Three’s a charm for Intel: On judicial review of economic evidence to rebut a legal presumption

In 2009, the European Commission imposed what was then a record-breaking fine of EUR 1.06 billion on Intel for abuse of dominance. Intel’s initial action for annulment was dismissed by the General Court (GC) of the EU in its 2014 judgment, but Intel’s appeal succeeded before the European Court of Justice (ECJ), which annulled the GC’s judgment and referred it back to the GC in 2017. The ECJ ordered the GC to do a complete and detailed judicial review of the various pieces of economic evidence used by the Commission to prove the infringement and submitted by Intel to rebut it. On 26 January 2022, the GC at long last and after an exhaustive judicial review, annulled the fine. This is the third judgment – but potentially not the last – in what can now be called the Intel saga. It is a story of judicial review of economic evidence in the light of legal presumptions. But it is also about the ‘as efficient competitor-test’ and the stringency with which it must be applied. In describing the Intel saga in this blog, we will touch on all these issues; we will also try to answer what is likely (or unlikely) to happen in this case going forward and whether the EU has to repay significant default interest (as well as the principle sum) as a result of the annulment of the fine. 

To guide you in following the saga, we have created the timeline below: 

Period Event
2002 – 2007 Period of alleged infringement
13 May 2009 Commission decision in case AT37.990
22 July 2009 – 12-June-2014 GC proceedings in case T-286/09
26 August 2014 – 6-September2017 ECJ proceedings in case C-413/14 P
14 November 2017 – 26 January 2022 GC proceedings in case T-286/09, RENV

Commission’s findings

According to the European Commission, Intel had at least a 70% market share in the x86 processors, which are used in computers. The high market share, in combination with other factors, led the Commission to the conclude that Intel held a dominant position. The Commission also decided that Intel had abused its dominant position in two ways. First, by granting: (i) fidelity/exclusivity rebates to four major computer manufacturers, provided that they equip all (or nearly all) of their computers with an Intel chip; and (ii) an exclusivity bonus to a major retailer to only sell computers with Intel chips. Second, the Commission found that Intel also engaged in naked restrictions of competition by paying the computer manufacturers to delay, cancel or restrict the distribution of computers powered by Intel’s competitor, Advanced Micro Devices (AMD). AMD complained to the Commission about these practices and used US discovery procedures to get hold of incriminating documents which were in Intel’s possession to hand them over to the Commission (in a milestone judgment the US Supreme Court even allowed US discovery rules for this purpose). The Commission found that these practices excluded potential competitors from entering the market (foreclosing effect) and thereby infringed Article 102 TFEU. It therefore imposed a fine of EUR 1.06 billion on Intel.

Change from per se abuse to rebuttable presumption with effects-based approach

The ECJ’s milestone case on fidelity (or loyalty) rebates is Case 85/76 Hoffmann-La Roche v Commission (ECLI:EU:C:1979:36). With respect to the foreclosing effect of such rebates, the ECJ held that 

[t]he fidelity rebate, unlike quantity rebates exclusively linked with the volume of purchases from the producer concerned, is designed through the grant of a financial advantage to prevent customers from obtaining their supplies from competing producers‘ (par. 90).

The ECJ assumed that the nature of fidelity rebates is to restrict competition. The proof of the anti-competitive nature was then enough for the Commission to prove an infringement, making an analysis of the effect unnecessary. Article 102 TFEU does not distinguish between competitive restrictions by object or competitive restrictions by effect as Article 101 TFEU does for anti-competitive agreements, but this approach towards fidelity rebates is in our view comparable to the object and effect restrictions under Art. 101 TFEU. With regard to the difference between those two types of restrictions in Article 101 TFEU, the ECJ made clear in its 2012 judgment in Case C-226/11, Expedia (ECLI:EU:C:2012:795) that the anti-competitive nature of the behaviour is the distinguishing factor between the two types of restrictions.

In that regard, the Court has emphasised that the distinction between ‘infringements by object’ and ‘infringements by effect’ arises from the fact that certain forms of collusion between undertakings can be regarded, by their very nature, as being injurious to the proper functioning of normal competition. It must therefore be held that an agreement that may affect trade between Member States and that has an anti-competitive object constitutes, by its nature and independently of any concrete effect that it may have, an appreciable restriction on competition‘ (par. 36-37).

By using the nature-criteria with regard to fidelity rebates, the ECJ seemed to have introduced a similar distinction in Article 102 TFEU by allowing the proof of the anti-competitive nature to be sufficient and independently of any concrete effect. For example, that is what happened in the fidelity rebate case concerning Tomra. In Case C549/10 P, Tomra (ECLI:EU:C:2012:221), the ECJ held that both the GC and the Commission were right in proving that a fidelity rebate is capable of restricting competition (its abusive nature) without having to make an economic analysis of the actual effect of the rebates on the relevant markets (par. 67-68). The ECJ added that

that the loyalty mechanism was inherent in the supplier’s ability to drive out its competitors by means of the suction to itself of the contestable part of demand. When such a trading instrument exists, it is therefore unnecessary to undertake an analyse of the actual effects of the rebates on competition given that, for the purposes of establishing an infringement of Article 102 TFEU, it is sufficient to demonstrate that the conduct at issue is capable of having an effect on competition’ (par. 79).

Although fidelity rebates were presumed to constitute a prohibited behaviour, the Commission expressed a change in the way it wanted to enforce Article 102 TFEU against foreclosure by dominant undertakings. Instead of relying on these kinds of per se abuses, it embraced a more economic and effects based approach in its 2009 Guidance on enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings. At the time, the Commission stated the following on possible abusive pricing behaviour by dominant undertakings.

In order to determine whether even a hypothetical competitor as efficient as the dominant undertaking would be likely to be foreclosed by the conduct in question, the Commission will examine economic data relating to cost and sales prices, and in particular whether the dominant undertaking is engaging in below-cost pricing‘ (par. 25).

In its 2009 decision against Intel (case AT37.990), the Commission admitted that its guidance only came out after the administrative procedure against Intel started. As a result, it stated that it did not apply its guidance to the Intel case. Nevertheless, the Commission concluded that the Intel decision was in line with the economic approach it developed in its Guidance(par. 916). Consequently, the Commission did not apply the Guidance itself, but the economic approach the Commission advocated in the Guidance and did not only prove the anti-competitive nature of the fidelity rebates, but also conducted an in-depth analysis of the individual circumstances of the case based on data from Intel.

First instance and appeal

Intel brought an action for annulment against that Commission decision before the GC. Its points of appeal narrowed down to two fundamental issues. First, Intel challenged the Commission’s jurisdiction to enforce EU competition law against it. Both the GC and the ECJ ruled that the Commission had jurisdiction to enforce EU competition law. (1)

Second, Intel accused the Commission of failing to meet the infringement’s standard of proof. According to Intel, the Commission had merely proven the anti-competitive nature of the contested conduct, but not that it had an actual anti-competitive effect. In Intel’s view, the Commission had to prove this on a case-by-case basis. This was necessary to distinguish fair competition on the merits from illegal fidelity rebates which exclude competitors. In other words: were the rebates which Intel had offered (and thereby lower prices to the benefit of the customers) fair price competition or illegal abuse to foreclose competitors from the market (in the long run). An economic test to distinguish between fair and unfair competition on prices, is the so-called ‘as-efficient-competitor’ (AEC) test. The Commission indicates in its 2009 Guidance that it will use this test to analyse price-based exclusionary conduct (par. 23-27). If the Commission had found that there had been competition on the merits, which could exclude less efficient competitors, there would not have been a problem, as this type of competition is not prohibited. The only type of behaviour that is prohibited is when equally efficient (or more efficient) competitors are excluded by the conduct.

Intel argued that, in terms of the rebates for each of the four computer manufacturers and for the bonus to the retailer, the Commission had to have proven that a competitor at least as efficient as Intel was foreclosed by the rebates and bonus applied. In other words, that a competitor such as AMD (provided it is as cost efficient or even more efficient than Intel) could not have offered processors at equivalent net prices without falling below cost. Despite the Commission’s primary view that the proven anti-competitive nature was sufficient to prove breach of Article 102 TFEU, the Commission had nevertheless conducted a thorough analysis of the individual circumstances of the case, including the AEC test. Intel blamed the Commission for making errors in the detailed analysis. 

On 12 June 2014, the GC ruled in Case T-286/09, Intel v Commission (ECLI:EU:T:2014:547), that the Commission did not have to prove the exclusionary effect of each of the rebates and bonus. Instead, the nature of a fidelity rebate and exclusivity bonus was sufficient to conclude that there had been abuse. The GC also held that the Commission did not have to prove that a theoretically equally efficient competitor would also have been excluded by the challenged exclusive rebates and bonus under the AEC test. In this respect too (and following Hoffman La Roche and equivalent precedents in jurisprudence), the GC found that the anti-competitive nature of the exclusive rebates and bonus alone was enough to prove the infringement, even if the entry of an equally efficient competitor would not have been economically impossible. As regards the naked restrictions, the GC found the anti-competitive object of the “anti-AMD payments” to have been proven and, thus, these payments to be a prohibited competition restriction, without also testing their effects. The GC therefore dismissed Intel’s appeal in its entirety.

Intel subsequently lodged an appeal against the GC’s findings. In its judgment of 6 September 2017 in Case C-413/14 P, Intel v Commission (ECLI:EU:C:2017:632), the ECJ embraced the AEC test as a tool to prove anti-competitive behaviour. 

In that respect, it must be borne in mind that it is in no way the purpose of Article 102 TFEU to prevent an undertaking from acquiring, on its own merits, the dominant position on a market. Nor does that provision seek to ensure that competitors less efficient than the undertaking with the dominant position should remain on the market. Thus, not every exclusionary effect is necessarily detrimental to competition. Competition on the merits may, by definition, lead to the departure from the market or the marginalisation of competitors that are less efficient and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation‘ (par. 133-134).

Already in the 2012 Post Danmark case, the ECJ expressed its acceptance of the AEC test to appreciate abuse behaviour through low, predatory pricing (see the previous European Law Blog post on that case here). The game-changer in the Intel case was that the ECJ shed clarity in its nearly 40-year old judgment in Hoffmann-LaRoche: that a fidelity rebate is presumed to have an anti-competitive effect is a rebuttable presumption (par. 138). (2) Since Intel had submitted evidence during the administrative proceedings to rebut the alleged foreclosure effects, the Commission and, consequently, the GC should have appreciated the rebuttable aspect according to the ECJ, and not simply adhere to the presumption. Furthermore, because the detailed economic analysis including the AEC test was an essential part of the Commission’s decision, the GC was wrong in holding that the AEC test was not relevant. As the GC had therefore failed to examine whether the Commission had carried out those analyses correctly, whether the Commission had proven the infringement had never been established. According to the ECJ, the GC had not satisfied the requirements for judicial review for a punitive fine as required by the European Convention of Human Rights (ECHR). The case was referred back to the GC for a judicial review of the facts.

Third time’s a charm:  the GC’s renvoi judgment

On 26 January 2022, the GC issued a ruling in Case T-286/09 RENV, Intel v Commission (ECLI:EU:T:2022:19). In accordance with the judgment of the ECJ, the GC reviewed each discount and bonus that the Commission subjected to the AEC test in its decision. In line with the ECJ’s judgment, the GC held that the anti-competitive nature of the conduct complained of is a rebuttable presumption. Therefore, it believes that the Commission should not have ignored the rebuttals provided by Intel, which could have plausibly explained its conduct other than that of a prohibited restriction on competition. 

The GC borrows the concept of “plausible explanation” from the 1993 Woodpulp II judgment on the merits of those joined cases (ECLI:EU:C:1993:120). In that case on Article 101 TFEU the Commission fined woodpulp producers for illegal collusion between competitors on prices. The Commission had only proved parallel behaviour on pricing and assumed that that was caused by illegal collusion. The ECJ held that the Commission could only rely on the presumption that parallel pricing behaviour was caused by collusion if there was no other plausible explanation for the parallel pricing (par. 126-127). In that case, independent economic experts explained the possibility of what could have caused the parallel behaviour, besides collusion, such as the high degree of transparency in the pulp market (par. 82-88). Since there was another plausible explanation, the ECJ annulled the fine imposed on the wood pulp producers. In the Intel matter, the GC in it 2022 judgment relied on this “plausible explanation” principle to assess whether the Commission’s decision met the required standard of proof (par. 165).

Furthermore, according to the GC, the AEC test is not needed to prove the exclusionary effect of exclusivity rebates. However, because the Commission used the AEC test to prove that exclusionary effect, the GC is now looking at it critically. In 300 paragraphs, the GC subjected to detailed judicial review the AEC tests conducted by the Commission on the fidelity rebates of each of the four computer manufacturers, and the retailer’s exclusivity bonus. The Court found that the Commission committed errors in applying the AEC test to each of the five customers. For example, with respect to the rebates given to computer manufacturer HP, the GC held that the foreclosure effect was not demonstrated for the entire infringement period. For computer manufacturer NEC and the retailer, the GC held that the Commission had erroneously used the data from only a one-quarter year period which it then had extrapolated to the entire infringement period instead of using the real data for the whole infringement period. As regards the naked restriction of competition caused by the “anti-AMD” payments, the GC notes that the ECJ did not rule on this aspect of the decision in the appeal, and that the GC’s original opinion in its initial 2014 judgment remains unchanged. At the time of the initial ruling, in calculating the fine, the GC had asked the Commission to arrive at a relative value of the naked restrictions on competition. As the Commission replied that it had looked at the conduct as a whole, the GC now considers itself unable to reach a partial annulment and set the fine at a lower amount. It therefore annulled the Commission’s decision in its entirety.

Is this the end of the Intel saga?

What remains for both the Commission and Intel to answer is: what happens now? There are three possible scenarios. First, the Commission can leave the case at rest and pay the costs (and possible default interest). Second, the Commission can start all over again and try to overcome the faults and flaws the GC found in the AEC test. If new economic testing again leads the Commission to the conclude that foreclosure of an as efficient competitor has taken place, the Commission could adopt a new decision with a fine. Such re-adoption after annulment can however be time consuming as previous re-adopted decisions demonstrate (one year in the carbonless paper case; two years for ICAP in the Yen Interest Rate Derivatives case; three years in the steel beams case; and even four years in the recently re-adopted fine for Telefónica and Pharol (formerly Portugal Telecom)).

A final possibility is somewhere in the middle, where the Commission adopts a new decision regarding only the naked restrictions, including an appropriate but lower fine, while leaving the fidelity rebates and exclusivity bonus out of the decision. 

With the last option, the Commission can perhaps offset the costs it has to pay to Intel with the new fine. Those costs would not only consist of the judicial costs which the GC order the Commission to pay Intel, but also possible default interest. Why interest? A week before the third 2022 Intel judgment, the GC handed down its judgment in Case T-610/19, Deutsche Telekom / Commission (ECLI:EU:T:2022:15) on a claim to pay default interest on a partially annulled fine. It held that the Commission had to pay default interest over the part of the fine which was no longer due. (3) This 2022 Deutsche Telekom-case on default interest made us wonder: how large will the default interest be which the EU has to pay to Intel after the annulment? In the Intel case, the fine of EUR 1.06 billion was due back in 2009. Intel might have paid the fine then, or opted to give a bank guarantee instead. The 2009 fine was annulled 13 years later and is therefore no longer due. If Intel paid the fine, a record-breaking amount of default interest might be payable to Intel by the EU in addition to two thirds of the costs incurred by Intel at the three judicial instances within the EU. A new decision on the naked restrictions alone, with a fine would at least be a budget-neutral solution, albeit one that would leave deep-seated frustration, particularly for the Commission’s case handlers and economist team.



(1)  Intel argued that there was no jurisdiction because it is US-based company which only sold its processors to non-European subsidiaries of computer manufacturers Acer and Lenovo. In its 2014 judgment, the GC tried to reconcile its own jurisdictional approach developed in the 1999 Gencor case (on EU Merger Control in a merger between two South African companies) and the criteria developed by the ECJ in its 1988 Woodpulp I judgment. In the latter case, the ECJ held that EU competition law can only fall under public international law in line with the principle of territoriality. According to the ECJ, that was the case in Woodpulp I because the implementation of the alleged anti-competitive behaviour took place within the internal market by, for example, directly selling to customers in the established internal market (even though the alleged illegal collusion took place outside of that market). In Gencor, the GC ruled that there is jurisdiction when the “immediate, substantial and foreseeable effects” can be located within the internal market (the qualified effects test). In its 2014 judgment in the Intel case, the GC held that the two tests are alternative tests to establish whether there is sufficient nexus with the internal market and therefore jurisdiction for the Commission to enforce EU competition law (par. 244). Consequently, the GC applied both tests to Intel’s two-pronged behaviour. On appeal, Intel indicated that the GC erred in law by applying the qualified effects test, because the implementation test was the one and only right test to look at. Appeal in its 2017 judgment, the ECJ held that both tests were indeed suitable and justified in establishing jurisdiction of the Commission to enforce EU competition law in relation to non-EU undertakings (par. 44-45). In its submission to the GC after the referral, Intel withdrew its pleas disputing the Commission’s jurisdiction (par. 74 of the 2022 GC judgment).

(2)  In competition law, many legal presumptions are used, which often lowers the Commission’s burden of proof (or that of the national competition authorities of the member states). Once the European Commission meets the burden, the accused undertaking bears the burden and has to rebut the presumption. A question on the extent of these legal presumptions do not contradict the presumption of innocence as set in Article 48 of the EU Charter of Fundamental Rights or the right to a fair trial of Article 47 of the Charter. Nevertheless, the ECJ has allowed rebuttable presumptions in competition law (see for a detailed analysis of these presumptions in EU competition law, the EU’s contribution to the OECD roundtable on legal presumptions). Since the ECJ has set rebuttable presumptions, few undertakings have been able to rebut all of them. Some may even seem incapable of being rebutted (such as the parental liability presumption developed in the 2009 AKZO case for a 100% shareholder company).

(3)  In 2018, the GC had to rule on a Commission decision regarding abuse of dominance on the Slovak internet market. For that behaviour, the Commission imposed a fine jointly and severally on Deutsche Telekom and its subsidiary Slovak Telecom. The GC partially annulled the fine in Case T851/14, Slovak Telekom (ECLI:EU:T:2018:929) and Case T-827/14, Deutsche Telekom (ECLI:EU:T:2018:930), upheld by the ECJ in 2021 in Case C-165/19 P (ECLI:EU:C:2021:239). The Commission repaid the portion of the overall fine of EUR 12 million which was no longer due, but it refused to pay default interest and additional damages. Deutsche Telekom challenged that refusal before the GC. A week before the third 2022 Intel judgment, the GC handed down its judgment on these claims in Case T-610/19, Deutsche Telekom / Commission (ECLI:EU:T:2022:15). It held that the Commission had to pay default interest. The GC used the 3.55% rate the European Central Bank (ECB) applies to refinancing operations for the four years between payment and repayment. The end amount was EUR 1.8 million. No additional damages were awarded.