters, that circumstance would not suffice to demonstrate that no royalty for the roasting IP ought to have been paid by SMBV to Alki.

359    For the reasons set out in paragraphs 279 to 358 above, it is therefore appropriate to hold that the Commission has not demonstrated, to the standard required by the case-law cited in paragraphs 194 to 196 above, that the royalty ought to have been zero. Therefore, it is necessary to uphold, on that basis, the actions of the Kingdom of the Netherlands and of Starbucks to the extent that they concern the second line of reasoning of the contested decision. Accordingly, it is not necessary to examine the arguments of the Kingdom of the Netherlands and of Starbucks by which they dispute the rejection of their arguments raised, during the administrative procedure, to justify the royalty’s existence (see paragraph 230 above).

(h)    Argument that the level of the royalty should have been lower than the level endorsed by the APA

360    As has been set out in paragraph 229 above, the Commission indicated, in the contested decision, that the variable nature of the royalty payment during the period from 2006 to 2014 gave a ‘first indication’ that the level of that payment bore no relation to the value of the roasting IP. In that regard, the Commission explained, during the hearing, that it was apparent from recitals 287 to 289 and from footnote 146 of the contested decision that the royalty should have been set at a lower level than that endorsed by the APA.

361    First of all, it must be noted that, in recital 287 of the contested decision, the Commission merely reproduced certain findings made in the opening decision, but without drawing any consequence whatsoever for the contested decision. Next, in recital 288 of the contested decision, the Commission explained that, for the period from 2006 to 2014, it calculated the annual amount of the royalty paid by SMBV to Alki as a percentage of the annual sales of roasted coffee by SMBV to stores, which, in its view, confirmed its doubts about the fluctuation of the royalty. Last, in recital 289 of the contested decision, the Commission added that the variable nature of the royalty payment gave a ‘first indication’ that the level of that payment bore no relation to the value of the IP for which it was being paid. Footnote 146 of the contested decision mentions, in essence, that, ‘for illustration ... no contract [examined by the Commission] was identified whereby remuneration was paid for coffee roasting technology licen[sed] on the market’.

362    It must therefore be pointed out that neither recitals 287 to 289 of the contested decision nor footnote 146 of that decision contains any argument according to which the level of the royalty ought to have been lower than the level endorsed by the APA. Those recitals merely state, first, that the variability of the royalty indicate that it was not linked to the value of the roasting IP, and, second, that that royalty in no way had to be paid.

363    On the contrary, it should be held that it apparent, inter alia, from recitals 290, 318, 339 and 445 of the contested decision that the Commission found that the royalty should have been precisely zero. A fortiori, in recital 340 of the contested decision, the Commission stated that the level of the royalty did not need to be estimated and that, in other words, the profits paid by SMBV to Alki as a royalty for the roasting IP should have been fully taxable in the Netherlands.

364    It follows from those considerations that the contested decision does not contain any consideration that would have been identifiable by the Kingdom of the Netherlands and Starbucks whereby the royalty should have been set at a lower level than that endorsed by the APA.

365    In any event, even assuming that it followed reasonably clearly from the contested decision that the royalty should have been set at a lower level than the level endorsed by the APA, the Kingdom of the Netherlands and Starbucks dispute, in essence, the Commission’s argument that the level of the royalty is dissociated from its economic value.

366    In that context, it should be noted that, indeed, it cannot be denied that the variable nature of the royalty raises questions regarding the economic rationality of the royalty. In the case at hand, the Kingdom of the Netherlands and Starbucks have not provided any convincing explanation justifying the choice of an unusual method to determine the level of the royalty.

367    However, the residual nature of that royalty simply means that it was calculated, in principle, from the determination of the level of other relevant charges and incomes as well as from an estimate of the level of SMBV’s taxable profit. If those parameters were correctly identified, the mere residual nature of the royalty would not preclude the level of the residual royalty from corresponding to its economic value.

368    It must be held that the findings made in recitals 287 to 289 of the contested decision were not sufficient to demonstrate that the royalty should have been set at a lower level than that endorsed by the APA for the entire period between 2006 and 2014, inter alia on the ground that the contested decision does not specify what is the level of royalty that the Commission would have regarded as appropriate.

369    In addition, it must be noted that, in the context of the finding made in recital 289 of the contested decision, the Commission made reference to footnote 146 of the contested decision, which indicated the following: 

‘[A]n analysis using RoyaltyStat, at 2Q 2015, shows that out of the 168 agreements available through the database across sector whereby only technology was licen[sed], the median value of the royalty was 5 % of sales (based on 143 of these agreements where the licence fee was determined as a percentage of the value of sales rather than amount paid per unit sold). Among all the contracts available through the RoyaltyStat database, no contract was identified whereby remuneration was paid for coffee roasting technology licen[sed] on the market. Such technology was only licen[sed] out in certain instances in combination with trademarks.’

370    In that regard, it must be stated, first, that those considerations were made only ‘for illustration’, second, that while the Commission argues that a royalty was paid ‘across sector whereby only technology was licen[sed]’ and that there were examples of a ‘licen[sing] out in certain instances in combination with trademarks’, it nevertheless does not explain what the appropriate level of such a royalty would be and, third, that the Commission has not explained the reasons why it is of the view that the data relating to 2015 were reasonably foreseeable at the time of the APA’s conclusion in 2008.

371    The Commission has thus not substantiated to a requisite legal standard its assertion that, for the entire period between 2006 and 2014, the level of the royalty was not linked to the value of the IP for which it had been paid and that, as a result, an economic advantage had been conferred on SMBV.

372    It follows that it is necessary to reject the Commission’s argument that it demonstrated in the contested decision that the royalty should have been set at a lower level than that endorsed by the APA.

373    Accordingly, it is appropriate to uphold the second part of the third plea in Case T‑760/15 and the fourth part of the second plea in Case T‑636/16, to the extent that the Kingdom of the Netherlands and Starbucks dispute that the Commission demonstrated, under the second line of reasoning, that the royalty paid by SMBV to Alki should have been zero and that it resulted in an advantage within the meaning of Article 107(1) TFEU, without it being necessary to examine Starbucks’ argument that the Commission was obliged to determine an arm’s length range for the royalty.

3.      Annual determination of the costs of green coffee beans (third line of reasoning)

374    The Kingdom of the Netherlands and Starbucks raise, in essence, two complaints against the analysis conducted by the Commission under the third line of reasoning of the contested decision, according to which the level of the price of green coffee beans was overvalued when the question of whether their price was in conformity with the arm’s length principle had not been examined in the APA. By the first complaint, Starbucks claims that the third line of reasoning concerns an element of SMBV’s costs that was outside the scope of the contested measure as defined in the contested decision. By the second complaint, the Kingdom of the Netherlands and Starbucks dispute the finding that the level of the mark-up applied to the costs of green coffee beans sold by SCTC to SMBV was not in conformity with an arm’s length level.

(a)    Whether the price of green coffee beans was outside the scope of the contested measure

375    With regard to the first complaint, Starbucks claims, in essence, that the Commission’s third line of reasoning, relating to the price of green coffee beans, concerns an element of SMBV’s costs that was outside the scope of the contested measure as defined in the contested decision. Starbucks notes that the Commission did not examine the matter of the price of green coffee from the perspective of the time of the APA’s conclusion, that is to say, in April 2008. It adds, in the reply, that certain arguments raised in the defence in Case T‑636/16 indicate that the tax advantages identified by the Commission resulting from the price of green coffee beans for the years 2011 to 2014 cannot be attributed to the APA. The alleged tax advantages resulting from the price of green coffee beans cannot be attributed to the APA, but only to the annual assessments endorsing those prices, and thus fall outside the scope of the ‘contested decision’.

376    The Commission argues that it is clear from the contested decision as well as from the defence in Case T‑636/16 that the price for green coffee beans should have been examined in order to establish whether the level of that price was too high and led to a reduction in SMBV’s taxable profit.

377    As regards the scope of the contested measure as established by the contested decision, it must be stated that, according to the wording of Article 1 of the contested decision, the measure which constitutes aid within the meaning of Article 107(1) TFEU and which was implemented by the Kingdom of the Netherlands in infringement of Article 108(3) TFEU is the APA, ‘entered into by the Netherlands on 28 April 2008 with [SMBV]’. It is apparent from that provision and from the definition set out in recital 40 of the contested decision that the contested measure is therefore made up solely of the APA.

378    In that regard, it must be noted that it follows from the provisions of the APA (see paragraphs 12 to 16 above) that that agreement determines the method for calculating SMBV’s remuneration for its production and distribution activities, which serves to establish the taxable base for SMBV’s payment of Netherlands corporate income tax. In that context, although the APA makes reference to the price of green coffee beans paid by SMBV to SCTC by noting that those costs are excluded from SMBV’s cost base, it does not resolve the question of what level of transfer pricing should be set for the purchase of green coffee beans. It is necessary to distinguish the question of whether the cost of green coffee beans is part of the cost base for the calculation of the taxable base from the question of what is the amount of the transfer price of those transactions which was actually determined for a given year. The APA does not contain any element, however, enabling that amount to be determined, meaning that the Netherlands authorities did not validate, in the framework of the APA, any transfer pricing method or price level in respect of green coffee beans.

379    It must be specified that, in the absence of determination of price level for the purchase of green coffee beans in the APA, the annual setting of the price of coffee beans, in particular for the years from 2011 to 2014, should have been carried out, if necessary, in the context of annual tax assessments.

380    It follows that the APA did not make provision for  determining annually the level of the costs of green coffee beans and that was, consequently, outside the scope of the contested measure. That finding is not called into question by the Commission’s arguments.

381    First, the Commission is of the view that the APA, which constitutes the contested measure, should have predetermined transfer pricing for green coffee beans from the 2011 fiscal year onwards. According to recital 447 of the contested decision, SMBV’s taxable profits resulting from the costs of green coffee beans for the fiscal years from 2011 onwards should have been set at a higher level. It is apparent from recitals 360 and 361 of the contested decision that, according to the Commission, the transfer pricing report failed to examine whether the price of green coffee beans, paid by SMBV to SCTC, was at arm’s length. According to the Commission, that ‘means’ that the method proposed in that report for determining SMBV’s taxable profits confers a selective advantage on it. In addition, the Commission maintained, in recital 348 of the contested decision, that the APA should have prescribed an arm’s length price in 2008 from which no deviation would have been possible in 2011, including an increase in the mark-up, unless those prices were replaced or amended in the APA.

382    However, it must be pointed out that the transfer pricing report contains no examination of the transfer pricing applicable for specific transactions such as the price of green coffee beans requested by SCTC of SMBV. On the contrary, it sets out the method for calculating SMBV’s remuneration for its production and distribution activities, which constitutes the taxable base for the purpose of Netherlands corporate income tax.

383    The APA consists merely of the acquisition of confirmation, in advance, of the tax treatment of a taxpayer. An anticipatory decision such as the APA does not necessarily cover all aspects of a taxpayer’s tax treatment, but may deal only with certain specific issues. It is moreover apparent from page 28 of the transfer pricing report that the Starbucks group’s tax advisor considered that the transactions relating to green coffee were different transactions from those for which the APA had been requested.

384    First, the Commission has not provided any evidence indicating that, under Netherlands law – the relevant law in this regard – the question of whether the level of the price of green coffee beans paid by SMBV to SCTC was at arm’s length should have been examined in the APA.

385    Second, the mere fact that the APA does not predetermine a level of transfer price for green coffee beans does not in itself mean that the APA, in setting the method for determining SMBV’s remuneration, conferred an advantage on SMBV for its production and distribution activities.

386    Second, the Commission observes, in its submissions, that the technical implementation of the APA, via the annual tax assessments, also constitutes the grant of aid. However, such a finding does not follow from the contested decision. The Commission refers in that regard to Article 1 of the contested decision, which makes reference to the fact that, on the basis of the APA, SMBV ‘[is] enable[d] ... to determine its corporate income tax liability in [the Kingdom of] the Netherlands on a yearly basis for a period of 10 years’. The Commission adds that the contested decision makes numerous references to SMBV’s taxable profit as determined by the APA. In its view, the APA has no value unless it is used ‘to prepare those tax declarations’. In that regard, it cites recital 225 of the contested decision, which states that the APA entails an acceptance by the Netherlands tax administration of a profit allocation proposed by Starbucks on the basis of which SMBV determines the amount of its corporate income tax liability to the Netherlands on a yearly basis.

387    Contrary to the assertions of the Commission as described above, it must be held that the annual tax assessments relating to SMBV do not implement the APA in a purely technical manner. While it is certainly true that the APA and the transfer pricing report on which it is based predetermine the method for calculating SMBV’s taxable profit on the basis of Netherlands corporate income tax, they in no way allowed for a forecast of the income and annual charges declared by SMBV with a view to the actual transactions that took place during the year in question.

388    Additionally, the Commission’s assertion that the technical implementation of the APA via the annual tax assessments also constitutes the grant of aid is incorrect. The annual tax assessments implementing the APA are not part of the measure at issue as defined by the Commission – the APA – as is apparent from Article 1 of the contested decision. More specifically, the APA did not determine SMBV’s taxable profit on the basis of the costs of green coffee beans nor did  it deal with the issue of the annual determination of the costs of green coffee beans. In addition, nowhere in the contested decision did the Commission criticise the Netherlands authorities for having conferred a selective advantage on SMBV on account of the exclusion of the costs of green coffee beans from the taxable base, but merely disputed the fact that the level of their price had not been verified by the Netherlands tax authorities.

389    In any event, it should be stated that there was nothing preventing the Commission from defining the contested measure more broadly so as to cover the annual tax assessments relating to SMBV, yet it limited the scope of the contested measure to the APA alone.

390    Furthermore, it has been set out in paragraph 248 above that the APA could be revoked or modified during its validity, from 2007 to 2017. It must be noted that, in the contested decision, the Commission did not find that the fact that the Netherlands authorities had not revoked or modified the APA during its validity on the ground that the costs of green coffee beans were excessive had conferred an advantage on SMBV.

391    Accordingly, it is necessary to uphold the complaint that the third line of reasoning concerns an element of SMBV’s costs that was outside the scope of the contested measure. In so far as the level of the costs of green coffee beans for the fiscal years from 2011 onwards was not part of the contested measure, the Commission was not entitled to request the Kingdom of the Netherlands, in accordance with Article 2(1) of the contested decision, read in conjunction with recitals 447 and 448 thereof, to recover the difference between the amount actually paid by way of corporate income tax and the amount that would have been due had the taxable profit of SMBV resulting from the costs of green coffee beans for the fiscal years from 2011 onwards been set at a higher level.

(b)    Whether the level of the mark-up applied to the costs of green coffee beans sold by SCTC to SMBV was not in conformity with an arm’s length level

392    In any event, assuming that the third line of reasoning concerns an element of SMBV’s costs that was covered by the contested measure, it must be stated that the second complaint set out in paragraph 374 above would have to be upheld. From the outset, it should be recalled that the cost of the green coffee beans purchased by SMBV is excluded from the cost base of SMBV determined in the APA. In essence, the price of green coffee beans to be paid by SMBV to SCTC is made up of the costs of SCTC’s goods and a mark-up on those costs.

393    The contested decision explains that, for the period from 2005 to 2010, the average mark-up on the costs of the green coffee beans delivered by SCTC was [confidential]%, compared to [confidential]% for the period from 2011 to 2014. The corresponding average gross margin on the cost of goods (‘COGS’) for the period from 2005 to 2010 was [confidential]%, whereas the average gross margin on COGS for the period from 2011 to 2014 was [confidential]%. According to the contested decision, Starbucks claimed that the mark-up of [confidential]% applicable on average for the period from 2005 to 2010 corresponded to an arm’s length mark-up. The Commission next assumed that the increase in the mark-up from 2011 onwards could have constituted a remuneration recorded by SMBV on the coffee roasting activities. As the [confidential]% mark-up was also within the range for the supply function put forward by Starbucks during the administrative procedure, the Commission concluded that the [confidential]% mark-up on the costs of green coffee beans during the period from 2005 to 2010 was at arm’s length. Since Starbucks had not provided, according to the contested decision, any ‘valid’ justification for the increase in the average mark-up to [confidential]% from 2011 onwards, the Commission considered that no corresponding deduction to SMBV’s accounting profits as a result of that increase ought to be accepted from that period onwards.  

394    However, to arrive at a reliable approximation of an arm’s length mark-up for the period from 2011 onwards, the Commission accepted increasing the [confidential]% mark-up for the period from 2005 to 2010 by the costs of the C.A.F.E. Practices programme and up to the amount of the costs of [confidential]. According to the Commission, those costs represented [confidential]% of the costs of green coffee beans purchased by SCTC at the end of 2014 and translated into [confidential]% of the price charged to SMBV. An arm’s length mark-up recorded by SCTC for the period from 2011 onwards would therefore be up to [confidential]% of the costs of green coffee beans purchased by SCTC, corresponding to a gross margin of up to [confidential]% on SCTC’s COGS, charged by SCTC to SMBV.

395    The Commission concluded that the [confidential]% average mark-up on the costs of green coffee beans supplied by SCTC to SMBV effectively applied from 2011 to 2014 did not reflect a reliable approximation of a market-based outcome in line with the arm’s length principle.

396    First, it must be stated that, as has been indicated in paragraphs 243 to 251 above, in the circumstances of the case at hand, the Commission was required to refrain from any assessment based on a situation subsequent to the APA’s conclusion. The Commission does not explain, however, in the contested decision, how the high level of the costs of green coffee beans for the fiscal years from 2011 onwards, which it discusses in recitals 342 to 359 of that decision, would have been foreseeable at the time of the APA’s adoption, when that was SMBV’s situation from 2011 onwards. The Commission has thus not demonstrated that it was entitled to rely on the fact that SCTC had applied a higher mark-up on the costs of green coffee beans for the fiscal years from 2011 onwards.

397    Second, even assuming that the gradual increase in the mark-up from 2011 onwards was foreseeable at the time of the APA’s conclusion, it must be stated that the Commission’s approach is unconvincing. As Starbucks rightly contends, the Commission suggests that the mark-up for SCTC should have been set at the level of the average profit before tax that SCTC earned on its inter-company sales in the years before 2008, when such previous inter-company (‘controlled’) transactions cannot serve as a benchmark for ‘marke[t] based’ transfer prices.  

398    In that regard, it should be noted that the Commission maintains that the price that SMBV paid to SCTC was too high from 2011 onwards. It must be recalled that that price is one paid within the Starbucks group. To determine a transfer price, however, the Commission should have compared the price paid by SMBV to SCTC to a price for green coffee beans that a stand-alone company would have paid on the market. It should have determined a price range for green coffee beans that a stand-alone roaster in a comparable situation to SMBV’s would have paid on the market. However, instead of determining and examining such an uncontrolled transaction, the Commission limited its analysis to the controlled transaction in question and simply verified the plausibility of the structure of the costs and mark-ups of the other (integrated) party to the controlled transaction in question, namely, of SCTC.

399    By way of illustration, it should be recalled that the 2010 version of the OECD Guidelines, to which the Commission repeatedly refers in the contested decision, set out the following in paragraphs 3.24 and 3.25 thereof:

‘3.24       A comparable uncontrolled transaction is a transaction between two independent parties that is comparable to the controlled transaction under examination. It can be either a comparable transaction between one party to the controlled transaction and an independent party (“internal comparable”) or between two independent enterprises, neither of which is a party to the controlled transaction (“external comparable”).

3.25             Comparisons of a taxpayer’s controlled transactions with other controlled transactions carried out by the same or another MNE group are irrelevant to the application of the arm’s length principle and therefore should not be used by a tax administration as the basis for a transfer pricing adjustment or by a taxpayer to support its transfer pricing policy.’

400    In that context, the Commission acknowledges that, in recitals 342 to 361 of the contested decision, the purpose of the analysis had not been to perform a rigorous transfer pricing analysis to establish the arm’s length price for green coffee at the time that the APA had been requested. Nevertheless, as has been set out in paragraph 154 above, it was for the Commission to justify the choice of the transfer pricing method it deemed appropriate in the case at hand in order to examine the level of transfer pricing for an intra-group transaction.

401    The Commission’s assertion that, for the purposes of its assessment, it had no need to identify comparable external transactions for green coffee beans since it had ‘understood’ that the [confidential]% average mark-up for the period from 2005 to 2010 would correspond to an arm’s length mark-up in 2008 does not suffice as justification in that regard. The purpose of the comparison of the controlled transaction with comparable external transactions, for the period after 2011, is to determine whether the controlled transaction was at arm’s length and the fact that another controlled transaction is presumed to be at arm’s length, for the period between 2005 and 2010, does not enable the examination of comparable external transactions for the period after 2011 to be avoided. The mere fact that, according to the Commission, the Starbucks correspondents did not provide any valid justification for the increase in the mark-up from 2011 onwards does not prove that the price of the green coffee beans paid by SMBV to SCTC for the fiscal years from 2011 onwards was set at a higher level than the prices that other comparable operators on the market would have had to pay.

402    Those considerations suffice to conclude that the second complaint set out in paragraph 374 above should also be upheld.

403    Consequently, as has been set out in paragraphs 391 and 402 above, it is necessary to uphold the complaint that the third line of reasoning concerns an element of SMBV’s costs that was outside the scope of the contested measure and that, moreover, the Commission has not demonstrated, by that line of reasoning, the existence of an advantage within the meaning of Article 107(1) TFEU in favour of SMBV.

404    Accordingly, it is necessary to uphold the plea alleging that, under its first to third lines of reasoning, the Commission has not demonstrated that the APA conferred an advantage on SMBV, within a meaning of Article 107(1) TFEU.

E.      Dispute of the subsidiary reasoning regarding the existence of a tax advantage in favour of SMBV (recitals 362 to 408 of the contested decision)

405    The fourth plea in Case T‑760/15 and the third part of the second plea in Case T‑636/16 concern the Commission’s subsidiary reasoning on the existence of an advantage, consisting in demonstrating that, even assuming that the TNMM could be used for the purposes of determining SMBV’s taxable profits, the detailed rules for the application of that method to SMBV, as validated in the APA, would be erroneous.

406    That subsidiary reasoning is divided into two parts. In the first part, the Commission found that the choice of SMBV as the ‘tested party’ for the purposes of applying the TNMM, rather than Alki, was incorrect (fourth line of reasoning). In the second part, the Commission took the view that, even assuming that the tested party was indeed SMBV, the profit margin of SMBV obtained after applying the TNMM is not at arm’s length. First, the Commission considered that the choice of operating costs as profit level indicator was incorrect (fifth line of reasoning). Second, it found that, in any event, the adjustments applied to the profit margin in order to increase SMBV’s comparability with comparable undertakings were inappropriate (sixth line of reasoning).

407    The TNMM, to which the Commission refers in recitals 72 to 74 of the contested decision, is an indirect transfer pricing method. It consists in determining, from an appropriate base, the net profit realised by a taxpayer in a controlled transaction or in controlled transactions that are closely linked or continuous. To determine that appropriate base, it is necessary to choose a profit level indicator, such as costs, sales or assets. The net profit indicator obtained by the taxpayer in a controlled transaction must be determined by reference to the net profit indicator realised by the same taxpayer or a stand-alone undertaking in comparable uncontrolled transactions. The TNMM involves identifying a party to the transaction for which an indicator is tested.

408    The Kingdom of the Netherlands and Starbucks, which maintain that the TNMM was applied correctly, dispute all the criticisms made by the Commission as part of its subsidiary reasoning on the existence of an advantage.

409    First, the first part of the fourth plea in Case T‑760/15 and the first complaint of the third part of the second plea in Case T‑636/16 concern the identification of SMBV as the least complex entity.

410    Second, the second and third parts of the fourth plea in Case T‑760/15 and the second complaint of the third part of the second plea in Case T‑636/16 concern the identification of SMBV’s main functions and the determination of SMBV’s profits on the basis of operating costs.

411    Third, the fourth part of the fourth plea in Case T‑760/15 and the third complaint of the third part of the second plea in Case T‑636/16 concern the choice of adjustments intended to increase SMBV’s comparability with comparable undertakings. It will be appropriate to examine each of those complaints in turn.

412    Moreover, Starbucks argues that the Commission’s subsidiary reasoning on the existence of an advantage (recitals 362 to 408 of the contested decision) is vitiated by inadequate reasoning. It complains that the Commission criticised the way in which the TNMM was applied but did not prove that a better application of that method would have resulted in a higher profit for SMBV.

1.      Identification of SMBV as the most complex entity (fourth line of reasoning)

413    By the first part of the fourth plea in Case T‑760/15, the Kingdom of the Netherlands claims that the Commission’s argument that, since SMBV was the most complex entity, it could not be identified as the tested party for the purpose of applying the TNMM is erroneous. It maintains that the choice of SMBV for the application of the TNMM was correct. First, the mere fact that Alki was the holder of the roasting IP as well as the Starbucks brand for the EMEA region justified not designating it as the tested party for the purpose of applying the TNMM. Second, SMBV’s functions are less complex than Alki’s. None of the arguments relied on by the Commission in the contested decision relating to the functions and risks assumed by SMBV is such as to call that finding into question. In addition, the Kingdom of the Netherlands maintains that the Commission did not calculate the profit that ought to have been allocated to Alki if the TNMM had been applied to it and, consequently, that it did not demonstrate that the application of the TNMM it advocates would have led to a higher profit for SMBV.

414    Under the first complaint of the third part of the second plea in Case T‑636/16, Starbucks argues that SMBV was rightly classified in the transfer pricing report as being the less complex entity as compared to Alki. First, it argues that SMBV carries out routine, low-risk coffee roasting and conditioning activities as well as administrative and logistical support. Second, Starbucks considers that Alki is necessarily the more complex entity, since it exploits the roasting IP – which the Commission does not dispute – and it bears the risks in relation to SMBV’s activities, in accordance with the terms of the roasting agreement. Starbucks criticises the Commission for not having carried out a proper analysis of SMBV’s or Alki’s roles.

415    In addition, Starbucks claims that the contested decision is vitiated by inadequate reasoning. It argues that the Commission has not demonstrated that its misclassification of SMBV as the less complex entity conferred an advantage on it. It also argues that the contested decision does not state what SMBV’s taxable income would have been had Alki been classified as the less complex entity, or how it should have been applied to Alki.  

416    The Commission disputes those arguments. It maintains that it demonstrated, to a requisite legal standard, in the contested decision, that the choice of SMBV as tested entity for the purposes of the application of the TNMM was incorrect and did not enable a reliable approximation of an arm’s length outcome to be reached.

417    First, the Commission argues that the circumstance that the transfer pricing report does not contain a complete functional analysis of SMBV and Alki is sufficient to consider that the method used in the APA does not enable an arm’s length outcome to be reached. Second, the Commission maintains that the OECD Guidelines do not support the position of the Kingdom of the Netherlands and Starbucks on the choice of tested party. Third, the Commission contends that the complexity of the tested party is relative to that of the other entity involved in the transaction to be tested and that, in that perspective, Alki is less complex than SMBV. Fourth, the Commission maintains that the argument that it failed to conduct a proper functional analysis of SMBV and Alki is inadmissible, in so far as it is a new argument put forward for the first time in the reply. It contends that, in any event, that argument is unfounded.

418    Fifth, regarding the inadequate reasoning raised by Starbucks, the Commission retorts that it concluded, in recital 377 of the contested decision, that, as the TNMM’s application was premissed on a flawed assumption, it did not enable a reliable approximation of a market-based outcome to be achieved and thus conferred an advantage on SMBV. It specifies that, had Alki been identified as the more complex entity, the analysis of its functions would have shown that it was not entitled to any remuneration, such that all the profits would have remained with SMBV.  

419    In essence, the parties are in disagreement, first, on whether the transfer pricing report, as validated in the APA, correctly identified SMBV as being the tested entity for the purposes of the TNMM and, second, on whether the Commission stated sufficient reasons why it considered that the error in the identification of the tested party led to a reduction in SMBV’s taxable profit.

420    It is necessary to examine in the first place, irrespective of whether the tested entity was SMBV or Alki, whether the Commission has satisfied its obligation to state reasons.

421    In that regard, it is settled case-law that the statement of reasons required by Article 296(2) TFEU must be appropriate to the act at issue and must disclose in a clear and unequivocal fashion the reasoning followed by the institution which adopted the measure in question in such a way as to enable the persons concerned to ascertain the reasons for the measure and to enable the competent court to exercise its power of review. The requirements to be satisfied by the statement of reasons depend on the circumstances of each case, in particular the content of the measure in question, the nature of the reasons given and the interest which the addressees of the measure, or other parties to whom it is of direct and individual concern, may have in obtaining explanations. It is not necessary for the reasoning to go into all the relevant facts and points of law, since the question whether the statement of reasons meets the requirements of Article 296 TFEU must be assessed with regard not only to its wording but also to its context and to all the legal rules governing the matter in question (see judgments of 15 July 2004, Spain v Commission, C‑501/00, EU:C:2004:438, paragraph 73 and the case-law cited, and of 22 January 2013, Salzgitter v Commission, T‑308/00 RENV, EU:T:2013:30, paragraphs 112 and 113 and the case-law cited).

422    In the case at hand, the Commission set out, in Section 9.2.3.4 of the contested decision, that the choice of SMBV as tested entity for the purposes of the application of the TNMM conferred an advantage on SMBV.

423    First of all, it noted, in recital 364 of the contested decision, that the transfer pricing report should have included a comparison of each of the parties in the related transactions.

424    Next, the Commission considered, in recitals 365 to 368 of the contested decision, that the choice of SMBV as tested party was not in line with the requirements of the 1995 and 2010 versions of the OECD Guidelines, pursuant to which the tested party is the one whose functions are the least complex among the entities that are party to the controlled transaction.

425    Last, after an analysis of the functions of SMBV and Alki, the Commission concluded that the transfer pricing report had erroneously classified SMBV as having the less complex function compared to Alki (recitals 369 to 376 of the contested decision).

426    In recital 377 of the contested decision, the Commission concluded that, since the methodology for determining SMBV’s tax base in the transfer pricing report was premissed on the flawed assumption that SMBV should be the tested party for the application of the TNMM, that methodology does not result in a reliable approximation of a market-based outcome in line with the arm’s length principle. The Commission adds that, since the APA’s endorsement of that methodology leads to a lowering of SMBV’s tax liability under the general Netherlands corporate income tax system as compared to stand-alone companies whose taxable profit under that system is determined by the market, the APA should be considered to confer a selective advantage on SMBV for the purposes of Article 107(1) TFEU.

427    As has been set out in paragraph 201 above, the mere finding of non-observance of the methodological requirements for the determination of transfer pricing is not sufficient to establish that there is State aid within the meaning of Article 107 TFEU. The Commission must also demonstrate that the methodological errors that it identified do not enable an approximation of an arm’s length outcome to be reached and that they resulted in a reduction of the taxable profit compared to a profit that would have been calculated in accordance with the arm’s length principle.

428    Accordingly, in the examination of advantage within the meaning of Article 107 TFEU, in order to satisfy its obligation to state reasons as set out in paragraph 421 above, the Commission was obliged to set out the reasons why it had found that the error as to the choice of the tested entity had had the effect of reducing the level of SMBV’s taxable profit in such proportions that it did not correspond to a reliable approximation of an arm’s length outcome, thereby leading to a reduction in SMBV’s tax burden.

429    It must be stated, however, that Starbucks is right to claim that the contested decision contains no element allowing an understanding of the reasons why the Commission found that the error in the identification of the tested entity for the purposes of the application of the TNMM conferred an advantage on SMBV.

430    First, while the Commission indicated in recital 377 of the contested decision that the error regarding the tested entity had conferred an advantage on SMBV, that recital does not contain sufficient reasons. As is apparent from paragraph 422 above, the Commission merely asserted that the error regarding the determination of the tested entity led to a reduction in taxable profit. It provides no element to establish that the application of the TNMM to Alki and the allocation of residual profits to SMBV led to a higher taxable profit for SMBV.

431    Second, the other recitals of the contested decision contain no element enabling an understanding of the reasons for which the Commission found that SMBV’s taxable profit would have been higher had the TNMM been applied to Alki rather than to SMBV.

432    In the light of those considerations, it must be stated that the Commission  did not set out the reasons why it considered that the choice of SMBV as tested party for the purposes of the application of the TNMM had led to a reduction in SMBV’s taxable profit. Therefore, the Commission did not demonstrate how that error conferred an advantage on SMBV, in breach of its obligation to state reasons, as follows from Article 296(2) TFEU. 

433    In any event, it must also be stated that the Commission’s reasoning on the choice of tested entity is erroneous. Even assuming that the Commission could review the APA in the light of the 1995 version of the OECD Guidelines, as were available in April 2008, and that it could infer the existence of an advantage from the non-conformity with the requirements contained in those guidelines, those guidelines do not lay down a strict rule on the identification of the tested party.

434    More specifically, as the Kingdom of the Netherlands and Starbucks rightly argue, paragraph 3.43 of the 1995 OECD Guidelines indicates that the associated company, to which the TNMM is applied, must be the company for which reliable data on the most directly comparable transactions can be identified. It is next specified that that will often entail selecting the associated enterprise that is the least complex of the enterprises involved in the controlled transaction and that does not own valuable intangible property or unique assets. It follows that the guidelines do not necessarily oblige choosing the least complex entity, but merely advocate choosing the entity for which the most reliable data are available.

435    The Commission does not, however, prove that more reliable data were available to apply the TNMM to Alki. It is necessary inter alia to note that, first, the purpose of the APA is to determine the level of taxable profit of SMBV and not that of Alki and that, second, Alki was a third party to the procedure aimed at determining SMBV’s fiscal situation in the Netherlands.

436    Moreover, the existence of that recommendation in no way means that the choice of either entity as tested entity will necessarily have an impact on the transfer price obtained and that the choice of the more complex entity as tested entity does not enable an arm’s length outcome to be achieved.

437    If the choice of the least complex entity as tested party purports to limit errors, it is not at all inconceivable that applying the TNMM to the more complex entity might lead to an arm’s length outcome. In addition, in so far as the residual profits are allocated to the other party, the outcome should in theory be the same no matter which entity is tested.

438    Accordingly, it is necessary to uphold the first part of the fourth plea in Case T‑760/15 and the first complaint of the third part of the second plea in Case T‑636/16, without it being necessary to examine Starbucks’ argument seeking to dispute the admissibility of certain arguments put forward by the Commission.

2.      Analysis of SMBV’s functions and determination of SMBV’s profit on the basis of operating costs (fifth line of reasoning)

439    By the second and third parts of the fourth plea in Case T‑760/15, the Kingdom of the Netherlands argues that the Commission erroneously considered, first, that SMBV’s main functions were the resale of coffee-derived products and non-coffee products as opposed to the roasting of coffee and, second, that operating costs were not the appropriate profit level indicator.  

440    First, the Kingdom of the Netherlands claims, in essence, that the Commission was wrong to consider that SMBV’s main function was resale rather than roasting. Second, the Kingdom of the Netherlands maintains that, since the Commission wrongly considered that SMBV’s main function was resale, its finding that the appropriate profit level indicator was sales is also incorrect. Third, the Kingdom of the Netherlands argues that the alternative comparability analysis proposed by the Commission in recitals 395 to 398 of the contested decision does not prove that the determination of SMBV’s profits on the basis of turnover would have resulted in a higher taxable profit for SMBV.

441    Under the second complaint of the third part of its second plea, Starbucks also criticises the Commission for wrongly considering that SMBV’s main function was the resale of non-coffee products rather than the roasting of coffee and thereby inferring that sales, and not operating costs, were the appropriate profit level indicator. It maintains in that regard that operating costs were the right profit level indicator for SMBV. In addition, Starbucks claims that the Commission has not demonstrated that the error in determining SMBV’s functions conferred an advantage on it, in so far as the Commission’s comparability analysis is vitiated by significant errors.

442    The Commission maintains that it correctly demonstrated that SMBV’s main function was resale and that, therefore, the relevant profit level indicator for SMBV was sales and not operating costs.

443    In the first place, the Commission contends that SMBV acted primarily as a reseller.

444    In the second place, the Commission disputes the arguments of the Kingdom of the Netherlands and of Starbucks purporting to demonstrate that it erroneously considered that sales were the relevant profit level indicator. It argues that, having established that SMBV’s main function was resale, it was right to criticise the Netherlands authorities for having validated the use of operating costs as profit level indicator and that it was entitled to regard sales as the relevant profit level indicator.

445    Furthermore, the Commission states that the profit from sales of products unrelated to coffee must be allocated to SMBV and cannot be paid, via a royalty, to Alki, which is not in a position allowing it to generate profits from the resale of non-coffee products.

446    In the third place, the Commission disputes the arguments of the Kingdom of the Netherlands and of Starbucks that its analysis of the points of comparison is vitiated by a number of errors.

447    In the fourth place, the Commission disputes the criticisms made by the Kingdom of the Netherlands and Starbucks according to which it did not demonstrate that a better application of the TNMM would have resulted in a higher taxable profit for SMBV.

448    First, the Commission argues that those criticisms are irrelevant to the assessment of the validity of the contested decision. It did not seem necessary to it to propose a recovery method for its subsidiary reasoning when it did not agree with using the TNMM in SMBV’s case.  

449    Second, the Commission maintains that those remarks are unfounded, given that it carried out an analysis on the basis of SMBV’s reselling function and calculated a remuneration for SMBV on the basis of a sales margin. Recital 400 of the contested decision, in which it acknowledged that that calculation was not intended to calculate SMBV’s arm’s length remuneration, does not call into question the fact that its comparability analysis was carried out to demonstrate that a better application of the TNMM would have given rise to a higher taxable profit for SMBV.

450    In essence, the parties are in disagreement on whether the Commission demonstrated that the errors it identified in respect of the analysis of SMBV’s functions and the choice of profit level indicator conferred an advantage on SMBV.

451    As a preliminary point, it must be found that, although, in the application in Case T‑760/15, those issues are the subject of two separate complaints dealt with in different sections, the issues of the identification of SMBV’s functions and the choice of profit level indicator are indissociable. It follows from recitals 386 and 400 of the contested decision that those two questions come under the same evidence according to which the APA conferred an advantage on SMBV.

452    As a first step, the Commission found that SMBV’s main function was the resale of non-coffee products and not coffee roasting (recitals 380 to 386 of the contested decision).

453    As a second step, on the basis of that finding, the Commission asserted that sales were a more appropriate profit level indicator than operating costs (recitals 387 to 391 of the contested decision). The Commission considered, in essence, that, for the period from 2008 to 2014, the choice of operating costs as profit level indicator did not reflect the significant increase in SMBV’s sales and profit derived from its reselling activity. The Commission inferred that the profit derived from the sales was unduly shifted to Alki by means of the royalty, as Alki was not in a position to generate those profits.

454    The indissociable nature of those two steps set out in paragraphs 452 and 453 above is apparent, first, from the fact that the Commission draws no conclusion on the existence of a selective advantage from the finding of the single error in the identification of SMBV’s functions and, second, from the fact that the Commission inferred the error in the choice of profit level indicator for SMBV’s profits from the error in the identification of SMBV’s functions.

455    As a third step, the Commission also sought to ‘illustrate’ the impact of the error in the determination of SMBV’s main functions and choice of profit level indicator for SMBV’s profits. In order to do that, it conducted its own functional analysis departing from the premiss that SMBV’s main function was resale (recitals 392 to 400 of the contested decision).

456    For the sake of clarity, it must be noted that, by that reasoning, first, the Commission does not call into question the choice of the TNMM in the case at hand and, second, it does not claim that the profit level indicator used in the APA – operating costs – should have included factors other than costs, but maintains that a profit level indicator wholly separate from costs should have been used in the APA.

457    In order to examine whether the Commission has managed to demonstrate that the choice of profit level indicator had led to an outcome not in conformity with the arm’s length principle, it is thus appropriate to examine, first of all, the demonstration by the Commission in the first and second steps (recitals 380 to 391 of the contested decision), and then its comparability analysis conducted in the third step of its reasoning (recitals 392 to 400 of the contested decision).

(a)    Choice of profit level indicator

458    In the contested decision, the Commission found that SMBV’s main function was the resale of non-coffee products. It based its reasoning primarily on the fact that, in 2007, only [confidential]% of SMBV’s income was derived from the sale of roasted coffee. By comparison, [confidential]% of SMBV’s income was derived from the sale of non-coffee products, which corresponded to what Starbucks identified as the provision of logistics and administrative services, and that a significant portion of SMBV’s employees were active in that activity.

459    On the basis of that finding, the Commission took the view that sales were the appropriate profit level indicator. In recital 387 of the contested decision, it first of all noted that, according to paragraph 2.87 of the 2010 version of the OECD Guidelines, sales or operating costs related to distribution might be an appropriate profit level indicator. Next, in recital 388 of the contested decision, the Commission found that, in the case at hand, sales were a more adequate indicator of SMBV’s profit generating reselling function, since its profits were generated and recorded through a margin on products distributed. In addition, according to the Commission, between 2008 and 2014, SMBV’s total sales practically tripled, while the ‘gross margin’ more than doubled over that same period and, by comparison, SMBV’s operating costs increased by only 6%. It thereby inferred that operating costs could not be regarded as an adequate profit level indicator. On the basis of that finding, in recital 389 of the contested decision, the Commission asserted that the payment of royalties to Alki, corresponding to residual profit, had the effect of shifting the profit SMBV derived from resale to Alki, when Alki, owing to its limited operating capacity, was not in a position to generate profits from that activity. It therefore concluded that the entirety of the profits should have been attributed to SMBV.

460    It must nevertheless be held that, even had the Commission not erred in finding that SMBV’s main function was the resale of non-coffee products, its analysis does not suffice to demonstrate that a profit level indicator based on operating costs could not lead to an arm’s length outcome.

461    First, it should be noted that, as the Commission itself found in recital 387 of the contested decision, it is apparent from paragraph 2.87 of the 2010 version of the OECD Guidelines that sales or operating costs related to distribution might be an appropriate profit level indicator. It follows that, even assuming the Commission’s premiss whereby SMBV’s main function was the resale of non-coffee products were correct, it is not inconceivable, in principle, that operating costs could constitute an appropriate profit level indicator.

462    To the extent that the Kingdom of the Netherlands disputes the Commission’s assessment that the resale of non-coffee products constituted an appropriate base for determining SMBV’s net profit, it should be recalled that it is apparent from the OECD Guidelines, on which the Commission based its analysis, and in particular from paragraphs 1.42, 3.2 and 3.26 of the 1995 version thereof, which correspond, in essence, to paragraphs 2.57, 2.58 and 3.9 of the 2010 version, that the TNMM consists in determining, from an appropriate base, the net profit realised by a taxpayer in a controlled transaction or in controlled transactions that are closely linked or continuous. It follows that the TNMM serves to determine the level of transfer pricing for a type of transaction or for closely linked or continuous transactions on the basis of an analysis of the main functions related to that transaction or to those transactions. However, its purpose is not to determine the level of profit for an undertaking’s overall activity, consisting in types of varied transactions, on the basis of the identification of a single main function, by disregarding the other functions performed by that undertaking. Such a practice would not be in line with paragraph 3.4 of the 1995 version of the OECD Guidelines, which corresponds to paragraph 2.7 of the 2010 version of those guidelines, which provides the following: 

‘In no case should transactional profit methods be used so as to result in over-taxing enterprises mainly because they make profits lower than the average, or in under-taxing enterprises that make higher than average profits. There is no justification under the arm’s length principle for imposing additional tax on enterprises that are less successful than average when the reason for their lack of success is attributable to commercial factors.’

463    In that context, first, it must be borne in mind that, in the contested decision, the Commission argued that SMBV’s functions relating to the resale of coffee-related products and those relating to roasting were not of negligible importance. Consequently, both of those functions – and not one or the other – had to be taken into account for the purposes of determining SMBV’s remuneration.

464    Second, in any event, in the contested decision, the Commission did not demonstrate that, in the circumstances of the case at hand, all of SMBV’s intra-group transactions that were relevant to the determination of its taxable profit were closely linked or continuous, such that a single pricing level could be determined for their remuneration.

465    That finding is sufficient to reject the position of the Commission according to which sales of non-coffee products were a profit level indicator that could be used for the entirety of SMBV’s activity.

466    Second, in any event, the arguments of the Commission seeking to call into question, in the case at hand, the use of operating costs as profit level indicator are unconvincing.

467    First, it must be stated that the Commission’s analysis, set out in recitals 388 and 389 of the contested decision, is based on data subsequent to the APA’s conclusion. As has been found in paragraph 251 above, however, in the circumstances of the case at hand, the Commission could not base its analysis on information that was not available or reasonably foreseeable in April 2008, when the APA was concluded. In the present case, the Commission has not demonstrated that the data relating to SMBV’s sales and its costs over the period from 2008 to 2014 were reasonably foreseeable, such that it was not entitled to base its analysis on those data.

468    Second, even supposing that the data relating to SMBV’s activity between 2008 and 2014 could be used by the Commission, it must be stated that the assertion that SMBV’s sales tripled while its operating costs increased by only 6% over the same period does not suffice to call into question the choice of operating costs as profit level indicator.

469    It must be recalled that, as has been found in paragraph 458 above, the Commission’s argument is based on the premiss that SMBV’s main function is the resale of non-coffee products. First, however, the figures relied on by the Commission concern, as it itself indicates, SMBV’s total sales and ‘gross margin’, which necessarily includes sales of coffee and coffee-related products. In addition, regarding ‘gross margin’, it is equal to gross profit, that is to say, sales turnover minus costs of goods sold, divided by sales (see footnote 70 of the contested decision) and thus does not constitute a rate indicating the profitability of sales before deduction of fixed costs. The Commission does not explain how those figures might be usable or relevant in the present case. Nor does it provide any evidence in support of those figures, or any indication as to their source.  

470    Third, the profit level indicator proposed by the Commission, namely, total sales, does not appear to be appropriate for determining SMBV’s remuneration, either.

471    As has been set out in paragraph 458 above, the Commission based its line of argument on the premiss that [confidential]% of SMBV’s income was derived from the function of reselling non-coffee products. It thereby inferred that that function was SMBV’s main function.

472    However, it is important to note that that figure on which the Commission based its reasoning concerns SMBV’s income, and not profits. It must be stated that a high proportion of income does not necessarily translate into a high proportion of profits, such that that finding alone is not sufficient to prove that SMBV’s main function is the resale of non-coffee products.

473    Furthermore, the probative value of that figure is all the more debatable given that, as has been found in paragraphs 275 to 277 above, the Commission should have taken into account the fact that a portion of the income and profits was derived from the sale of roasted coffee by third parties.

474    In the light of the findings made in paragraphs 458 to 473 above, it must be held that the Commission has not demonstrated to a requisite legal standard that the choice of operating costs as profit level indicator did not allow a reliable approximation of a market-based outcome to be reached.

475    Since the Commission has not demonstrated that the choice of profit level indicator was incorrect, it could not take the view, in recital 389 of the contested decision, that a portion of SMBV’s profits, relating to its resale activity, was unduly shifted to Alki by means of the royalty. It has not demonstrated that SMBV’s profit would have been higher than the level of profit determined under the application of the APA.

(b)    Commission’s comparability analysis

476    As has been found in paragraph 455 above, it must be stated that the Commission carried out, in recitals 392 to 399 of the contested decision, its own comparability analysis departing from the premiss that SMBV’s main function was the resale of non-coffee products.

477    The Commission sought to determine the arm’s length range for SMBV, by comparing it to undertakings whose main function was the wholesale of coffee-derived products and using sales as profit level indicator.

478    In order to do so, the Commission reproduced the analysis of the tax advisor with a corrected group of comparable companies, which it referred to as a ‘corrected peer group’, identified on the basis of SMBV’s resale functions, then calculated, from the corrected peer group, the range of return on sales that corresponded, in the light of its analysis, to an arm’s length outcome. The interquartile range obtained for the return on sales corresponded to a range of between 1.5% and 5.5%. The Commission then applied it to SMBV’s results obtained from 2007 to 2014. It found that, for each of the years covered by the APA, SMBV’s tax base calculated on the basis of the APA was lower than the quartile range of SMBV’s tax base, as followed from applying the method selected by the Commission.  

479    The Commission’s approach consisting, first, in conducting its own analysis and, second, in comparing SMBV’s situation under the APA with the results of its analysis is such as to satisfy the Commission’s requirements regarding the demonstration of the existence of an advantage. The results of the Commission’s analysis demonstrate that SMBV’s taxable profit, as obtained under the APA for the years 2007 to 2014, is lower than SMBV’s taxable profit, as calculated for the years 2007 to 2014, applying the arm’s length range obtained by the Commission, from the corrected peer group.

480    However, in the first place, it must be held that, as the Kingdom of the Netherlands and Starbucks argue, the Commission’s comparability analysis is lacking in reliability.

481    First, it must be pointed out that, in recital 400 of the contested decision, the Commission clarified that the ‘purpose of [its analysis] ... [wa]s not to calculate an arm’s length remuneration for the functions performed by SMBV within the Starbucks group’. It thus ‘acknowledge[d] that the range presented above [wa]s not backed by a sufficient comparability analysis’. Such a clarification – made by the Commission itself – weakens the probative value of its analysis for demonstrating that the errors identified as regards SMBV’s functions and the choice of profit level indicator led to an advantage being conferred on SMBV.

482    Second, the impossibility, alleged by Starbucks, of replicating the corrected peer group search, done by the Commission, and of obtaining the same results as the Commission is such as to confirm the lack of reliability of the Commission’s comparability analysis. When Starbucks’ tax advisor attempted to replicate the Commission’s comparability analysis using the same criteria as the Commission, it obtained a list of 87 companies. Of the 12 companies identified by the Commission for the purpose of its comparability analysis, however, only 3 were present in the list of 87 companies.

483    It is true that the Commission tried to replicate, at the stage of the defence in Case T‑636/16, the corrected peer group search, in order to demonstrate the reliability of its comparability analysis. However, even supposing that using the ‘Orbis’ database instead of the ‘Amadeus’ database had made no difference, since the first database contains the same data as the second, it must be pointed out that five of the companies identified in recital 394 of the contested decision did not appear when it replicated its search of comparable companies. The Commission has moreover admitted this, in paragraph 179 of its defence in Case T‑636/16.

484    The arguments relied on by the Commission to justify that difference in results between its own comparability analysis and the replication of that analysis are therefore incapable of undermining the finding of lack of credibility and reliability of its comparability analysis. The Commission maintains that that difference in results is explained by the fact that those five companies were reclassified in the database subsequent to its comparability analysis.

485    First, it is apparent from Starbucks’ reply – without being contradicted on that point by Commission in the rejoinder in Case T‑636/16 – that it is possible to consult the historical versions of the ‘Orbis’ and ‘Amadeus’ databases, such that the change in the companies’ situation should not have any impact on the replicability of the Commission’s comparability analysis. Therefore, since those historical versions of the ‘Amadeus’ database cannot be updated retroactively, the results could not have been different from those obtained in the Commission’s comparability analysis.

486    Second, Starbucks claims that the Commission used, both for its own comparability analysis and for the replication of the comparability analysis, versions of the ‘Amadeus’ and ‘Orbis’ databases dating from 2015 and 2017, respectively, which the Commission does not call into question. It follows that the Commission’s analysis is based on versions of the databases subsequent to 2008. To the extent that, as the Commission itself maintains, the classification of companies can vary depending on the version of the database, the results of the comparability analysis could be distorted by using a more recent version. Furthermore, as is apparent from paragraphs 243 to 251 above, only information that was available on the day of the adoption of the APA could be taken into account by the Commission.

487    Therefore, the circumstances, on the one hand, that the Commission was not in a position to replicate its comparability analysis and, on the other hand, that those five companies represented a significant portion of the corrected peer group, used for the comparability analysis, as well as the resulting impossibility for the Kingdom of the Netherlands and Starbucks or for this Court to know the precise method used by the Commission in its reasoning and to reproduce that analysis in order to verify whether it correctly identified the comparable companies are such as to call into question its reliability and credibility.

488    In the second place, in any event, it must be held that, as the Kingdom of the Netherlands and Starbucks argue, the Commission’s analysis is vitiated by a number of errors.

489    First, it must be stated that the corrected peer group used by the Commission for its comparability analysis is inconsistent with the findings it made as to SMBV’s functions and is not such as to prove that the errors it identified led to a reduction in SMBV’s taxable profit.

490    It is important to point out, first, that the Commission found, in the contested decision, that SMBV’s main function was the resale of non-coffee products. In recital 382 of the contested decision, the Commission clearly stated that SMBV’s main function was resale, [confidential]% of the company’s income in 2007 having derived from that activity. In recital 384 of that decision, the Commission clarified its position according to which the majority of SMBV’s activities related to the sale or resale of non-coffee products, such as cups and paper napkins. That finding is, moreover, supported by the Commission’s submissions, in which the Commission asserts that SMBV’s main function is reselling non-coffee products and that it is the main reason why it criticised the Starbucks group’s tax advisor for having chosen operating costs as profit level indicator.  

491    Second, the Commission set out in recitals 393 and 394 of the contested decision that, as SMBV’s functions had been incorrectly identified in the transfer pricing report, the peer group used to apply the TNMM, identified by the NACE (Nomenclature générale des activités économiques dans les Communautés européennes) code ‘processing of tea and coffee’ was inappropriate. The Commission then reproduced the analysis carried out in the transfer pricing report, using the corrected peer group, identified by the NACE code ‘wholesale of coffee, tea, cocoa and spices’. It then excluded companies mainly distributing products other than tea and coffee from the corrected peer group. The result was a corrected peer group made up of 12 companies, each engaged in roasting, as the Commission found in recital 394 of the contested decision.

492    It must be pointed out that the companies making up the corrected peer group have different functions than SMBV’s main function, as identified by the Commission, namely, the resale of non-coffee products. This means that those companies are not in a comparable situation to SMBV’s. Those companies therefore cannot be regarded as relevant for calculating the profit SMBV would generate under market conditions. Consequently, the alternative comparability analysis consisting in replicating the tax advisor’s analysis by using a corrected peer group engaged in the sale of coffee and roasting is necessarily incorrect.

493    Second, it must be held, as Starbucks argues, that, even supposing that the corrected peer group could be used by the Commission, the results of the comparability analysis carried out by the Commission are necessarily distorted, since it compared incomparable data, comparing the operating profits of comparable companies with SMBV’s taxable profit.

494    In that regard, first, it must be noted that it is settled between the parties that the interquartile range calculated by the Commission for the period from 2005 to 2007, corresponding to a range of between 1.5% and 5.5% of sales, was calculated on the basis of the operating profits of the companies comprising the corrected peer group. That finding is moreover supported by Table 12 of the contested decision. Second, it is settled that it is SMBV’s taxable profit, determined under the APA, and not its operating profit that the Commission compared with the operating profit of the comparable companies from the corrected peer group. That is moreover apparent from Table 13 of the contested decision.

495    The Commission does not dispute, however, that operating profits are not comparable to pre-tax profits, merely asserting that it replicated the analysis of the Starbucks group’s tax advisor. Furthermore, it must be stated that operating profits and taxable profits are two separate notions which are reflected, in principle, in the recording of different amounts in the corresponding accounting tables, as is apparent from recital 82 of the contested decision and from Table 1 of that decision.

496    The circumstance that the Commission asserted, in recital 397 of the contested decision, that it compared SMBV’s taxable profit calculated on the basis of the APA with the taxable profit calculated on the basis of the interquartile range determined by the Commission cannot call into question the finding made in paragraph 493 above. Given that the interquartile range was calculated on the basis of the operating profits of comparable undertakings, the result obtained in respect of SMBV when applying that range corresponds not to its taxable profit, but to its operating profit.

497    It follows that the comparison of SMBV’s taxable profit with the interquartile range obtained from the operating profit of the companies of the corrected peer group is necessarily distorted.

498    Moreover, for the year 2007-2008, the 1.2%  figure is fairly close to the lower part of the range calculated by the Commission. In view of the numerous approximations in the Commission’s analysis, that result does not demonstrate a situation clearly contrary to market conditions. It should be borne in mind that, where the Commission checks whether the taxable profit of an integrated undertaking in application of a tax measure corresponds to a reliable approximation of a taxable profit realised under market conditions, it may determine the existence of an advantage within the meaning of Article 107(1) TFEU only provided that the discrepancy between the two factors of comparison goes beyond inaccuracies inherent to the method used to obtain that approximation.

499    Furthermore, even assuming that the error consisting in comparing SMBV’s taxable profit with the operating profit of comparable companies was indeed contained in Starbucks’ transfer pricing report – which Starbucks disputes – the existence of that error in the transfer pricing report does not prevent the Court from verifying that the contested decision is not vitiated by error. In addition, if the Commission deemed problematic the fact that operating profits were compared with taxable profits, it ought to have examined that matter in the contested decision.

500    It is therefore necessary to consider, on the basis of the findings made in paragraphs 480 to 499 above, that the comparability analysis conducted by the Commission in recitals 392 to 399 of the contested decision, first, lacks reliability and, second, is vitiated by a number of errors.

501    Consequently, in the light of the considerations set out in paragraphs 457 to 500 above, it is necessary to uphold the complaints of the Kingdom of the Netherlands and of Starbucks according to which the Commission has not demonstrated that the APA’s validation of SMBV’s functions and of the choice of profit level indicator, proposed in the transfer pricing report, conferred an advantage on SMBV. Accordingly, it is no longer necessary to examine whether the Commission was correct to find that the identification of SMBV’s functions and the choice of profit level indicator used in the APA were erroneous. It is therefore not necessary to examine the argument of Starbucks disputing the admissibility of certain arguments put forward by the Commission.

3.      Choice of adjustments (sixth line of reasoning)

502    Under the fourth part of the fourth plea in Case T‑760/15, and the third complaint of the third part of the second plea in Case T‑636/16, the Kingdom of the Netherlands and Starbucks claim, in essence, that the Commission has not demonstrated that the adjustments proposed in the transfer pricing report to increase the comparability between SMBV and the comparable companies were such as to confer an advantage on SMBV.

503    The Kingdom of the Netherlands argues that the Commission erroneously found that two of the adjustments proposed in the transfer pricing report for increasing the comparability between SMBV and the 20 comparable unrelated companies did not enable an approximation of an arm’s length outcome to be reached. First, the exclusion, from the relevant cost base, of the costs of products related and unrelated to coffee is justified inter alia by the fact that SMBV acts as service provider, does not perform any purchasing function and does not bear stock-related risks, unlike the comparable companies. Second, the mark-up adjustment is justified by the fact that the mark-up before adjustment concerns operating profits, whereas the objective of the APA was to determine taxable profits. That adjustment had the effect of increasing the mark-up.

504    First, Starbucks adds that, in the contested decision, the Commission did not call into question the adjustments applied to the cost base, chosen as profit level indicator. Therefore, the argument of the Commission set forth in paragraph 183 of the defence in Case T‑636/16, according to which the cost-base adjustment is inappropriate due to the absence of risk transfer from SMBV to Alki, is inadmissible since it does not appear in the contested decision. Second, it considers that the figures presented by the Commission in paragraphs 184 and 185 of the defence in Case T‑636/16, purporting to show that SMBV’s taxable income would have been higher if a mark-up had been applied on total costs, are also inadmissible in so far as they do not appear in the contested decision.  

505    So far as concerns the adjustments at issue, Starbucks alleges inadequate reasoning. The Commission merely asserts that the adjustments were not appropriate, without demonstrating how SMBV’s taxable profit would have been higher if more appropriate adjustments had been made.

506    The Commission disputes those arguments. It maintains that the two adjustments proposed in the transfer pricing report are inadequate and lead to a reduction in SMBV’s taxable profit. It argues that the Kingdom of the Netherlands and Starbucks have not demonstrated that it committed an error of assessment.

507    First, regarding the adjustments applied to the cost base, the Commission argues that it did indeed challenge that point in recitals 319 to 332 of the contested decision, asserting that Alki could not bear any of SMBV’s entrepreneurial risk. The Commission refers also to recitals 59 and 159 of the contested decision, which  explain that the cost-base adjustment is justified, according to the transfer pricing report, by SMBV’s status as toll manufacturer, and not assuming any risk. In addition, the Commission disputes the arguments of the Kingdom of the Netherlands and of Starbucks according to which SMBV’s profit could be calculated on the basis of operating costs and not total costs.

508    Second, the Commission notes that, although the adjusted mark-up led to a higher percentage, that percentage was applied to a much narrower cost base. It adds that, given that the cost of green coffee beans, the remuneration paid to third parties and non-coffee products should have been included in the cost base, there was no reason to apply the ‘working capital adjustment’. Even assuming that the Starbucks group’s tax advisor, having prepared the transfer pricing report, committed no error in excluding those different costs from the cost base, the ‘working capital adjustment’ would not have been appropriate, either. In addition, the Commission argues that it adequately explained, in recitals 402 to 406 of the contested decision, how the ‘working capital adjustment’, together with the cost-base adjustment, lowered SMBV’s annual corporate income tax liability.

(a)    Preliminary observations

509    First of all, it must be pointed out that, in recitals 407 and 408 of the contested decision, the Commission considered that, even if SMBV’s functions and the profit level indicator had been correctly identified, two adjustments proposed in the transfer pricing report meant that the methodology proposed in the transfer pricing report did not result in an arm’s length outcome.

510    Based on the finding that the two adjustments were erroneous, the Commission concluded that, by accepting that methodology, which led to a lowering of SMBV’s tax liability under the general Netherlands corporate income tax system as compared to stand-alone companies whose taxable profit under that system is determined in market conditions, the APA conferred a selective advantage on SMBV for the purposes of Article 107(1) TFEU.

511    It should be noted that it is thus apparent from recitals 407 and 408 of the contested decision that the Commission’s approach, consisting in comparing SMBV’s taxable profit under the APA with that of a stand-alone company whose profit under the general Netherlands corporate income tax system is determined in market conditions, appears, at first sight, to satisfy the requirements incumbent on the Commission regarding the existence of an advantage.

512    However, it should be recalled that it follows from the considerations set out in paragraphs 151 and 152 above that, in order to determine whether the APA conferred an advantage on SMBV in the case at hand, it is for the Commission to show that the transfer pricing method endorsed in the APA led to a reduction of SMBV’s tax burden and, more specifically, to show that the level of SMBV’s profits, calculated under the transfer pricing method, is reduced to such an extent that it cannot be regarded as being a reliable approximation of a market-based outcome. As has been found in paragraph 498 above, where the Commission checks whether the taxable profit of an integrated undertaking in application of a tax measure corresponds to a reliable approximation of a taxable profit realised under market conditions, it may determine the existence of an advantage within the meaning of Article 107(1) TFEU only provided that the discrepancy between the two factors of comparison goes beyond inaccuracies inherent to the method used to obtain that approximation.

513    It is therefore appropriate to examine whether the Commission has sufficiently shown that the two adjustments made by the Starbucks group’s tax advisor conferred an advantage on SMBV.

(b)    Cost-base adjustment

514    The first adjustment proposed in the transfer pricing report concerns the cost base (the ‘cost-base adjustment’). It consists in excluding certain costs from the cost base used as profit level indicator for the purposes of the application of the TNMM. However, it must be stated that it is apparent from recitals 406 and 407 of the contested decision that the criticisms identified by the Commission concern only the exclusion of the costs of unaffiliated manufacturing company 1 from the cost base used for the application of the TNMM. In essence, the Commission found that there was no explanation why  the costs of unaffiliated manufacturing company 1 were excluded when they had been taken into account in the preceding APA.

515    In the first place, it should be recalled that, contrary to what the Commission argues, the conclusion in recital 407 of the contested decision, according to which the adjustments proposed in the transfer pricing report and validated in the APA confer an advantage on SMBV, is explicitly limited to excluding the costs of unaffiliated manufacturing company 1 from SMBV’s cost base. It is not apparent from the text of the contested decision that the Commission based the finding of advantage on the exclusion of other costs from the cost base used as SMBV’s profit level indicator.

516    The circumstance, relied on by the Commission, that that institution called into question, in recitals 319 to 332 of the contested decision, the fact that SMBV’s entrepreneurial risks were transferred to Alki does not permit the conclusion that it considered, for those reasons, that certain costs had been erroneously excluded from the cost base used as profit level indicator. That finding is supported by the circumstance that the issue of the adjustments is presented by the Commission itself as being a line of reasoning (see recital 407 of the contested decision) subsidiary to the line of reasoning examined in recitals 319 to 332 of the contested decision.

517    Moreover, contrary to what the Commission essentially maintains, it is not apparent from recitals 59 and 159 of the contested decision that that decision based the finding that the APA had conferred an advantage on SMBV on the cost-based adjustments. So far as concerns recital 59 of the contested decision, it must be stated that, although it does refer to the said adjustments, it merely presents the content of the transfer pricing report. As regards recital 159 of the contested decision, that recital – which is in the section introducing the administrative procedure – merely indicates that the Commission had expressed doubts as to the adjustments proposed in the transfer pricing report, without it being possible to infer the Commission’s position in the context of the contested decision.

518    On the basis of the findings made in paragraphs 515 to 517 above, it must therefore be held that the Commission did not contend in the contested decision – let alone prove – that the cost-based adjustment, apart from the exclusion of the costs of unaffiliated manufacturing company 1, had conferred an advantage on SMBV. Accordingly, it is necessary to reject the arguments of the Commission, presented in the defence, according to which the use of operating costs instead of total costs (including the cost of coffee beans, the fees paid to third parties and the costs of non-coffee products) led to a reduction in SMBV’s taxable base.  

519    In the second place, regarding the exclusion of the costs of unaffiliated manufacturing company 1, the Commission found, in recital 406 of the contested decision, that the transfer pricing report had accepted a considerable reduction in the cost base, by excluding those costs.

520    The Commission, however, is limited to asserting, in recital 406 of the contested decision, without any further clarification, that those costs had been taken into account in the previous arrangement for determining SMBV’s tax base, used prior to the APA’s conclusion, and that the exclusion of those costs had not been reasoned. It is not clear from the text of the contested decision to what the Commission is referring when it invokes an absence of reasoning of the exclusion of costs, and in particular whether it considers that such explanations should have been contained in the APA or provided during the administrative procedure.

521    In that regard, first, it must be held that the finding as to the insufficient justification of the adjustment, whether it be by the Starbucks correspondents or by the Netherlands authorities, does not suffice to demonstrate, as such, that that adjustment was erroneous, or that it led to a reduction in SMBV’s tax burden.

522    Second, in any event, it must be held that it is apparent from recital 407 of the contested decision that the Commission’s examination of the erroneous nature of the exclusion of the costs of unaffiliated manufacturing company 1 is a subsidiary analysis that would be at issue if SMBV’s main function was indeed the roasting of coffee.

523    First, it is apparent from the Commission’s defence in Case T‑636/16 that unaffiliated manufacturing company 1 mainly manufactured products such as flavoured coffee, powder for a coffee-based product protected by a registered trademark or soluble coffee and that it roasted green coffee beans only in ‘limited volumes’. The Commission did not, however, explain how the costs of unaffiliated manufacturing company 1 were relevant to calculating SMBV’s taxable profit, as roaster.

524    Second, it must be stated that the arguments put forward by the Commission in its submissions regarding the exclusion of the costs of unaffiliated manufacturing company 1 are based on the premiss that SMBV’s main activity is resale. Those different arguments must therefore be rejected.

525    Third, it is apparent from the transfer pricing report that the tax advisor excluded from the cost base, used for the purposes of applying the TNMM, the costs relating to activities for which SMBV does not provide added value. The Kingdom of the Netherlands and Starbucks moreover contend, in their respective submissions, that the exclusion of the costs of unaffiliated manufacturing company 1 is justified by the fact that SMBV does not provide any added value. They maintain that the costs relating to the transaction between SMBV and unaffiliated manufacturing company 1 merely pass through SMBV’s accounts, but cannot be attributed to SMBV’s activity. The purchase of the products of unaffiliated manufacturing company 1 thus constitutes a neutral transaction for the determination of that company’s taxable profit.

526    In that regard, it must be stated that it is not ruled out that the income derived from the products provided by unaffiliated manufacturing company 1 are equivalent to the costs of unaffiliated manufacturing company 1, such that SMBV does not generate any profit from that company’s products. The Commission, however, has not demonstrated that SMBV added any value to the products of unaffiliated manufacturing company 1 and that it did indeed generate profit from the exploitation of those products, such that the costs of unaffiliated manufacturing company 1 should necessarily have been taken into account for the purposes of the application of the TNMM.

527    Nor does the Commission demonstrate that the differences cited in the transfer pricing report between the functions of SMBV and of the 20 companies on the basis of which the comparability analysis was conducted do not justify the application of the adjustment involving the exclusion of the costs of unaffiliated manufacturing company 1.

528    To the extent that the Commission does not provide evidence showing that SMBV generated profit from the transaction with unaffiliated manufacturing company 1 or that the mark-up should apply to a cost base including the costs of unaffiliated manufacturing company 1, it must be held that it was not entitled to conclude that the exclusion of those costs was erroneous and led to a reduction in SMBV’s profit.

529    In the third place, it must be stated that, as Starbucks argues, the figures contained in the table produced in paragraph 184 of the Commission’s defence in Case T‑636/16, which are calculations based on the figures contained in Table 3 of the contested decision, cannot be taken to support the Commission’s position. First, those figures relate to SMBV’s total costs (operating expenses and COGS) and not only to the operating costs to which the costs of unaffiliated manufacturing company 1 would have been added. Second, those figures demonstrate only that the level of profits would have been higher if the cost base had been broader and do not support the proposition that SMBV would have made a profit from exploiting the products of unaffiliated manufacturing company 1.

530    In the fourth place, it must be noted that the exclusion of the costs of unaffiliated manufacturing company 1 was combined with the adjustment – increasing  – of the rate  of return. It therefore cannot necessarily be concluded that the adjustments applied in the APA, taken as a whole, necessarily led to a reduction in SMBV’s taxable base. The Commission, however, has not quantified the costs of unaffiliated manufacturing company 1 or, at least, the proportion of SMBV’s costs they represent. It is thus not apparent from the contested decision that the costs of unaffiliated manufacturing company 1 represent such a proportion of SMBV’s costs that merely excluding them would have an impact on SMBV’s profits to such an extent that their level would no longer be representative of a profit resulting from an arm’s length situation.

531    In the light of those observations, it must be held that the Commission has not managed to demonstrate that the exclusion of the costs of unaffiliated manufacturing company 1 conferred an advantage on SMBV, without it being necessary to examine whether the Commission vitiated its decision by a failure to state reasons.

(c)    ‘Working capital adjustment’

(1)    Scope of the adjustment at issue

532    With regard to the scope of the second adjustment, it must be stated that, in recital 407 of the contested decision, the Commission maintained that the application of the ‘working capital adjustment’ meant that the method proposed in the transfer pricing report did not result in a reliable approximation of a market-based outcome in line with the arm’s length principle. In that regard, it must be noted that neither the transfer pricing report nor the APA uses the expression ‘working capital adjustment’.

533    First of all, in the contested decision, the Commission maintained that, in the transfer pricing report, the Starbucks group’s tax advisor proposed a conversion mark-up adjustment, presented by the Netherlands authorities as a ‘working capital adjustment’ (recital 401 of the contested decision). It follows from that finding that the expression ‘working capital adjustment’ as used in the contested decision must be understood in the sense used by the Netherlands authorities during the administrative procedure.

534    Next, it is apparent from recital 403 of the contested decision that the Commission had expressed doubts on the ‘working capital adjustment’ in recitals 101 to 113 of the opening decision. First, it should be noted that, in recitals 101 and 102 of the opening decision, the Commission discussed the ‘raw material cost mark-up’, while the adjustment relating to the exclusion of the costs of green coffee from the cost base was discussed in recitals 99 and 100 of the opening decision. The contested decision therefore does not refer, in recital 403 thereof, to the latter adjustment. That finding is moreover supported by point (iii) of recital 269 and by footnote 130 of the contested decision.

535    It is true that recitals 103 to 113 of the opening decision also concern, in part, the adjustment relating to the exclusion of the costs of green coffee from the cost base. However, according to recital 107 of the opening decision, the arguments of the Netherlands regarding the ‘working capital adjustment’ are set out in recital 59 of the same decision. According to recital 59 of the opening decision, the Netherlands authorities stated that ‘the adjustment in this case [was] a combination of two comparability adjustments: it combine[d] a working capital adjustment for raw materials inventory on the return of the comparable companies with an adjustment for the raw material costs in the cost base of the comparable companies’. It is apparent from the description of the arguments of the Kingdom of the Netherlands during the administrative procedure that, for it, the expression ‘working capital adjustment’ concerned only the ‘raw materials cost adjustment’, identified in the transfer pricing report.

536    Last, it must be pointed out that, in recital 407 of the contested decision, the Commission itself notes a difference between the ‘working capital adjustment’ and the exclusion of the costs of unaffiliated manufacturing company 1 from SMBV’s tax base.

537    It must therefore be concluded that the expression ‘working capital adjustment’, used in recital 407 of the contested decision, makes reference to the ‘raw materials cost adjustment’, identified in the transfer pricing report.

538    In any event, even supposing that the expression ‘working capital adjustment’, used in recital 407 of the contested decision, were also be understood as making reference to the raw materials cost adjustment in SMBV’s cost base, it must be held that recitals 401 to 406 of the contested decision contain no argument on the cost base other than the one involving the exclusion of the costs of unaffiliated manufacturing company 1. It has already been found, in paragraphs 514 to 531 above, however, that the Commission has not managed to demonstrate that the exclusion of those costs conferred on advantage on SMBV. In recitals 404 and 405 of the contested decision, the Commission simply rejects the arguments of the Kingdom of the Netherlands on the relevance of a comparability study on the basis of total costs and a scientific article. Moreover, recitals 401 to 403 of the contested decision contain no reference to SMBV’s cost base.

(2)    Complaint alleging the absence of a reduction in SMBV’s tax burden

539    In the first place, it should be recalled that, in so far as, first, the ‘working capital adjustment’ corresponds to the raw materials cost adjustment in the cost base, identified in the transfer pricing report (see paragraph 537 above) and, second, the argument regarding the exclusion of the costs of unaffiliated manufacturing company 1 has been rejected (see paragraphs 514 to 531 above), that adjustment translated into an increase in the cost base mark-up from [confidential]% to [confidential]%. The use of a higher mark-up for the purposes of determining SMBV’s taxable profit could not lead to a reduction in SMBV’s taxable profit. That adjustment alone, taken in isolation, is thus not such as to confer an advantage on SMBV.

540    It follows that the Commission has not managed to show that the ‘working capital adjustment’ had the effect of lowering the level of SMBV’s profits or, consequently, that that adjustment conferred an advantage on that company.

541    In the second place, it should be stated that the Commission’s reasoning regarding the ‘working capital adjustment’, set out in recitals 401 to 405 of the contested decision, is not capable of showing that the ‘working capital adjustment’ had the effect of lowering the level of SMBV’s profits and that it had, consequently, conferred an advantage on that company.

542    First of all, in so far as the Commission based its reasoning on the finding that the method used to determine the ‘working capital adjustment’ did not take into account the amount of the working capital of the comparable undertakings, or that of SMBV, it is sufficient to state that the Commission does not explain how that fact could demonstrate a lowering of the level SMBV’s profits.

543    Next, although the Commission found that there was no constant relation between the COGS used in the adjustment and working capital needs, it should be noted that the Commission has not explained how that fact could concretely demonstrate a lowering of the level of SMBV’s profits.

544    In addition, by its assertions that the ‘working capital adjustment’ made by the Starbucks group’s tax advisor is not suited to the declared purpose of adjusting for differences in working capital use, the Commission is limited to general and approximate considerations, such as its contention that that adjustment is ‘ill-fitted’ or that ‘a company with a high amount of raw material cost might have low working capital needs if it processes its stock efficiently’.

545    Last, regarding the finding, in recitals 402 to 405 of the contested decision, that there is no justification for the ‘working capital’ adjustment in the set of facts presented in the transfer pricing report or in any of the arguments raised by the Kingdom of the Netherlands in the context of the administrative procedure, it must be held that the mere finding of the absence of such justification similarly does not demonstrate that the ‘working capital adjustment’ led to a reduction in SMBV’s taxable profit.

546    It follows that, contrary to what the Commission concluded in recital 407 of the contested decision, it has not proved that the ‘working capital adjustment’ led to a reduction in SMBV’s taxable profit.

547    That finding cannot be called into question by the Commission’s arguments. It must be stated that it is apparent from recital 407 of the contested decision that the Commission’s examination of the ‘working capital adjustment’ is a subsidiary analysis that would be at issue if SMBV’s main function was indeed the roasting of coffee. The arguments put forward by the Commission, however, in its submissions, regarding the ‘working capital’ adjustment, are based on the premiss that SMBV’s main activity is resale. Those various arguments must therefore be rejected.

548    In the light of the considerations set out in paragraphs 502 to 547 above, it is necessary to uphold the complaints of the Kingdom of the Netherlands and of Starbucks according to which the Commission has not demonstrated that the APA’s validation of the working capital adjustment and that relating to the exclusion of the costs of unaffiliated manufacturing company 1 conferred an advantage on SMBV.

549    Accordingly, it is necessary to uphold the plea alleging that, under its fourth to sixth lines of reasoning, the Commission has not demonstrated that the APA conferred an advantage on SMBV, within the meaning of Article 107(1) TFEU.

F.      Whether the APA derogated from Article 8b of the CIT and from the transfer pricing decree (reasoning in respect of the limited reference system, recitals 409 to 412 of the contested decision).

550    The Kingdom of the Netherlands claims that it puts forward its pleas regarding the absence of an advantage in the case at hand against both the Commission’s main position – the first six lines of reasoning – and its reasoning in respect of the limited reference system, in which the Commission concluded that there was an advantage, in the present case, under Article 8b of the CIT and under the transfer pricing decree. Starbucks, for its part, argues that the Commission should have examined the APA in the light of Article 8b(1) of the CIT and the transfer pricing decree, which it did not do.

551    The Commission maintains that it assessed, in recitals 409 to 412 of the contested decision, the APA against Article 8b(1) of the CIT and that it found, following that assessment, that the APA conferred a selective advantage on SMBV.

552    It must be observed in that regard that, as a further subsidiary point, in Section 9.2.4 of the contested decision, entitled ‘Subsidiary line of reasoning: Selective advantage due to a derogation from the [transfer pricing] Decree’ (recitals 409 to 412 of the contested decision), the Commission found that the APA conferred an advantage on SMBV under an assessment based on the limited reference system of Article 8b(1) of the CIT and the transfer pricing decree (recital 412 of the contested decision).

553    In recital 410 of the contested decision, the Commission stated that, ‘in a subsidiary line of reasoning, ... the SMBV APA also grant[ed] SMBV a selective advantage in the context of the more limited reference system composed of group companies applying transfer pricing to which Article 8b(1) of the CIT and the [transfer pricing] Decree appl[ied]’. In recital 411 of the contested decision, the Commission recalled that Article 8b(1) of the CIT and the transfer pricing decree had established ‘the “arm’s length principle” under Dutch tax law, according to which transactions between intra-group companies should be remunerated as if they were agreed to by independent companies negotiating under comparable circumstances at arm’s length’. In the same recital, the Commission noted that the preamble to the transfer pricing decree explained that the OECD Guidelines applied directly to the Netherlands. In recital 412 of the contested decision, the Commission referred to the reasoning set out in recitals 268 to 274 of the contested decision, which summarise the first six lines of reasoning, to conclude that the APA also gave rise to a selective advantage under the more limited reference system of Article 8b(1) of the CIT and the transfer pricing decree.

554    It is clear from those findings that the Commission concluded that the APA at issue conferred a selective advantage on SMBV, since it resulted in a lowering of the tax liability as compared to the situation where the arm’s length principle laid down in Article 8b CIT and in the transfer pricing decree was properly applied.

555    It must be noted that the Commission based that conclusion on its examination of the APA conducted as part of its primary analysis. It thus asserted that it had already demonstrated, in Section 9.2.3.1 of the contested decision, that the APA did not enable a reliable approximation of an arm’s length outcome to be reached.

556    It is true that the reasoning set out in recitals 409 to 412 of the contested decision concerns, first and foremost, an argument of the Kingdom of the Netherlands and of Starbucks on the choice of reference system, forming part of the analysis of the selectivity of the measure at issue.

557    However, it must be pointed out that the Kingdom of the Netherlands and the Commission are of the view that recital 412 of the contested decision must be interpreted as meaning that the latter concluded, on the basis of an examination under the relevant national law – Article 8b(1) of the CIT and the transfer pricing decree – that the APA conferred an advantage on SMBV, the analysis conducted by the Commission in the first six lines of reasoning applying mutatis mutandis. That finding is moreover supported by the wording of recital 416 of the contested decision.

558    Without it being necessary, in the case at hand, to take a position on the exact nature and scope of the reasoning in respect of the Commission’s limited reference system, set out in recitals 409 to 412 of the contested decision, it is sufficient to state that, assuming that, by that reasoning, the Commission has examined the errors it had identified in the context of the first six lines of reasoning under Article 8b of the CIT and the transfer pricing decree, which fall under the arm’s length principle under Netherlands law, the Commission has not demonstrated, for the same reasons as those set out in paragraphs 173 to 549 above, which apply mutatis mutandis to such an examination, that the APA had conferred an advantage on SMBV, within the meaning of Article 107(1) TFEU.

G.      Conclusion

559    First, it follows from paragraphs 404 and 549 above that the six lines of reasoning of the contested decision did not suffice to demonstrate that the APA had conferred an advantage on SMBV, within the meaning of Article 107(1) TFEU.

560    Second, it follows from paragraphs 550 to 558 above that the Commission has not demonstrated that the APA derogated from Article 8b of the CIT and from the transfer pricing decree and that it had thus conferred an advantage on SMBV, within the meaning of Article 107(1) TFEU.

561    It therefore follows from all the foregoing that the Commission has not managed, by any of the lines of reasoning set out in the contested decision, to demonstrate to a requisite legal standard the existence of an advantage within the meaning of Article 107(1) TFEU. Accordingly, it is necessary to annul the contested decision in its entirety, without it being necessary to examine the other pleas raised by the Kingdom of the Netherlands and by Starbucks.

IV.    Costs

562    Under Article 134(1) of the Rules of Procedure, the unsuccessful party is to be ordered to pay the costs if they have been applied for in the successful party’s pleadings. Since the Commission has been unsuccessful, it must be ordered to bear its own costs and to pay those of the Kingdom of the Netherlands and of Starbucks, in accordance of the forms of order sought by the applicants.

563    In accordance with Article 138(1) of the Rules of Procedure, Ireland shall bear its own costs.

On those grounds,

THE GENERAL COURT (Seventh Chamber, Extended Composition)

hereby:

1.      Joins Cases T‑760/15 and T‑636/16 for the purposes of the judgment;

2.      Annuls Commission Decision (EU) 2017/502 of 21 October 2015 on State aid SA.38374 (2014/C ex 2014/NN) implemented by the Netherlands to Starbucks;

3.      Orders the European Commission to bear its own costs and to pay those incurred by the Kingdom of the Netherlands, Starbucks Corp. and Starbucks Manufacturing Emea BV;

4.      Orders Ireland to bear its own costs.

Van der Woude Tomljenović Bieliūnas

Marcoulli       Kornezov

Delivered in open court in Luxembourg on 24 September 2019.

E. Coulon M. van der Woude

Registrar President

Table of contents

I. Background to the dispute and legal framework

A. National legislative framework

B. Advance pricing arrangement

C. Background to the dispute

1. Administrative procedure before the Commission

2. Contested decision

(a) Description of the contested measure

(b) Assessment of the contested measure

(c) Recovery of the State aid

(d) Conclusion

II. Procedure and forms of order sought

A. Written part of the procedure in Case T760/15

1. Composition of the formation of the Court and priority treatment

2. Interventions

3. Applications for confidential treatment

4. Forms of order sought

B. Written part of the procedure in Case T636/16

1. Composition of the formation of the Court and priority treatment

2. Applications for confidential treatment

3. Forms of order sought

C. Joinder for the purposes of the oral part of the procedure, and the oral part of the procedure

III. Law

A. Procedural matters

1. Joinder of the present cases for the purposes of the decision closing the proceedings

2. Request that Starbucks’ observations on the report for the hearing be withdrawn from the case file

3. Admissibility of Annex A.7 to the application in Case T760/15

B. Pleas raised and structure of the examination of the present actions

C. Existence of an arm’s length principle in the field of State aid control and compliance with the principle of Member States’ fiscal autonomy

D. Dispute as to the principal reasoning regarding the existence of a tax advantage in favour of SMBV (recitals 275 to 361 of the contested decision)

1. Choice of the TNMM in the case at hand and the lack of examination of the intra-group transaction for which the APA had in reality been requested (first line of reasoning)

(a) Preliminary observations

(b) Burden of proof

(c) Intensity of review to be conducted by the Court

(d) Failure to identify and analyse the royalty paid by SMBV to Alki in the APA

(e) Necessity of prioritising the CUP method over the TNMM

2. Whether the royalty paid by SMBV to Alki should have been zero (second line of reasoning)

(a) Preliminary observations

(b) SMBV’s royalty-related functions

(c) Normal taxation rules under Netherlands law

(d) Use of evidence by the Commission that was not available when the APA was concluded

(e) Whether the roasting IP represented a value for SMBV

(1) Whether SMBV exploited the roasting IP directly on the market

(2) Whether SMBV has been loss-making on its roasting activities

(f) Comparison with coffee roasting agreements concluded by Starbucks with third parties and against similar licensing arrangements ‘on the market’

(1) Contracts concluded subsequent to the APA

(2) Contracts concluded with undertakings which do not roast coffee

(3) Contracts with undertakings that did not engage in the sale of roasted coffee to stores or to consumers

(4) Contracts concerning products other than roasted coffee

(5) Contract which provides for the payment of a royalty for the use of the roasting IP

(6) Contract concluded with unaffiliated manufacturing company 2

(g) Arrangements between Starbucks’ competitors and third-party roasters

(h) Argument that the level of the royalty should have been lower than the level endorsed by the APA

3. Annual determination of the costs of green coffee beans (third line of reasoning)

(a) Whether the price of green coffee beans was outside the scope of the contested measure

(b) Whether the level of the mark-up applied to the costs of green coffee beans sold by SCTC to SMBV was not in conformity with an arm’s length level

E. Dispute of the subsidiary reasoning regarding the existence of a tax advantage in favour of SMBV (recitals 362 to 408 of the contested decision)

1. Identification of SMBV as the most complex entity (fourth line of reasoning)

2. Analysis of SMBV’s functions and determination of SMBV’s profit on the basis of operating costs (fifth line of reasoning)

(a) Choice of profit level indicator

(b) Commission’s comparability analysis

3. Choice of adjustments (sixth line of reasoning)

(a) Preliminary observations

(b) Cost-base adjustment

(c) ‘Working capital adjustment’

(1) Scope of the adjustment at issue

(2) Complaint alleging the absence of a reduction in SMBV’s tax burden

F. Whether the APA derogated from Article 8b of the CIT and from the transfer pricing decree (reasoning in respect of the limited reference system, recitals 409 to 412 of the contested decision).

G. Conclusion

IV. Costs

*      Languages of the case: Dutch and English.

1      Confidential information omitted.