The Meaning Of State Aid Is Anyone’s Guess After Apple

In European Commission v. Ireland, C-465/20 P, the EU's highest court has confirmed that member states enjoy broad autonomy to design their transfer pricing regimes however they see fit — except when they don't.

The Court of Justice's judgment in Commission v. Ireland reinstates a 2016 decision by the commission, ruling that Ireland granted illegal state aid to Apple Inc.'s subsidiaries through an advance pricing agreement granted in 1991 and renewed in 2007. The commission's decision, which drew widespread scorn for its technical shortcomings and allegedly political motivation, was annulled by the EU General Court's 2020 decision in Ireland v. Commission and Apple Sales International and Apple Operations Europe v. Commission, joined cases T-778/16 and T-892/16.

In ruling for the commission, the Court of Justice largely followed the surprising opinion by Advocate General Giovanni Pitruzzella urging the Court to annul the General Court judgment. The key difference was that the Court issued a final judgment in the commission's favor instead of remanding the case as Pitruzzella recommended.

The reason Pitruzzella's opinion and the Court's Commission v. Ireland judgment were so surprising is that they seemingly defy case law that isn't even two years old. In Fiat Chrysler Finance Europe v. Commission and Ireland v. Commission, joined cases C-885/19 P and C-898/19 P (2002), the Court pointedly rejected the imposition of parameters external to member state law when assessing the selectivity of APAs under 107(1) of the Treaty on the Functioning of the European Union. The Court affirmed this principle as recently as December 2023 in Commission v. Luxembourg, C-457/21 P, which rejected a similar state aid ruling concerning Amazon.

Unless justified by the logic of the member state's tax system, national tax measures that deviate from the reference system by selectively favoring some enterprises over others violate 107(1). The commission's 2016 decision regarding Apple, like its decisions on alleged aid granted to other high-profile multinationals, found that the unjustified selective deviation was Ireland's failure to enforce the arm's-length principle in the APAs. The commission based its assessment almost entirely on OECD guidance on transfer pricing and the attribution of profit to a permanent establishment, or at least the commission's controversial interpretation thereof.

A lightning rod for criticism in the Commission v. Ireland judgment is the Court's decision to accept as given that Irish law incorporated the arm's-length principle and something substantially similar to the authorized OECD approach (AOA) to profit attribution. It's contentious because Ireland had no domestic transfer pricing regime formally adopting the arm's-length principle when the APAs were granted, and it would have been impossible to have branch profit attribution rules based on an AOA that didn't yet exist.

Although the General Court ultimately ruled against Ireland and Apple in 2020, it rejected the objections to the commission's invocation of the AOA and OECD transfer pricing guidelines. Opportunistically invoking the principle of res judicata, the Court of Justice declined to revisit the General Court's purported holding because it was never challenged in a cross-appeal.

The Court's refusal to scrutinize the General Court's holding on narrow procedural grounds was especially striking considering the Fiat judgment, which forcefully rejected a similar attempt by the commission to equate domestic law with OECD transfer pricing guidance.

"Parameters and rules external to the national tax system at issue," paragraph 96 of the 2022 Court of Justice judgment proclaimed in Fiat, are relevant only when the "national tax system makes explicit reference to them." Although the objective of the relevant national measure in Fiat "correspond[ed], in general terms, to that of the arm's length principle, the fact remains that, in the absence of harmonisation in EU law, the specific detailed rules for the application of that principle are defined by national law."

Because the version of section 25 of Ireland's Taxes Consolidation Act 1997 in effect when the APAs were granted didn't (and couldn't) make explicit reference to later OECD guidance, most observers assumed that the commission had almost no chance of winning after Fiat. But most observers were wrong: The Court of Justice essentially flouted its recent holding in Fiat without resolving, or even acknowledging, the apparent contradiction. Unless and until future judgments provide more clarity, EU tax administrations and taxpayers can only guess exactly how OECD transfer pricing guidance will factor into selectivity assessments.

The Court's decision to let the commission foist the AOA and arm's-length principle onto Irish law will undoubtedly prompt a flood of critical commentary, and understandably so. However, this focuses on whether the APAs granted by Ireland were inconsistent with OECD profit attribution and transfer pricing guidance. The accepted meaning of the arm's-length principle is always evolving, and there has historically been a range of plausible interpretations to choose from at any given time. If Fiat retains any meaning at all, then it seems that a member state should enjoy the discretion to apply any one of these plausible interpretations.

Examining the issue isn't simply a matter of piling more criticism on a deeply flawed decision. It may also shed light on the extent of the likely fallout from Commission v. Ireland for other APA-related state aid cases.

Filling the Void

It's important to recall that, in the famous words of former U.S. Treasury financial economist Michael McDonald, " 9 ain't 7." The arm's-length principle is the basis for allocating income among commonly owned or controlled enterprises under 9 of the OECD model tax convention, and the OECD transfer pricing guidelines are widely accepted as the canonical interpretive authority on the principle's meaning.

Under 9, a jurisdiction may tax resident enterprises on the income they would have earned had they transacted at arm's-length prices with any enterprises under common ownership or control. The OECD guidelines provide that the arm's-length price must be commensurate with the functions, assets, and risks of the transacting parties, and it should account for economic conditions and any other circumstances that may materially affect prices.

Article 9's arm's-length principle is also relevant to 7, which deals with the attribution of profit to a PE or branch, but its relevance is indirect. In its 2008 and 2010 reports on profit attribution, the OECD established the AOA as the approved method for allocating income to a local PE of a nonresident enterprise.

The AOA hypothesizes the PE as a functionally separate entity and applies the arm's-length principle under 9, as adapted to account for the branch-versus-subsidiary distinction. The objective is to leave the PE with the profit that it would have earned had it been an independent enterprise dealing at arm's length with the rest of the legal entity of which it is part.

The key adaptation necessary to use the arm's-length principle as a profit attribution mechanism under 7 concerns the allocation of assets and risks. The OECD transfer pricing guidelines emphasize that the distinct legal rights and obligations of commonly controlled enterprises, including their ownership of assets and contractual assumptions of risk, should generally be respected by tax administrations. But a single legal entity can't contract with itself or split asset ownership across its different physical locations. A different asset and risk allocation mechanism is needed.

The AOA's solution is to allocate assets and risks based on the relevant "significant people functions," which comprise the key decision-making activities involved in acquiring, developing, or managing assets and in assuming or managing risk. The location of the economic owner of an asset, and the deemed bearer of a risk, is the location where the enterprise carries out these active decision-making functions.

Economic ownership of intangible property and IP development risk should thus be allocated to the part of the enterprise where active management of the development program takes place. These active decision-making functions may be "performed below the strategic level of senior management," according to the OECD report, but they must still entail the organizational authority and practical competence to direct the intangible development process.

However, there's a gap in the AOA, as articulated in the OECD's 2010 profit attribution report: It seemingly assumes that, for every significant asset and risk, the significant people functions must lie somewhere within the enterprise. The OECD report doesn't address situations in which neither the PE nor any other part of the enterprise carries out the critical decision-making activities that typically dictate economic ownership and risk assumption, and it's unclear whether the report's drafters considered the possibility.

But as the facts in Commission v. Ireland clearly attest, it's entirely possible for the key decision-making functions required to establish economic ownership or risk assumption to be made by commonly controlled but legally separate enterprises.

In the variation of the double-Irish structure put in place by Apple, two group entities that were incorporated in Ireland but effectively managed elsewhere — Apple Sales International (ASI) and Apple Operations Europe (AOE) — avoided tax residency in any jurisdiction. ASI and AOE entered a cost-sharing arrangement (CSA) with Apple Inc., under which the two entities agreed to make cost-sharing payments in exchange for rights to exploit the cost-shared intangible property outside the Americas.

Although the two entities' "head offices" were essentially cash boxes that reaped the returns attributable to Apple's cost-shared IP rights, each entity also maintained an Irish branch that carried out substantial but routine activities.

It's beyond any real dispute that the significant people functions necessary to develop cost-shared IP under Apple's CSA took place entirely in Cupertino, California. The head offices of AOE and ASI, as the commission rightly observed, were little more than tax planning vehicles responsible for holding board meetings and attending to legal formalities. The only substantial economic activities carried out by AOE and ASI took place at their Irish branches, which were tasked with important but routine manufacturing, procurement, marketing, and support services. Although the branches' functions were necessary to successfully exploit the cost-shared IP, they had no substantial role in deciding what IP to develop or how it should be developed.

This raises the question left unaddressed by the OECD profit attribution report. Must the assets legally owned and risks contractually borne by an enterprise be allocated somewhere within that enterprise under the AOA, even when none of the enterprise's personnel at any location perform what would otherwise be the necessary significant people functions? If so, then the cost-shared IP rights must be allocated either to each entity's head office or to its Irish branch. Some language used in the report, including in paragraph 18, arguably supports this strict intra-enterprise interpretation:

"The functional and factual analysis will examine all the facts and circumstances to determine the extent to which the assets of the enterprise are used in the functions performed by the PE and the conditions under which the assets are used, including the factors to be taken into account to determine which part of the enterprise is regarded as the economic owner of the assets actually owned by the enterprise."[Emphasis added.]

But elsewhere, including in paragraph 15, the report's language suggests that a PE can only economically own the assets and assume the risks that its own significant people functions support:

"The authorised OECD approach attributes to the PE those risks for which the significant functions relevant to the assumption and/or management (subsequent to the transfer) of risks are performed by people in the PE and also attributes to the PE economic ownership of assets for which the significant functions relevant to the economic ownership of assets are performed by people in the PE." [Emphasis added.]

Although readers with enough creativity and motivation can find support for whichever answer they prefer, attaching IP income to a PE that doesn't perform the necessary significant people functions seems to be the less plausible of the two. But the Court, which offered only a superficial and conclusory explanation for its assessment, evidently saw the matter differently. It accepted the commission argument summarized in paragraph 137 of the judgment:

"It follows from the separate entity approach and the arm's length principle that Apple Inc., on the one hand, and ASI and AOE, on the other, should be treated as separate entities for tax purposes, and that their commercial and financial relations, which are governed by intra-group transactions, should be priced at arm's length. When attributing profits of ASI and AOE between their respective head offices and branches, all that matters are the functions performed by those head offices and branches." [Emphasis added.]

The Court, like the commission, thus appeared to consider the exclusively intra-enterprise interpretation to be an unavoidable corollary of the arm's-length principle and the separate-entity approach on which it is based. Although the Court never bothered to explain its reasoning, the assumption appears to be that the principles of 9 effectively shrink the universe under 7 to the legal entity with the PE.

In other words, respecting the stateless CSA participants' separate legal existence and cost-shared IP exploitation rights (as the arm's-length principle under 9 would generally require) means accepting that the IP rights are stuck somewhere in ASI and AOE. They cannot drift back to Apple Inc., even if the only significant people functions took place in Cupertino.

It would follow that the AOA assessment must find significance in the people functions carried out somewhere within the enterprise, whether the significance is really there or not. Because something is more significant than nothing, the preordained winners of the two-way significance test must be the Irish branches that perform routine but substantial functions over the head offices that do almost nothing.

It's unclear how the Court, faced with a clear void in direct OECD guidance, found its absolute conviction that the strict intra-enterprise interpretation was the only possible way to approach this situation under the AOA. The illogical implications should have given the Court pause.

The most glaring is the implication that income attributable to cost-shared IP must be allocated not to the jurisdiction where it was actually developed or funded, but to PEs that had no role in funding or developing the IP. This is contrary to the fundamental principle that assets and risks cannot be decoupled from functions under the AOA.

The strict intra-enterprise interpretation would also lead to drastically different results in substantially similar cases. If ASI and AOE had each established an Irish subsidiary instead of a branch to carry out the same routine activities, then there would be no plausible argument for allocating any of the cost-shared IP income to Ireland. Although 9 isn't 7, the AOA is still based on the arm's-length principle. It's hard to justify a huge disparity in the treatment of routine activities performed by a branch and routine activities performed by a subsidiary.

Another awkward implication is that ASI and AOE, which the commission understandably insisted had too little substance to support the allocation of cost-shared IP rights under the AOA, nevertheless had the substance necessary to acquire the rights from Apple Inc. in the first place.

The activities performed at each Irish branch, which were trivial when the corresponding legal entity joined the CSA, cannot resolve the apparent discrepancy. The only potential justification is the distinction between 9 and 7, and that justification is probably no longer tenable after the base erosion and profit-shifting project.

The second interpretation, and the one that better reflects economic reality, would recognize that neither the branches nor the head offices performed the functions necessary to establish economic ownership of the cost-shared IP or assume the corresponding development risk. Assets, risks, and the corresponding returns follow significant people functions under the AOA, and the Irish branches simply did not perform those functions. The branches' routine functions may have been more significant than nothing in a general sense, but they were equally insignificant in relation to the specific assets and risks at issue. This interpretation suggests that if the only significant people functions relevant to Apple's CSA took place in Cupertino, then the only jurisdiction that had the right to tax the returns was the United States.

Letting Apple's income go untaxed is regrettable. But whether the United States actually exercised its taxing rights really isn't any of Ireland's business. It can't be illegal state aid for Ireland to refrain from taxing something it never had the right to tax.

Choose Your Arm's-Length Principle

It's true that the second interpretation of the AOA, which the Court somehow overlooked or perceived as too frivolous to warrant acknowledgment, may not strictly adhere to traditional interpretations of the arm's-length principle and the separate-entity approach. However, since the OECD issued its final BEPS in 2015, neither does the interpretation set out in the OECD transfer pricing guidelines.

Although interpretations of the OECD's post-BEPS transfer pricing guidance differ widely, the guidelines have clearly repudiated the idea that simply making cost-sharing payments should entitle cash boxes to astronomical returns. The commission hasn't shied away from retrospectively applying this principle in other state aid cases, making its selective devotion to the separate-entity approach all the more curious.

According to the post-BEPS guidance on risk in Chapter I of the OECD guidelines, tax authorities can reallocate risk contractually assumed by a group entity (along with any upside returns) that cannot or does not exercise control over the risk.

As explained in paragraph 1.65 of the post-BEPS transfer pricing guidelines, an entity must exercise the ability to decide whether to "take on, lay off, or decline a risk-bearing opportunity" and how to "respond to the risks associated with the opportunity" to establish control. Although this control test applies to the contractual assumption of risk in general, the revisions to Chapter VI of the guidelines confirm that it is especially relevant for intangible development risk.

The post-BEPS version of Chapter VI emphasizes that legal ownership, by itself, doesn't entitle the owner to retain the returns generated by the intangible. A group entity's right to share in the returns depends on the functions it performs or controls, the risks it assumes, and the assets it contributes to the development, enhancement, maintenance, protection, and exploitation (DEMPE) of intangible property. Given the interrelated nature of intangible development risk and intangible development functions, the most significant of which typically involve deciding whether to assume development risk or how to manage it, the underlying control standard is essentially the same.

The principles described in the revised Chapter VI for compensating asset contributions (particularly those that take the form of intangible development funding) rely on this concept of control as well. Chapter VI recognizes that funding intangible development, for example by making cost-sharing payments like those made by ASI and AOE, is an asset contribution that requires compensation. However, the extent of that compensation scales with the level of control over intangible development risk and the associated development functions.

Because a group entity has no right to a return on risk that it never properly assumed, funding intangible development without controlling any operational or financial risks generally entitles the funding entity to no more than a risk-free return on capital. But higher returns are available if the funding entity at least controls the financial risk associated with investing in the intangible development project. In that case, the group entity funding the development project may be entitled to a risk-adjusted expected return on capital invested. However, to claim a full share of any upside returns, the funding entity must also control specific operational and development risks.

If these principles were applied to Apple's CSA, it's unlikely that any rights to the cost-shared IP would have ended up in ASI or AOE in the first place. Nothing alleged by the commission or found by the General Court suggests that any part of ASI or AOE was ever capable of controlling the risks, either operational or financial, associated with intangible development under the CSA. They are like the funding entities that, under the post-BEPS version of Chapter VI, should earn no more than a risk-free return on their cost-sharing payments.

ASI and AOE didn't and couldn't control risk for essentially the same reason the commission gave for allocating the IP rights to Ireland: The head offices, which served the sole purpose of allowing ASI and AOE to participate in the CSA, never actually did anything substantial. As stated in paragraphs 118 and 119 of the commission's 2016 decision, the agreement signed by the CSA participants covered technology, copyrights, trade secrets, updates, processes, quality standards, and marketing intangibles along with similar items created under the "Development Programme."

However, according to the commission's 2016 decision and the General Court's 2020 judgment, ASI and AOE board meetings dealt with unrelated housekeeping subjects like paying dividends, appointing directors, managing bank accounts, carrying out audits, and handling other legal formalities. None of the handful of more substantive business decisions that, according to paragraph 306 of the General Court judgment, occasionally came up at board meetings had anything to do with actively formulating, amending, or executing the CSA Development Programme.

The Irish branches didn't control any of the development risk associated with the CSA either. They were responsible for important procurement, marketing, distribution, and specialized manufacturing functions, but none of these important functions involved deciding which IP development opportunities to pursue under the Development Programme or how to pursue them.

Without performing the activities necessary to establish control over DEMPE functions or risks under the CSA, the Irish branches' functions provide no basis for claiming the returns attributable to intangible development. Describing their activities as DEMPE functions doesn't change their character as routine contributions that can be reliably compensated using the transactional net margin method.

Even the less exacting threshold for control over financial risk, which entitles a funding entity to a risk-adjusted return on capital, would require decision-making functions that ASI and AOE didn't and couldn't perform. As explained in paragraph 6.64 of the OECD guidelines, control over financial risk still requires the exercise of real decision-making capabilities:

"The relevant decisions relating to taking on, laying off or declining a risk bearing opportunity and the decisions on whether and how to respond to the risks associated with the opportunity, are the decisions related to the provision of funding and the conditions of the transaction. . . . The higher the development risk and the closer the financial risk is related to the development risk, the more the funder will need to have the capability to assess the progress of the development of the intangible and the consequences of this progress for achieving its expected funding return, and the more closely the funder may link the continued provision of funding to key operational developments that may impact its financial risk. The funder will need to have the capability to make the assessments regarding the continued provision of funding, and will need to actually make such assessments, which will then need to be taken into account by the funder in actually making the relevant decisions on the provision of funding."

Although the General Court found that board members of ASI and AOE signed various versions of the CSA agreement at meetings in Cupertino, paragraph 1.66 of the OECD guidelines specifically states that signing off on decisions does not establish control:

"Neither a mere formalising of the outcome of decisionmaking in the form of, for example, meetings organised for formal approval of decisions that were made in other locations, minutes of a board meeting and signing of the documents relating to the decision, nor the setting of the policy environment relevant for the risk . . . qualifies as the exercise of a decisionmaking function sufficient to demonstrate control over a risk."

If neither the head offices nor the Irish branches of ASI and AOE controlled either the development risk or the financial risk associated with the CSA, then the post-BEPS version of Chapter VIII of the guidelines requires their exclusion from the arrangement. For a group entity to participate in a "development cost-contribution arrangement" (CCA), which is the OECD guidelines' equivalent of a CSA, paragraphs 8.14 through 8.18 require that the entity control the specific risks it assumes under the CCA. But it was Apple Inc., not ASI or AOE, that retained control over the intangible development risk associated with the Development Programme formulated and executed in Cupertino.

This means both entities would be excluded from the CSA under current OECD guidance, and they would be entitled only to a risk-free return on the capital contributed in the form of cost-sharing payments. This would effectively void the CSA's transfer of rights to exploit cost-shared IP and leave the United States with the exclusive right to tax the vast majority of the returns.

The commission's arguments in other state aid cases show that it is familiar with these principles and, at least when they support its position, willing to apply them retrospectively. In Commission v. Luxembourg and in the ongoing Nike state aid case, the CSA participants are analogous to the head offices of ASI and AOE without any Irish branch.

Each is a separately incorporated subsidiary that acquired territorial rights to exploit cost-shared intangible property, which the Amazon and Nike subsidiaries sublicense to local affiliates throughout the region. Evidently undeterred by the violation of the separate-entity principle, the commission has argued that these CSA participants have no right to retain the royalty income they collect from sublicensing the cost-shared IP rights to affiliates.

Is Fiat Still Relevant?

Of course, the BEPS project's updated guidance on control, risk, DEMPE functions, and CCAs didn't exist when the OECD drafted its 2010 profit attribution report on the AOA or when Ireland granted the APAs. But these updates form part of the canonical interpretation of the arm's-length principle today. For the Court to have been correct that the arm's-length principle requires a strictly intra-entity comparison, and thus that the intercompany transfer of cost-shared IP rights to ASI and AOE was beyond scrutiny, then the version of the principle that Ireland was bound to enforce must irreconcilably differ from the version now endorsed by the OECD.

The implication that the arm's-length principle relevant in Commission v. Ireland means something completely different from what the OECD now says it means should have been a red flag. It's not that the OECD's post-BEPS iteration of the arm's-length principle was always the one and only reasonable interpretation, or that the AOA must be construed consistently with this new iteration. The point is that there's a range of different plausible interpretations, and the current front-runner flatly contradicts the interpretation assumed by the Court.

It's hard to seriously contend that the only possible meaning of the arm's-length principle in 1991 and 2007 was something that conflicts with the OECD's clarified interpretation as of 2015. And it would be almost absurd to suggest that the arm's-length principle's meaning was so manifestly unreasonable that it can justifiably be injected into prior Irish law.

Choosing between different plausible interpretations is precisely the kind of judgment call that the Court of Justice so pointedly reserved to EU member states less than two years ago, or at least so everyone thought at the time. According to the Court of Justice's 2022 Fiat judgment, as explained in Advocate General Juliane Kokott's 2023 opinion in Commission v. Luxembourg, Fiat leaves critical details like the criteria for selecting transfer pricing methods to the discretion of member states.

And as cases like Coca-Cola Co. v. Commissioner, 155 T.C. 145 (2020), and Medtronic Inc. v. Commissioner, T.C. Memo. 2022-84, demonstrate, the choice between competing methods can be every bit as consequential as the choice between competing interpretations of the AOA in Commission v. Ireland.

There were certainly methodological flaws in the profit attribution approach approved by the APAs that Ireland granted to Apple, and it's entirely possible that these amounted to selectivity favorable treatment. However, with a void in on-point OECD interpretive guidance, the wide latitude recognized in Fiat and confirmed in Commission v. Luxembourg should have given Ireland the discretion to apply any AOA-like domestic legislation it may have had in a manner consistent with the OECD's later articulation of the arm's-length principle. Exactly why the Court of Justice believed otherwise is — and will likely remain — shrouded in mystery.

Still, there are reasons to be optimistic that Commission v. Ireland won't turn the entire post-Fiat world upside down. For one thing, it's no longer possible under Irish law to escape tax residency using the double Irish stricture. And tax ruling practices are now subject to the minimum standards established by action 5 of the BEPS project, which require spontaneous exchange of unilateral APAs. Both developments significantly reduce the likelihood that similar cases will arise in the future.

Another reason for guarded optimism is the general convergence in EU member states' transfer pricing rules with OECD standards. Aligning domestic law with post-BEPS OECD transfer pricing guidance, which most EU member states have effectively done one way or another, incorporates the revised OECD guidance on risk allocation and DEMPE functions. If an EU member state has duly incorporated this guidance into domestic law, and the multinational group concerned doesn't exercise control over intangible development in its jurisdiction, it's hard to see what basis the commission could have for claiming that the country's tax administration should have taxed the returns.

This means the separate-entity approach may soon lose its viability as a justification for trapping IP rights in a legal entity that performs no significant people functions. Future agreement on a proposed directive standardizing transfer pricing and profit attribution rules could further limit the commission's ability to raise exotic arguments and courts' opportunity to accept them.

Another factor that could contain the fallout from Commission v. Ireland is the decisive importance of Apple's ill-fated reliance on a branch structure. The identity of the legal entities concerned, as opposed to any real functional overlap or integration between the Irish branches and head offices, was the only link between the branches and the cost-shared IP rights.

Had Apple instead incorporated the Irish branches of AOE and ASI as legally separate subsidiaries, there would have been no plausible basis under any version of the OECD guidelines to attribute CSA-related returns to them. And there is nothing in the Commission v. Ireland judgment that suggests otherwise.

Although it may be of little comfort to Apple, there's also the possibility that the Court was more concerned with securing a specific outcome in this particular case than with establishing general precedent. The Court's rigid refusal to consider important arguments on res judicata grounds has led to grumbling about possible political motivations. If there's any truth in these suggestions, then the result may be a one-off.

However, it may be unfair to infer bad faith from a poorly reasoned, wrongly decided, tax-related judgment by the Court of Justice. After all, this is the same court that decided Hornbach-Baumarkt AG v. Germany, C-382/16 (2018), which endorsed the ludicrous premise that propping up a financially distressed subsidiary may be a valid excuse for mispricing transactions between it and the parent. It's also the same court that decided Cadbury Schweppes plc and CSO Ltd. v. Commissioners of Inland Revenue, C-196/04 (2006), which has created havoc for EU countries' anti-base-erosion regimes ever since.

Whatever the Court's motivations, erratic and misguided EU case law applying the TFEU to national tax measures has been a fact of life for member states for decades. Hopefully, member state convergence in transfer pricing and the factual peculiarities of the case will make Commission v. Ireland easier to live with, at least until next time.