This is the third Op-Ed of a Symposium on the Apple State Aid case (C‑465/20 P). Previous Op-Eds have been authored by Romero J. S. Tavares and Scott Wilkie. More Op-Eds will follow soon in EU Law Live.
The decision of the Court of Justice of the European Union (‘the Court of Justice’ or the ‘Court’) in the Apple State aid case caught international tax experts by surprise. While the Opinion of Advocate General (‘AG’) Pitruzzella in November 2023 seemed to be ‘somewhat unexpected’ after ‘an 8-year laborious effort of the Commission…to employ State aid rules against the advance transfer pricing agreements’ hardly anyone still envisaged the Court’s decision to favor the Commission’s ruling that illegal state aid had been granted by Ireland to Apple.
A question, however, is whether these sentiments of surprise were indeed triggered by the obviousness of the ‘right’ outcome in the Apple case or, rather, by the saga of the Commission’s courtroom losses in similar State aid investigations. What went wrong for Apple in this case, and why does the grand finale appear to be opposite of the outcomes in other state aid investigations like Fiat, Starbucks, or Amazon?
This contribution aims to evaluate the transfer pricing aspects of the case and determine whether the Court’s decision aligns with what could be expected under arm’s length conditions. The author does not intend to draw conclusions regarding the existence of state aid, as such an assessment necessitates a comprehensive analysis of various elements, including selectivity, economic advantage, and the significance of any alleged erroneous application of Irish law by the Irish tax authorities that could amount to state aid. In addition, the author is not an expert in Irish tax law and hence this contribution relies primarily on the OECD AOA and OECD Guidelines. It is also important to highlight that while transfer pricing regulations may dictate certain outcomes of the functional analysis, some factual aspects might be viewed differently from a competition law perspective (see, for example, para. 285 concerning the own decision of Apple Group to allocate the IP to Ireland).
Facts
There are several key elements in the Apple case. Apple Operations Europe (‘AOE’) and Apple Sales International (‘ASI’), subsidiaries of Apple Inc., incorporated in Ireland, but are not tax residents in Ireland. From 1991 to 2014, they participated in a cost-sharing agreement (‘CSA’) with Apple Inc., sharing R&D costs and risks for Apple’s products. Under this CSA, ASI and AOE, retaining intangible assets like intellectual property (‘IP’) rights, were licensed to sell Apple products outside North and South America. These entities operated in Ireland through branches considered Irish permanent establishments (‘Irish branches’). The core issue in the case was determining the profits attributable to these Irish branches, which ultimately depended on one question: whether the IP licenses of AOE and ASI under the CSA were to be attributed to the Irish branches.
Following the Advanced Pricing Agreements (‘APAs’) with the Irish tax authority, the licenses were not attributed to the Irish branches and consequently also the profits attributed to these licenses were not subject to corporate taxation in Ireland.
Irish Domestic Tax Law and the relevance of the OECD AOA
The most crucial difference between the Apple case and the other State aid cases is the reference framework. So far, there has not been much opportunity to assess what the Court had to say about the substance and transfer pricing technicalities in the Commission’s State aid investigations. Most cases were resolved and fell apart for the Commission on EU law grounds. In the Amazon case, for example, the CJEU concluded that the reference framework was wrongly identified: ‘the arm’s length principle cannot be applied for the purposes of examining tax measures in the context of Article 107(1) TFEU unless it is recognized by the national law concerned and in accordance with the rules defined by the latter.’ The Court further noted that the OECD Guidelines could only be relevant if they were incorporated into domestic law. In the cases of Fiat and Engie, which also concerned Luxembourg domestic legislation, the Court similarly did not endorse the idea that the arm’s length principle (‘ALP’) is inherent in Article 107 TFEU and may stem from outside the domestic reference framework.
In contrast, in Apple, the General Court (‘GC’) first found and then the Court of Justice subsequently endorsed that Section 25 of the Irish Taxes Consolidation Act 1997 (‘TCA’) was comparable to the OECD Guidance on the functional approach for the attribution of profits to a permanent establishment (the so-called ‘OECD AOA’). After all, both Section 25 TCA and the OECD AOA required a functional attribution of income to the Irish branches as PEs of ASI and AOE.
The Court’s reference to the AOA has been seen by some commentators as a contradiction to earlier judgments, in which the Court repeatedly concluded that ‘the analysis in the state aid cases should be restricted to the domestic law system of the state concerned’ (Collier, R. (2024, September 19), A bad Apple ruling, TaxJournal).Yet, it seems that in the present case, the reliance on the OECD AOA, or at the very least on the functional (functions, assets, risks) approach to the allocation of profits to a PE, was not challenged in a cross-appeal by Apple and Ireland after the General Court’s conclusion that Section 25 TCA and the OECD AOA were effectively similar. Hence, in the absence of a cross-appeal, the judgment of the General Court had the force of res judicata, with no further assessment by the Court of Justice (para. 279, C-465/20 P).
Act 1. Who is Allowed to the Garden to Collect the Apples?
The Commission on one side and Ireland and Apple on the other had differing views regarding the application of the identified reference framework for the attribution of income to a permanent establishment (‘PE’). The seed of discord lay in the question of (i) whether only the functions of the Irish branches mattered for income allocation, (ii) whether the functions of the Irish branches should be considered in the context of the functions of their head offices, AOE and ASI, or (iii) whether, as argued by Apple and Ireland, the functions should be viewed through the prism of the entire Apple Group, with Apple Inc. as the key decision-making and IP-holding company.
The Court of Justice concluded that the GC applied the Irish domestic law test for the attribution of income to the PE, according to which the functions performed by the Irish branches of ASI and AOE should have been assessed relative to the functions performed by their head offices in relation to the Apple Group’s IP licenses held by those companies (paras. 198-200, 256). Yet, the GC, according to the Court, did not follow its own identification of the reference framework by failing to consider the functions of Apple Inc. The Irish law, according to the decision, ‘precludes the role of separate entities, such as a parent company of the non-resident company, from being taken into consideration’ (para. 256, C-465/20 P).
The Commission raised several arguments to justify this limited scope, namely the so-called separate entity approach and the ALP, which both focus on specific transactions. However, the Court did not provide much clarification on whether these arguments were decisive or relevant to the ultimate identification of the reference framework. In any case, this conclusion of the Court endorsed a comparative view on the attribution of profits to a PE but limited the scope of comparison only to the functional profiles of the PE and its respective head office, leaving the rest of the functions performed in the group value chain irrelevant for the assessment.
Act 2. Who Takes Care of the Apple Garden?
This leads us to several points of discussion. First, if the OECD AOA should have been applicable, the question is whether it is indeed the case that the ALP and OECD AOA require placing the focus on one specific transaction, leaving out the functions performed by any other entity in the group. Second, whether the functional comparison between the head office and a branch is prescribed by the ALP and the OECD AOA.
The answer to the first question may be both ‘yes’ and ‘no’, or, as is often the favourite answer of lawyers, ‘it depends’. From a strict legal point of view, the ALP generally applies to individual transactions. For example, Article 9 of the OECD Model Tax Convention (‘MTC’) applies to controlled transactions between two associated enterprises of two contracting states, while Article 7 OECD MTC applies to intra-group dealings between the head office and its PE. Thus, the scope of assessment under the ALP is limited to a specific controlled transaction. It primarily analyses the functions performed, assets used, and risks assumed by both parties in that controlled transaction to determine the price. The OECD AOA prescribes a comparative approach for the attribution of profits to a PE, but similarly specifies:
‘Under the first step, the functional and factual analysis must identify the economically significant activities and responsibilities undertaken by the PE. This analysis should, to the extent relevant, consider the PE’s activities and responsibilities in the context of the activities and responsibilities undertaken by the enterprise as a whole, particularly those parts of the enterprise that engage in dealings with the PE’ (para. 10 OECD AOA).
In this context, ‘the enterprise as a whole’ appears to refer to the entity of which the PE is a part and not the multinational enterprise as a whole. Otherwise, requiring tax administrations to assess the functional profile of the taxpayer in relation to every other entity in a multinational group could impose a disproportionately high and unnecessary administrative burden.
At the same time, the allocation of an asset to a PE under the OECD AOA necessitates identifying ‘significant people functions,’ as pointed out by the Commission. However, not all functions are suitable for attribution of economic ownership.
For the allocation of economic ownership of intangible assets, the OECD AOA requires that a PE performs specific functions, which consist, in particular, of active decision-making (the Irish law appears to take a distinct approach by emphasizing the necessity for control over the asset in question. However, it does not provide a comprehensive definition of what constitutes “control,” leading to potential discrepancies with the OECD AOA) in managing risks related to the development of IP (para. 85 OECD AOA) or enhancing the acquired IP (para. 94 OECD AOA). If these specific functions – active decision-making with respect to IP and associated risks – were not performed by the Irish branches, and there was sufficient evidence that these functions were performed elsewhere (as demonstrated by the taxpayer and recognised by both the General Court and the CJEU), there is no AOA-based ground to allocate the economic ownership to the Irish branches.
In this context, the demonstration by the taxpayer that Apple Inc. was the entity performing the decision-making and risk control functions, as considered by the General Court, could have been intended not to extend the scope of the ALP and the separate entity approach but to provide evidence that the functions relevant for attributing economic ownership of the IP did not rest with the Irish branches.
Instead, in the final decision, the Court considered the functions of Apple Inc. to be fully irrelevant for the application of Section 25 TCA. Furthermore, the Court did not appear to assess whether the specific functions identified by the Commission were sufficient for attributing economic ownership of the Apple licenses to the Irish branches. The Court only stated that ‘the Commission has therefore succeeded in showing that, in light of … the activities and functions actually performed by the Irish branches … the profits generated by the exploitation of the Apple Group’s IP licenses should have been allocated to those branches’ (para. 287, C-465/20 P).
As a separate argument in favour of considering the functions of Apple Inc., it must be noted that the entitlement of ASI and AOE to the proceeds from the Apple IP – of which Apple Inc. remained the legal owner – was based on the CSA.
If ASI and AOE needed to perform active decision-making functions with respect to this IP to be entitled to the proceeds under the CSA, as argued by the Commission, and if they did not perform these functions, then we must consider to which entity these functions were outsourced. If these functions were with another entity under the CSA, ASI and AOE would never have had ownership of the IP along with their Irish branches in the first place. Following the same separate entity approach, ASI and AOE would not be entitled to returns on the IP based on such functional approach.
Had the decision-making functions of ASI and AOE under the CSA indeed been outsourced to the Irish branches, the Commission would have to assume that the Irish branches became parties to the CSA by taking over these functions. The Commission took a similar position in the Amazon case (para. 258, GC, Case T-816/17 and T-318/18), for example. Consequently, if the Irish branches were assumed to be participants in the CSA, the attribution of profits would have not strictly been between the head office and the PE; the contributions of each participant in the CSA must have been analysed.
In conclusion, the Court did not seem to have analysed whether the functions related to the Apple IP performed by the Irish branches were sufficient for attributing economic ownership to them under the OECD AOA. By adhering to the separate entity approach, the Court overlooked the fact that these functions could have been (i) outsourced to another entity in the group, meaning that ASI and AOE, along with their Irish branches, were never entitled to this IP under the CSA; or (ii) if performed by the Irish branches, they would effectively become parties to the CSA, necessitating a comparison of their functions with those of the other CSA participants.
Act 3. Who Collects, and Who Owns the Apples?
The second question raised above is whether a functional comparison of the activities of a PE with those of the head office or any other entity in the group was necessary under the OECD AOA and OECD Guidelines. Is the allocation of the IP licenses to ASI and AOE subject to the same standards as the allocation of profits to a PE under the OECD AOA?
ASI and AOE were bound by the CSA, to which Apple Inc. was also a party. Apple and many commentators have argued that if the value of the IP could not be attributed to ASI and AOE, then Apple Inc. should have been entitled to all the proceeds, as it made all critical decisions regarding the Apple IP in its capacity as the parent entity and as a party to the CSA. This would be accurate if decision-making functions were the golden standard for allocating income also among affiliated parties, similar to the OECD AOA.
However, it is important to note that a sophisticated functional analysis for transactions involving intangibles is a relatively recent development. While the OECD AOA has required significant people functions for the attribution of income to a PE, the same standard did not apply to income allocation among affiliated entities back then. In 2010, the OECD Guidelines introduced specific guidance on the allocation of profits in CSAs. For example, paragraph 8.9 of the 2010 version stated that each participant’s contribution in a CSA could be in cash or in kind. Even back in 2010, and not to mention the 1990s, decision-making functions and active involvement in developing the IP were not prerequisites for entitlement to returns from the IP in a CSA (R. Ökten, A Comparative Study of Cost Contribution Arrangements: Is Active Involvement Required To Share in the Benefits of Jointly Developed Intangible Property?, 20 Intl. Transfer Pricing J. 1).
Moreover, paragraph 8.12 of the 2010 OECD Guidelines further clarified that in cases where all or part of the CSA activity, such as contract research and/or manufacturing, was outsourced to another company that was not a participant in the CSA, that separate company could be compensated with an arm’s length cost-plus fee for the services it provided to the CSA participants (R. Ökten, A Comparative Study of Cost Contribution Arrangements: Is Active Involvement Required To Share in the Benefits of Jointly Developed Intangible Property?, 20 Intl. Transfer Pricing J. 1, Sec. 3.2.2.3). This raises an important question: why could not the domestic Irish provision – which stated that branch income is computed based on what an independent party could have expected in a stand-alone context – rely on this OECD guidance, which allowed for cost-plus remuneration under the same factual circumstances, instead of the OECD AOA Guidance (… both not available at the time of the APAs)?
Next, the OECD AOA does not merely allocate profits; it also requires an understanding of the nature of intra-group dealings. It necessitates hypothesizing and accurately delineating the transaction. In the absence of decision-making functions regarding the development and enhancement of the IP at the level of the Irish branches, which could effectively make them parties to the CSA (see above), potentially a license transaction for the use and exploitation of the Apple IP should have been assumed between ASI, AOE, and the Irish branches. In this scenario, ASI and AOE would fund IP developments by licensing the right to use the IP to the Irish branches in exchange for a licensing fee. In a comparable transaction between two affiliated entities, as seen in the Amazon case for example, the General Court concluded that a return above the operational costs to the operating companies (‘OpCo’) would not correspond to a market outcome (para. 284, T-816/17 and T-318/18). It seems however that for Apple, a reliable approximation of a “market outcome” was determined differently…
In conclusion, at the time of the relevant transaction, there was no requirement for active involvement to allocate the economic entitlement to proceeds from the IP under the CSA. ASI and AOE would be entitled to these proceeds merely by contributing cash, even without performing any decision-making functions. The Commission could potentially challenge this allocation by arguing that the funding under the CSA came from the operations of the Irish branches, similar to the Commission’s position in the Amazon case. Yet, in that case, the GC concluded that it is ultimately the entity’s decision to license out the right to use intangibles in exchange for royalty payments, and the fact that these payments are not subject to tax due to a hybrid mismatch is insufficient to demonstrate a selective advantage (paras. 280-281, GC, T-816/17 and T-318/18).
Finale
The decision in the Apple case appears to be heavily influenced by the current emphasis on a functional approach for allocating income between affiliated parties and attributing income to a PE. It is evident that Irish domestic law lacked a clear standard for addressing all relevant elements of the case. Moreover, neither the OECD Guidelines nor the OECD AOA were consistent in prescribing the significance of people functions when assessing intra-group transactions compared to intra-group dealings.
From a transfer pricing perspective, the main shortcomings of the current decision include: (i) the failure to identify the nature of the transaction in question – specifically, whether the Irish branches were merely licensing the right to use the IP from the head office or if they possessed sufficient functionality to be considered participants in the CSA; (ii) the lack of analysis regarding whether the functions performed by the Irish branches were indeed those necessary to trigger the reallocation of economic ownership under the OECD AOA; and (iii) an undue focus on people functions for the purposes of the CSA, which was not the prevailing practice even in 2010, well after the contested APAs.
Any shortcomings, digressions and leaps of thought inevitably present in this Op-Ed remain the responsibility of the author. Note that the viewpoint in this contribution is not necessarily the viewpoint of her employer. The author would like to thank Natalie Reypens, Partner at an international lawfirm, for her valuable comments on this contribution.
