On 22 September 2023, the European Commission adopted a decision imposing a fine of €376.36 million on microprocessor manufacturer Intel. This decision relates to the facts that had already been sanctioned in an earlier decision of 13 May 2009. How can this apparent inconsistency be explained? In fact, the first decision was annulled in January 2022 by the General Court after a referral by the Court of Justice. This referral was based on the judgment of 6 September 2017, which constitutes one of the decisive turning points in the application of the effects-based approach to certain price-based practices of dominant firms that can lead to the foreclosure of competitors from the market.
So why a new decision? In fact, the 2009 decision dealt with two different types of practices that were considered to be anti-competitive at the time. On the one hand, the decision analysed conditional loyalty rebates, which the Commission considered to constitute an abuse of a dominant position per se, and on the other, practices involving naked restrictions, in other words, manoeuvres aimed at hindering the placing on the market and obstructing the distribution of PCs equipped with competing CPUs.
The judgment of the General Court finally delivered in 2022 annulled the part of the Commission’s decision relating to the qualification as abusive of the loyalty rebates introduced by the dominant operator. However, the annulment concerned only that part of the decision. The part relating to the naked restrictions was not affected. The fine imposed in 2009 covered both practices. It was not possible to attach a specific quantum to each of them. Therefore, as it was not possible to distinguish in the fine what related to each of the two practices, the entire financial penalty was annulled, leaving the Commission to take a new decision relating only to these practices. It is this decision, which was published in September 2023, that we should revisit.
This post considers three successive points. Firstly, it briefly reviews the background of the case, emphasising that the annulment related to only one of the two parts of the decision and not to all of the practices that the Commission had described as anti-competitive in May 2009. Secondly, it will return to the setting of the quantum of the financial fine, specific to naked restrictions in the decision handed down in September 2023. Thirdly, it will discuss the specific features of this market practice with regard to the application of competition rules.
Laying one more milestone in a long legal battle
The Commission’s decision of 22 September 2023 is the latest development in a case that gave rise to an initial decision on 13 May 2009. It is by no means the end of the proceedings, as the company may refer the matter to the General Court for judicial review, and the General Court’s referral judgment of 26 January 2022 was itself the subject of an appeal to the Court of Justice, in which Advocate General Medina delivered her opinion on 18 January 2024.
Bearing in mind that the proceedings were opened on 26 July 2007 in respect of practices some of which began in October 2002, it is necessary to briefly review the background to the case in order to explain the basis of the present decision.
The Commission’s decision of 13 May 2009 sanctioned anti-competitive practices implemented by Intel to the detriment of its competitor on the x86 CPU market, AMD, between October 2002 and December 2007. The Commission considered that all the practices implemented by Intel, whether loyalty rebates or naked restrictions (i.e., direct payments to equipment manufacturers and distributors to hinder access to the market for chips equipped with competing processors), constituted an abuse of dominant position and fined Intel €1.06 billion for these two practices taken together.
What were these direct payments?
With regard to distributors, Intel had granted rebates to Saturn (MSH – Media Saturn Holding) conditional on the exclusive sale in its shops of PCs equipped with Intel x86 chips. As for the payments to OEMs, they were conditional on a number of requirements ranging from the postponement of market launches of PCs equipped with AMD chips to the specialisation of deliveries according to the type of demand or the cancellation of orders for competing processors for certain types of PC. The Commission established in its May 2009 decision that Intel had conditioned the payments granted to HP on the prioritisation of PCs equipped with AMD processors to less strategic customers or on the acceptance of a six-month delay in the marketing of these same PCs in certain geographical areas. Acer and Lenovo were offered the same conditions linked to delayed market launches, with an additional requirement for Lenovo to refrain from launching notebooks equipped with AMD processors.
In its judgment of 12 June 2014 (Case T-286/09), the General Court upheld the Commission’s decision. However, the judgment of the Court of Justice of 6 September 2017 ruled in favour of Intel’s requests and referred the case back to the General Court. In the Court’s view, the General Court had committed procedural irregularities which affected Intel’s ability to exercise its rights of defence. The CPU manufacturer considered that the General Court had not addressed its claim that its rebate mechanisms were unable to exclude a competitor as efficient as Intel from the market.
On 26 January 2022, the General Court delivered a new judgment (the referral judgment – Case T-286/09 RENV) in which it annulled the part of the May 2009 decision relating to the rebates but not to the direct payments intended to impede access to the market for computers equipped with competing chips (para 4 of the decision).
The General Court’s judgment, therefore, rejected Intel’s claim that the naked restrictions should be subject to the same test as that applicable to conditional rebates. The question was, therefore, as follows: do naked restrictions need to be proven to be capable of having, at least, potential foreclosure effects in order to be sanctioned under Article 102 TFEU? (para 87of the referral judgment)
Intel’s position was clear: it was requesting the application of the same test to the two pricing practices that could result in the foreclosure of a competitor, i.e., conditional rebates and conditional payments. Following this argument would have meant applying the as-efficient competitor test to the latter. In other words, the direct payments could only have been penalised if it had been proved that they could not have been replicated by Intel’s competitors with the same cost function. However, in the General Court’s view in its initial judgment, the Court did not require the same test to be applied to naked restrictions as to loyalty rebates (para 93 of the referral judgment).
If the naked restrictions were not affected by the annulment of the decision by the General Court, why was the entire fine annulled? As explained in the introduction, in its decision of May 2009, the Commission had not identified and separated the part of the fine relating to these restrictions. Insofar as the General Court was not in a position to recalculate the amount of the appropriate penalty for these practices alone, it was obliged to annul the decision in its entirety (para 529 of the referral judgment). The General Court could, however, have exercised its unlimited jurisdiction to determine the appropriate amount unilaterally, but preferred to remit the matter to the Commission (para 9 of the Commission’s reinstated decision). As a result, the Commission was called upon to take a new decision to impose a penalty that related only to naked restrictions and was no longer applicable to the infringement relating to conditional rebates (para 530 of the referral judgment).
Insofar as the appeals lodged against the referral judgment do not relate to the naked restrictions but only to the fidelity rebates, their characterisation is now of res judicata. The Commission must, therefore, reopen the proceedings at the stage of calculating the penalty for this particular practice.
Calculating a revised fine quantum
While the anti-competitive characterisation of the practice is based on its object in this specific case, the penalty is calculated, at least in part, on the basis of its effects on the market. The rules for determining financial penalties are based in particular on the setting of a basic amount, determined on the sales affected by the practice at stake, its duration, and on the application of a coefficient based on an assessment of its gravity.
The Commission’s approach to establishing the quantum of the fine is based on its 2006 Guidelines, as applied in the initial 2009 decision. In this context, the basic amount is determined by applying a variable coefficient of up to 30% of the company’s sales on the market affected by the practice. The result is then multiplied by the number of years during which the practices were carried out.
In this case, the practice concerned x86 processors produced between 2002 and 2008 and marketed in the European Economic Area (EEA). The relevant market in this case is that for processors for desktops and chromebooks, not servers. The turnover for 2006, the last year in which the naked restrictions were implemented was used to set the fine base amount. This one calculated by the Commission was €2.977 billion. What percentage should be applied to this base? As the Commission points out, this depends on a number of parameters relating to the nature of the practice implemented, the market power of the company concerned and the geographical scope covered by the practice.
The first point relates to the nature of the practice in question. The Commission considers that the naked restrictions constitute practices that deviate from the principles of competition on the merits and have an anti-competitive object (para 44 of the Commission’s decision). The second point relates to the market power of the undertaking concerned. At the material time, Intel controlled 80% of the market, faced only the competitive pressure of one competitor and the relevant market was characterised by significant barriers to entry and expansion (para 46). Lastly, in terms of geographical scope, the Commission considers that the entire EEA was affected by the practice.
While the Commission had chosen a coefficient of 5% in 2009 for the entire set of practices, in September 2023 it opted for a coefficient of 4%. In the Commission’s view, this lower severity coefficient can be explained by the fact that the practices lasted less time than the loyalty rebates and were much smaller in scope (para 50). It should be noted, as the Commission itself does in its decision, that this coefficient is not particularly high in relation to its recent decision-making practice. In both the 2018 Google Android decision (AT.40099) and the 2019 Google AdSense decision (AT.40411), the rate used was 11%. The difference is largely explained by the fact that the elements of duration and market coverage mentioned above were completely different.
In accordance with its 2006 Guidelines, the Commission considers the duration of the practices in order to apply the multiplier to the basic amount (i.e., to the base of relevant sales adjusted by the gravity coefficient). In view of the duration of the infringement, from November 2002 to December 2006, a coefficient of 3.16 is adopted (para 60). This gives an amount of 376.358 million euros.
According to the fine-setting methodology, one needs to consider any mitigating or aggravating circumstances and check that the fine imposed is actually below the legal ceiling of 10% of turnover.
Only one of these steps is worth discussing. It relates to the arguments put forward by the defendant company regarding the need to account for the excessive length of the proceedings and the costs incurred by the defence. Although the Commission rejected these requests, the fact remains that these elements are interesting to consider. The first point of interest relates to the duration of the case. Over and above the effects of the various appeals, there is the question of the duration of the administrative phase. As the Commission points out, in the specific case of LVM (Joined Cases C-238/99 P, C-244/99 P, C-245/99 P, C-247/99 P, C-250/99 P to C-252/99 P and C-254/99 P) the Court of Justice considered that a duration of eleven years and four months was not excessive in itself. A second point of interest relates to the costs associated with the defence. The Commission did not accept the firm’s arguments that the costs incurred in organising its defence should be recognised. Notwithstanding this assessment, a relevant question remains that of the costs of competitive proceedings for all of the stakeholders. The issue of legal certainty (in terms of the legibility of the rule, the predictability of its application and the robustness of the resulting decisions on appeal), the duration of cases and the costs incurred must be considered when assessing the collective effectiveness of the application of competition rules.
Discussing the differences in treatment between the two types of abusive practices
Beyond the calculation of the appropriate fine to sanction the naked restrictions, several economic and legal questions can be raised. Ultimately, the legitimacy of the method of competition is at stake. Competition on the merits and abusive competition can have the same result: the strengthening of a dominant position. The resulting foreclosure is not in itself anti-competitive insofar as it can be considered that consumers can only be harmed by the exit from the market of a competitor who is less efficient than the dominant operator. While this assumption may be questionable insofar as the competitive threat posed by an operator present on the market is more significant than that posed by potential entrants, it has underpinned many of the Opinions of Advocates General in recent years (see AG Rantos in Servizio Elettrico Nazionale, Case C-377/20, for instance). However, this distinction does not exhaust the whole question of competition.
Should the legality of a given market practice be judged solely on the basis of its effects on efficiency, or should it also be judged on the basis of its legitimacy? In other words, can practices be sanctioned per se under the competition rules on the basis of their anti-competitive purpose?
In the Commission’s view, the anti-competitive object can serve as a basis for characterising a practice as violating Article 102 TFEU. This is the position it defends in para 10 of its decision: “It should be recalled that the naked restrictions were confirmed already in the initial judgement as pursuing an anti-competitive object and clearly falling outside the scope of a competition on the merits, capable by their very nature of restricting competition […]“. It had already confirmed this position in its initial decision of May 2009, laying down the principle that an infringement of Article 102 TFEU may also result from the anti-competitive object of the practice (para 1643), relying in particular on the Michelin II judgment (Case T-203/01) and the judgments linked to the Compagnie Maritime Belge case (Cases T-24/93, T-25/93, T-26/93 and T-28/93).
Beyond the legal discussion on the possible distinction between Articles 101 and 102 TFEU in this respect, the economic question could be formulated in these terms: should strategies that do not correspond to a standard of performance competition but of impediment competition not be qualified as anti-competitive without it being necessary to verify that the targeted undertakings are as efficient as the dominant operator? The Commission responded to the question by establishing that: “it is therefore undisputed that the naked restrictions pursued an anticompetitive object, clearly fell outside the scope of competition on the merits and were very capable by their very nature of restricting competition” (para 44).
In this case, there is a presumption of anti-competitive effects. In its initial decision of May 2009, the Commission responded to Intel’s arguments relating to the need to demonstrate that the conduct in question had effectively an anti-competitive foreclosure effect on AMD (para 1667) and that consumers had suffered harm in terms of their freedom of choice through the hindrance of access to the markets for products equipped with competing CPUs (para 1670). The Commission then emphasised the absence of convincing arguments from Intel as to any objective justification for its behaviour or as to any efficiency gains associated with it (para 1680).
The more fundamental question is whether competition on the merits should be based on a standard of behaviour, i.e., a requirement of fairness or loyalty, e.g. a standard of conduct, or whether it should be judged solely in terms of efficiency. The answer is not obvious. The issue of payments in this case could also be put into perspective with the issue of pay-for-delay agreements in the pharmaceutical industry.
If we consider that the naked restrictions should be treated solely on the basis of their anti-competitive effects, should we then apply the same rule as for exclusivity rebates, i.e., implement the equally efficient competitor test? There is some debate in the literature as to whether this test should become a standard applicable to both price-based and non-price-based strategies (see Auer and Radic here) (2023). In this case, the debate could move upstream: is the test valid for strategies whose purpose is to hinder market access for third-party products?
The Commission has rejected this option in the present case, but the question can be left open since the economic effects of an exclusivity rebate or a payment may in principle be equivalent. However, it is undeniable, that in the case of the naked restriction, the very notion of competition on the merits or the dominant operator’s particular responsibility to maintain the level playing field must be considered. Once again, the fundamental difference between impediment strategies and performance strategies that emerged from the ordoliberals provides us with a useful compass.