The Intel case

This article was first published in Competition Law Insight on 14 February 2017.

There is an ongoing tension in competition law between the desire for legal certainty as to the categories of behaviour that will give rise to infringements, and the need to assess the impact of such behaviour in context to determine whether it has anticompetitive effects. The economic reality that certain conduct may be either beneficial or harmful to competition depending on context complicates enforcement decisions. Overzealous enforcement on the basis of categories of infringement without exploring effects will lead to false positives – the penalising of arrangements which were, conversely, procompetitive. Likewise, excessive caution in seeking to demonstrate conclusively anticompetitive effects before any infringement finding will lead to false negatives, with anticompetitive conduct escaping punishment.

Against this backdrop, Advocate General Nils Wahl’s (AG Wahl) opinion on the Intel case (Case C-413/14) on 20 October 2016 sought to clarify the standard that the European Commission should be adopting in finding rebate schemes to be an abuse of dominance – and the extent to which the Commission was required to explore effects, rather than penalising on the basis of a form-based assessment. The potential impact of AG Wahl’s opinion should not be understated. If the EU Court of Justice’s (ECJ) subsequent judgment in the case follows AG Wahl’s opinion, this will necessitate a shift in enforcement policy from the Commission, and a move to a more effects-based analysis in future investigations. The implications for companies of such a shift would be potentially a reduction in legal certainty, but also greater confidence that procompetitive schemes would not fall foul of form-based infringement findings.

In this article, we explore (1) the legality of rebate schemes employed by dominant undertakings prior to Intel; (2) the status of the law following the Commission’s Intel decision; and (3) AG Wahl’s opinion, and its implications for the future.

Background

The Intel case in 2009 led to the Commission imposing a fine on Intel of €1.06bn (the largest single fine imposed by a competition regulator on an individual company) in respect of rebates and other “naked restrictions” which the Commission found had  been intended  to exclude  competition  by Intel’s rival, AMD, in the manufacture of a particular type of computer microprocessor (x86 CPUs) (C(2009) 3726 decision of 13 May 2009).

The Intel decision was considered controversial by some as the Commission appeared to follow a form-based line of reasoning, finding anticompetitive conduct on the basis of the legal parameters of previous cases, and explicitly stating that, although it had considered the effects of the conduct in question, it did not need to rely on this to establish an infringement. Intel countered that the Commission’s effects analysis was flawed and that these arrangements had no anticompetitive effects. However, in Intel’s first appeal of the Commission’s decision in 2014, the Commission’s approach was endorsed by the EU General Court (GC), which found that a dominant undertaking cannot employ rebate schemes conditional on exclusivity (or quasi-exclusivity), as these by their very nature restrict competition.

From a competition policy perspective, it would seem uncontroversial that companies should not face sanction – and certainly not €1bn in penalties – for conduct which has no anticompetitive effect. However, there is a legitimate question where types of conduct ostensibly appear to have anticompetitive effects, such as obvious price-fixing cartels, as to how far an authority is required to go to prove those effects. The concern for many observers is that, as has also been observed in cartel case law, there may be an inclination on the part of the Commission to follow the path of least resistance in framing its cases, relying on narrow interpretations of legal precedent to avoid protracted consideration of the anticompetitive effects of the arrangements in question. The key question for rebates, which fundamentally entail  a reduction in price to customers, is whether they should be judged against the same legal standard as price-fixing cartels if they involve explicit exclusivity requirements and are provided by a dominant undertaking.

The legality of rebates by dominant companies prior to Intel

Over the last 35 years, the extent to which dominant companies are entitled to offer rebates to their customers has been an ongoing area of controversy. The jurisprudence of the Commission and EU courts leading up to the Commission’s Intel decision in 2009 had tended to favour a form-based, over an effects-based, approach in assessing the legality of rebates where these involved requirements to commit the majority of business to the dominant supplier.

In the leading case of Hoffmann-La Roche (Case 85/76), the ECJ ruled that loyalty rebates are intrinsically unlawful where they are “conditional on the customer’s obtaining all or most of its requirements … from the undertaking in a dominant position”, with the intention of giving “the purchaser an incentive to obtain his supplies exclusively from the undertaking in a dominant position”. This was on the basis that such schemes “are designed to deprive the purchaser of or restrict his possible choices of sources of supply and to deny other producers access to the market”. In that judgment, the ECJ distinguished between “loyalty”  and  “volume”  rebates, on the basis that loyalty rebates are “designed through the grant of a financial advantage to prevent customers from obtaining their supplies from competing producers”. An exclusivity rebate is intended to prevent competitor sales – but at an economic level is it actually any different than a low price which will also drive sales to the dominant undertaking, and which is only illegal if it is set at a predatory (ie below cost) level? Moreover, this binary approach is open to gaming: a dominant company which sets a volume target rebate at 1,000 units (a level which it expects to be close to the total requirements of a customer) may escape censure, while an equivalent rebate requiring the customer to commit 95% of purchases (even if this is effectively the same as 1,000 units) will be illegal.

In the subsequent Michelin I case (Case 322/81), the rebates in question did not fall into either the “loyalty” or “volume” rebates category identified in Hoffmann-La Roche, leading the ECJ to characterise a new category of rebates in its judgment in Michelin I, namely rebates that may have a loyalty-inducing effect, and therefore be illegal, but only when judged against “all the circumstances”.

The GC in Michelin II (Case T-203/01) was concerned with how best to establish the anticompetitive effect of loyalty rebates, and deemed it sufficient to show that the abusive conduct of the dominant firm “tends to” restrict competition or, in other words, that the conduct “is capable of ” having that effect. Hence, for the purposes of applying article 102, “establishing the anticompetitive object and the anticompetitive effect are one and the same thing”.

The 2008 enforcement priorities guidance paper

The approach adopted by the EU courts was widely criticised (in particular in the 2005 EAGCP report, authored by leading academic economists), on the basis that loyalty rebates by dominant undertakings could have procompetitive effects, and a detailed economic analysis was therefore required on a case-by-case basis to determine their legality.

The European Commission acknowledged this criticism and sought to move away from its prior approach when it published its guidance paper in 2008, in the context of the Commission’s enforcement priorities when deciding what cases to investigate.

The guidance paper stipulates that the Commission should follow an effects-based approach when assessing whether to open an investigation into a potential abuse of dominance, including in respect of loyalty rebates. Indeed, the Commission clarifies that a loyalty rebate will not necessarily be anticompetitive, even if it has the effect of making goods and/or services below average total cost.

Further, the “as efficient competitor” (AEC) test should be applied in assessing whether a loyalty rebate is anticompetitive, to determine whether a loyalty rebate is capable of excluding a competitor that is as efficient as the dominant company. The guidance paper therefore acknowledges that the carrot – the size of the rebate which leads to exclusivity – must be judged as a price-based abuse, rather than the simple fact of exclusivity amounting to an abuse.

The Intel decision

When the Commission issued its decision against Intel in 2009, it was clear that it was not intending to depart from its prior decisional practice.

The Commission  found Intel  guilty  of two  separate infringements intended to freeze out AMD:

• rebates conditional on major Intel customers buying all or the vast majority of their x86 CPUs from Intel to be used in respect of specific product ranges of the customers in question; and
• payments made to certain customers conditional on their postponing or cancelling the launch of products based on CPUs produced by AMD (so-called “naked restrictions”).

In reaching its decision, the Commission stated that it was not bound by the guidance paper, on the basis that it did not apply to proceedings opened before its publication. Instead, the Commission relied on the EU precedents to identify three possible categories of rebates: (1) loyalty rebates, which are presumptively illegal; (2) volume-based rebates, which are presumptively legal; and (3) rebates having a loyalty-inducing effect, which may be illegal, but only when judged against “all the circumstances”. The Commission’s case characterised Intel’s rebates as loyalty rebates by a dominant undertaking, making it “unnecessary to verify whether they are capable of restricting competition in a specific case”.

The Commission nevertheless undertook an assessment of the effects of the loyalty rebates in line with its guidance paper, concluding that a hypothetical AEC would be unable to match them without pricing its products below average total costs – so that these rebates were effectively predatory, as well as explicitly linked to exclusivity. The Commission rejected Intel’s arguments that it would have been more suitable to run this test on actual data relating to AMD rather than to a hypothetical competitor, finding in any event the effects-based assessment “should not be regarded as a necessary or absolute test”, pursuant to the jurisprudence of the EU courts. 

The status of loyalty rebates following the Commission’s Intel decision

The requirement of a more effects-based approach was evident in the ECJ’s rulings in TeliaSonera (C-52/09) and Post Danmark I (C-209/10) in 2014. These cases concerned pricing abuses rather than exclusivity rebates, and applied the AEC test to establish whether the prices were effectively predatory. Many considered these judgments indicative that the EU courts and Commission would be following a more effects-based approach going forward.

However, the more recent cases concerning exclusivity rebates have continued to rely on a more formalistic approach. In the lower tier appeal in the Intel case in 2014 (T-286/09), the GC upheld the Commission’s decision against Intel in full, and actually went further than the Commission in relying solely on a form-based analysis, stating that “exclusivity rebates granted by an undertaking in a dominant position are by their very nature capable of restricting competition”, and rejecting the requirement for any effects-based analysis (including the AEC test). The GC disregarded that AMD had grown its market share during the period of the abuse (suggesting that Intel’s  exclusivity  rebates  were  not  particularly  effective),  in addition to Intel’s arguments that AECs would have been able to match the loyalty rebates offered (ie that the size of the carrot needed to be considered). While the GC acknowledged that “exclusivity conditions may, in principle, have beneficial effects for competition, so that in a normal situation on a competitive market, it is necessary to assess their effects on the market in their specific context”, it deemed that this was not applicable in this case, because competition was already restricted due to Intel’s dominant position.

In 2015, the ECJ’s judgment in Case C-23/14 Post Danmark II (the current leading EU authority on rebates) was hotly anticipated, with some expectation that the ECJ would reverse the approach adopted by the GC in Intel, and lean toward its own previous findings in TeliaSonera and Post Danmark I by requiring consideration of the AEC in a rebates case.

Post Danmark II concerned rebates offered in 2007-08 by the Danish postal incumbent to its direct-mail customers, contributing to the exit of its only rival, Bring Citymail. However, the ECJ ruled that there was no requirement to conduct an effects-based assessment of loyalty rebates, and no absolute requirement to apply the AEC test to determine whether such rebates have a loyalty-inducing effect. The ECJ also harked back to Michelin II in stating that the anticompetitive effect of a rebate scheme must be “probable”, but there was “no need to show that it is of a serious or appreciable nature”.

This was a real setback for the proponents of an effects-based approach, as the threshold outlined by the ECJ confirmed that any rebate imposed by a dominant company which was linked to exclusivity was likely to be found illegal. However, the circumstances of the case, whereby the incumbent had a 95% market share in bulk mail and whose rebates contributed to its rival’s exit, made it understandable that the ECJ deemed it unnecessary to examine the effects of the conduct in any detail.

AG Wahl’s Intel opinion

AG Wahl’s opinion comes as part of Intel’s appeal to the ECJ – and, while not binding, the ECJ follows attorney general opinions in around three-quarters of its judgments. AG Wahl, strikingly, finds the Commission and GC wrong on five of Intel’s six grounds of appeal. Significantly, AG Wahl’s opinion looks again at the case law on which the Commission and GC had relied, and interprets it differently, finding a requirement that the case against Intel considers more closely whether the practices in question were “capable” of restricting competition, when considering “all the circumstances”, the “legal and economic context”, and whether the agreements had any “immediate, substantial and foreseeable anticompetitive effect in the EEA”.

AG Wahl found that the GC had erred in creating a “super-category” of exclusivity rebates that were unlawful, with no possibility of rebuttal, and that the illegality of all rebates needed to be judged in “all the circumstances” in which they exist. The Commission’s failure to provide robust evidence on anticompetitive effects therefore meant its decision was legally incorrect. This failure also tainted the reasoning that the rebates were “capable” of restricting competition, as the Commission had failed to establish that the conduct had “in all likelihood” had an anticompetitive effect. The message was clear: the Commission needed to demonstrate in context the harm arising from these allegedly illegal practices – as the old cases had done in AG Wahl’s view, by considering “all the circumstances” before establishing rebates in the “exclusivity” category.

AG Wahl further criticised the GC’s assessment that Intel’s behaviour was illegal, even during the two years of the alleged infringement, where as little as 14% of the market was affected by Intel’s restrictive clauses. AG Wahl did not conclude whether 14% market coverage was sufficient to lead to an abuse, stating that finding exclusivity rebates automatically abusive where they account for a very small proportion of the market (“let us say, for the sake of argument, 3%”) could lead to “unwarranted” outcomes – again, highlighting the Commission’s failure to consider the actual effects of the rebates in establishing a long duration of infringement.

AG Wahl criticised the Commission’s decision further: (1) it artificially segmented the market in order to characterise rebates as “exclusivity” rebates for specific products, even though customers could still buy a significant proportion – or majority – of their overall x86 CPUs from AMD; (2) the Commission made procedural errors in not recording third- party interviews which it relied on in its decision; and (3) the Commission failed to establish any impact in the EU (or EEA) of Intel’s agreements with Lenovo (which related to sales in China), which put them outside its jurisdiction. The first and third of these criticisms can again both be linked to the Commission’s failure to consider adequately the effects of Intel’s conduct: if customers could actually still buy most of their CPUs from AMD without restriction, was this really an exclusivity rebate? And if Intel entered an agreement with a Chinese computer company which foreclosed AMD’s ability to supply in China, but not in the EU, why was the Commission seeking to punish this?

Although AG Wahl agreed with the GC and ECJ’s findings in previous cases that an AEC test may not always be required to assess the effects of a loyalty rebate, he found the GC had erred in law in the present circumstances to have rejected the AEC test, “precisely because that test was carried out by the Commission in the decision at issue” (even if the Commission chose not to rely on its own analysis).

Intel judgment – the final word?

The ECJ’s Intel judgment is expected soon. If the ECJ supports AG Wahl, this will be hailed as a watershed in favour of a more economics-driven analysis. But is this really the case? Or will the Commission and other agencies in future just be more careful in establishing that, where a dominant company imposes exclusivity rebates, the context needs to be considered more thoroughly to ensure these are what they appear to be? It is to be expected that – on at least some of these grounds – the Commission will be defeated. In such circumstances, it is likely the ECJ will remit the case to the GC, and the GC may then reformulate the case against Intel to justify the original finding and impose a (probably reduced) penalty.

But for companies that may be in a dominant position, the 17-year Intel saga in itself (from the initial AMD complaint) sends a message that rebates linked to loyalty or exclusivity can cause significant problems, and are probably better structured in ways which do not explicitly link to a high proportion of customers’ purchases.