The Intel saga: what went wrong with the Commission’s AEC test (in the General Court’s view)?

Abstract

ABSTRACTThe General Court’s annulment of the European Commission’s finding that Intel’s conditional rebate scheme was abusive underscores the Court’s readiness to scrutinize in detail the economic analysis, including the as-efficient competitor (AEC) test. This paper critically reviews some of the key errors that the Commission, according to the Court, made in relation to the implementation of that test, focusing on some of its main ingredients, namely the contestable share of the market, the conditional portion of the rebates, and the relevant cost benchmark. We conclude that the Court’s assessment provides useful lessons for how to perform thorough and robust economic analysis not only within the context of an AEC test but in competition cases more generally. At the same time, considering the test’s intrinsic limitations, we find that, aside from the test’s implementation, its informative value should also be explored, based on economic theory and the facts of the case.KEYWORDS: Abuse of dominant positionas-efficient competitor testexclusivity rebatesforeclosure strategymicroprocessors marketJEL: D43K21L12L42 AcknowledgmentsFor helpful comments and suggestions, I would like to thank Nicola Tosini, David Matthew, C.-Philipp Heller, and an anonymous referee. The opinions and views stated in this text do not necessarily reflect those of my employer. All remaining errors are my own.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 See Decision of the European Commission of 13 May 2009 – Intel (COMP/37.990).2 See Judgment of the General Court of 26 January 2022 – Intel v Commission (T-286/09 RENV, EU:T:2022:19).3 See Judgment, Article 1.4 See GC 2014: Judgment of the General Court of 12 June 2014 – Intel v Commission (T-286/09, EU:T:2014:547).5 See CJEU 2017: Judgment of the Court of Justice of 6 September 2017 – Intel v Commission (C-413/14 P, EU:C:2017:632).6 Judgment, para. 482.7 Judgment, para. 524.8 Judgment, para. 526.9 See James Killick, Assimakis Komninos, and Peter Citron, ‘EU General Court demands a vigorous effects-based analysis for rebates cases and annuls the European Commission’s Intel decision and the €1.06 billion fine’ (2022) White & Case .10 See Judgment, para. 529 and Article 1.11 See Official Journal of the European Union from 7 June 2022, Appeal brought on 5 April 2022 by European Commission against the judgment of the General Court (Fourth Chamber, Extended Composition) delivered on 26 January 2022 in Case T-286/09 RENV, Intel Corporation v Commission (case C-240/22 P).12 See Judgment, section III. B. (pp. 25–76), out of a total of 88 pages. By contrast, the GC’s assessment of the criteria set out by the CJEU receives considerably less attention; see Judgment, section III. C. (pp. 76–82).13 See Decision, para. 5.14 See Decision, Article 1 a) – d).15 See Decision, Article 1 e).16 See Decision, Article 1 f) – h).17 Giorgio Monti, ‘Taming Digital Monopolies: A Comparative Account of the Evolution of Antitrust and Regulation in the European Union and the United States’ (2022) 67(1) Antitrust Bulletin 40–68, observes that the Commission, in focusing on Intel’s commercial relations with downstream customers, took a narrower scope than the Federal Trade Commission (‘FTC’). By contrast, the FTC also zoomed in on Intel’s behaviour vis-à-vis market access for producers of Graphic Processing Units (‘GPUs’) as well as IP licensing issues. For a comparison of the theories of harm pursued by the FTC and the European Commission, see also Patrick DeGraba and John Simpson, ‘Loyalty Discounts and Theories of Harm in the Intel Investigations’ (2014) 2(1) Journal of Antitrust Enforcement 170–202.18 See GC 2014, Article 1.19 GC 2014, para. 85.20 CJEU 2017, para. 138.21 See CJEU 2017, para. 139.22 For this interpretation, see Judgment, para. 126.23 See CJEU 2017, para. 142.24 CJEU 2017, para. 143.25 See Decision, section VII. 4.2.3 (pp. 302–453), out of a total of 517 pages (in the non-confidential version of the Decision).26 CJEU 2017, para. 144.27 Judgment, para. 521.28 See Judgment, para. 482.29 See Judgment, para. 86–96.30 See Judgment, para. 529 and Article 1.31 See Germain Gaudin and Despoina Mantzari, ‘Google Shopping and the As-Efficient-Competitor Test: Taking Stock and Looking Ahead’ (2022) 13(2) Journal of European Competition Law & Practice 125–35.32 See Pablo Ibáñez Colomo, ‘On Case C-377/20, Servizio Elettrico Nazionale (III): as efficient competitor principle and effects’ (2022) Chillin’Competition .33 See Pablo Ibáñez Colomo, ‘Anticompetitive Effects in EU Competition Law’ (2021) 17(2) Journal of Competition Law & Economics 309–63.34 See Miroslava Marinova, Fidelity Rebates in Competition Law – Application of the ‘As Efficient Competitor’ Test (Wolters Kluwer 2018), and Kai-Uwe Kühn and Miroslava Marinova, ‘The Role of the ‘As Efficient Competitor’ Test after the CJEU Judgment in Intel’ (2018) 4(2) Competition Law & Policy Debate 64–72.35 CJEU 2017, para. 139.36 See Miroslava Marinova, ‘The EU General Court’s 2022 Intel Judgment: Back to Square One of the Intel Saga’ (2022) 7(2) European Papers, European Forum, Insight 627–39.37 See Judgment of the Court of Justice of 14 October 2010 – Deutsche Telekom v Commission (C-280/08 P, EU:C:2010:603).38 See Judgment of the Court of Justice of 17 February 2011 – Konkurrensverket v TeliaSonera Sverige AB (C-52/09, EU:C:2011:83).39 See Judgment of the Court of Justice of 27 March 2012 – Post Danmark A/S v Konkurrencerådet (Case C–209/10, EU:C:2012:172).40 See Miroslava Marinova, ‘The EU General Court’s 2022 Intel Judgment: Back to Square One of the Intel Saga’ (2022) 7(2) European Papers, European Forum, Insight 627–39.41 See CJEU Post Danmark II: Judgment of the Court of Justice of 6 October 2015 – Post Danmark A/S v Konkurrencerådet C-23/14, EU:C:2015:651), para. 51–62.42 CJEU Post Danmark II, para. 61.43 GC Google Android: Judgment of the General Court of 14 September 2022 – Google and Alphabet v Commission (Google Android) (T-604/18, EU:T:2022:541), para. 643.44 See Patrick Rey and Jean Tirole, ‘A Primer on Foreclosure’ in Mark Armstrong and Robert Porter (eds) Handbook of Industrial Organization (Vol. 3, Elsevier 2007) 2145–220.45 See Chiara Fumagalli, Massimo Motta, and Claudio Calcagno, Exclusionary Practices: The Economics of Monopolisation and Abuse of Dominance (Cambridge University Press 2018).46 See Chiara Fumagalli and Massimo Motta, ‘Buyers’ Miscoordination, Entry and Downstream Competition’ (2008) 118(531) Economic Journal 1196–222.47 See Eric Rasmusen, John Ramseyer, and John Wiley, ‘Naked Exclusion’ (1991) 81(5) American Economic Review 1137–145.48 See Philippe Aghion and Patrick Bolton, ‘Contracts as a Barrier to Entry’ (1987) 77(3) American Economic Review 388–401.49 See John Simpson and Abraham Wickelgren, ‘The Use of Exclusive Contracts to Deter Entry’ (2001) Federal Trade Commission – Bureau of Economics, Working Paper No. 241.50 David Spector, ‘Loyalty Rebates: An Assessment of Competition Concerns and a Proposed Structured Rule of Reason’ (2005) 1(2) Competition Policy International (at p. 99).51 See Chiara Fumagalli and Massimo Motta, ‘Exclusive Dealing and Entry, when Buyers Compete’ (2006) 96(3) American Economic Review 785–95.52 See John Simpson and Abraham Wickelgren, ‘Naked Exclusion, Efficient Breach, and Downstream Competition’ (2007) 97(4) American Economic Review 1305–320.53 See Jose Miguel Abito and Julian Wright, ‘Exclusive Dealing with Imperfect Downstream Competition’ (2008) 26(1) International Journal of Industrial Organization 227–46.54 See Patrick DeGraba, ‘Naked Exclusion by a Dominant Input Supplier: Exclusive Contracting and Loyalty Discounts’ (2013) 31(5) International Journal of Industrial Organization 516–26.55 Decision, para. 1003–1005.56 See Communication from the European Commission from 24 February 2009, Guidance on the Commission’s enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings, para. 44.57 Ibid, para. 45.58 Ibid, para. 27.59 This example starts with a (retroactive) percentage discount to determine the customer’s savings. Equivalently, one could take a fixed payment (lump sum) as the starting point and work out the implied percentage rebate.60 In this example, at a price at the level of the cost ($35), the customer can save $65 per unit by ordering from the as-efficient competitor. By purchasing about 0.77 m units, it can compensate the rebate payments it would otherwise have obtained by staying with the dominant firm ($65 × 0.77 m ≈ 50 m). The required share for the as-efficient competitor to be able to match the dominant firm’s rebate is therefore about 7.7% (0.77 m / 10 m), which is less than the contestable share (25%) – the AEC test is passed.61 For an in-depth review of the numerous criticisms, see Miroslava Marinova, Fidelity Rebates in Competition Law – Application of the ‘As Efficient Competitor’ Test (Wolters Kluwer 2018) (at pp. 168–71 in particular).62 See Patrick DeGraba, ‘Naked Exclusion by a Dominant Input Supplier: Exclusive Contracting and Loyalty Discounts’ (2013) 31(5) International Journal of Industrial Organization 516–26.63 See Steven Salop, ‘The Raising Rivals’ Cost Foreclosure Paradigm, Conditional Pricing Practices, and the Flawed Incremental Price-Cost Test’ (2017) 81 Antitrust Law Journal 371–421.64 Cristina Caffarra, ‘The EU General Court Confirms Android Abuse of Dominance through Tying, with the Real Legacy of the Case Extending Far Beyond (Google Android)’ (2022) Concurrences (Preview) .65 See Giacomo Calzolari and Vincenzo Denicolò, ‘Loyalty Discounts and Price-Cost Tests’ (2020) 73 International Journal of Industrial Organization 102589.66 See Joshua Wright, ‘Simple but Wrong or Complex but More Accurate? The Case for an Exclusive Dealing-Based Approach to Evaluating Loyalty Discounts’ (2013) Speech delivered at the Bates White 10th Annual Antitrust Conference, Washington DC.67 Wright (n 66) (at p. 18).68 See Kai-Uwe Kühn and Miroslava Marinova, ‘The Role of the ‘As Efficient Competitor’ Test after the CJEU Judgment in Intel’ (2018) 4(2) Competition Law & Policy Debate 64–72.69 For a summary of this debate, see Miroslava Marinova, Fidelity Rebates in Competition Law – Application of the ‘As Efficient Competitor’ Test (Wolters Kluwer 2018) (chapter 6 in particular), and Miroslava Marinova, ‘What Can We Learn About the Application of the as Efficient Competitor Test in Fidelity Rebate Cases from the Recent US Case Law?’ (2018) 41(4) World Competition 523–48.70 See Decision, para. 1202–213. See Judgment, para. 171–73.71 This is on the following two grounds. First, the ASP was around $100–200 (see Judgment, para. 305, which refers to an ASP of $165 (for HP); generally, CPUs are sold within a wide range of prices; see Decision, para. 796), and the cost benchmark used by the Commission, average avoidable cost (AAC), appears to amount to roughly 35% of the ASP (see Decision, para. 1043; but note that the precise cost share remains undisclosed; see Decision, para. 1145–153). In Figure 3, this is reflected by setting the ASP to $100, and the cost to $35. Second, the Commission calculated the contestable share required for Intel to have passed the AEC test at 7.9% (see Judgment, para. 176, with reference to the AEC test calculated for 2005 Q1; the required share is of course higher than the actual contestable share adopted by the Commission, which is 7.1%, since otherwise Intel would have passed the Commission’s AEC test). For this reason, in Figure 3, the discount is set such that the AEC test would indeed be passed if the contestable share was just slightly higher than the Commission’s 7.1% (see also n 60).72 See explanation in n 59.73 In this example, at an exclusivity rebate higher than 5%, the required discount offered by a hypothetical as-efficient competitor quickly becomes more than 100% – that is, an effective price that is negative.74 See Judgment, para. 176. The required share calculated by the Commission refers to the first quarter of 2005.75 See Judgment, para. 207–11.76 See Judgment, para. 269.77 See Judgment, para. 177, 277.78 Judgment, para. 217.79 Judgment, para. 234.80 See Judgment, para. 233.81 Judgment, para. 256.82 See Judgment, para. 260.83 See Judgment, para. 265–66.84 See Judgment, para. 268–69.85 Judgment, para. 271.86 See Judgment, para. 282.87 Judgment, para. 287.88 See Decision, para. 1261. See Judgment, para. 280.89 Of course, when the Commission issued its Decision, it was operating on the presumption that Intel’s conditional rebates were anticompetitive by their very nature.90 In addition to this more quantitative exercise, understanding the intensity of competition in the downstream markets would also be relevant for an evaluation of the incentive to exclude in the first place. See the economic literature discussed in Section 3.2.91 For instance, and although not directly related to the Commission’s AEC test, the Commission noticed that Intel’s operating margin (31%) was significantly higher than AMD’s (4%). See Decision, para. 878.92 See (n 71).93 See Judgment, para. 126. Another question is whether the AEC framework, which does not allow for a return on investment, gives undue discretion to dominant firms’ pricing practices. That is, whether a hypothetical competitor (or a hypothetical investor) would, in the real world, indeed seek to expand its customer base if the margin (over average avoidable costs) it could expect to earn was at about 0% or slightly higher.94 See Decision, Article 1 c).95 The AEC test was restricted to just one quarter, and the results of that test were extrapolated to the rest of the infringement period. See Decision, para. 1408–411. The GC, however, ruled that the Commission made an error of assessment when it extrapolated its findings, as the Commission has not proven that NEC’s figures in that quarter were representative for the remainder of the period. See Judgment, para. 390–411. This extrapolation, which constitutes an error that is separate from the conditional portion of the rebates, will not be discussed further in this paper.96 See Judgment, para. 347.97 See Judgment, para. 342.98 Judgment, para. 354.99 Judgment, para. 380.100 Judgment, para. 372.101 Judgment, para. 389.102 See Judgment, para. 339–40, reproducing Intel’s position. Little can be inferred from public information as to whether this was, from a quantitative perspective, a close call or not in terms of passing the AEC test.103 Unlike in the Dell test, where the information disclosed in the Judgment and Decision allows one, based on some simple calculations, to draw indicative conclusions about the responsiveness of the outcome of the test to a change in the variable of interest (in that case, the contestable share), this does not appear possible with the NEC test (or in the Lenovo test, see below). Therefore, from a purely quantitative perspective, it is difficult to judge whether my reading of the Judgment is warranted or not. It would, however, appear rather odd to reject the conclusions of an analysis (namely, that the AEC test was failed) on the basis of an error, the correction of which, while affecting the numerical outcome of the analysis, would leave its conclusions unchanged.104 A related issue arose in the Commission’s AEC test with Dell. Because it was not clear how Intel would have reacted if Dell were not, in the counterfactual, to stay all-Intel, the Commission assumed, taking into consideration Dell’s own assessments and assumptions, that the rebate payment would have halved. See Decision, para. 1175–192.105 See Judgment, para. 356 (‘[the evidence] cannot support the Commission’s finding regarding the conditionality of ECAP’), as one of several examples. Reference to market shares is then made in the following sentence in that same paragraph (‘(…) it appears that that email exchange corroborates Intel’s argument that MDF was the only rebate dependent on Intel’s market share of NEC’s purchases’.)106 See Decision, Article 1 d).107 See Judgment, para. 415–17.108 See Judgment, para. 420.109 See Judgment, para. 417.110 See Judgment, para. 426. The Commission also remarked that, although it appears that AMD had ‘facilities’ in China, that is not the point because the AEC test is concerned with a hypothetical competitor and not any specific one. In any event, the Commission argued that AMD’s facilities may not be equivalent to Intel’s supply hub. See Judgment, para. 427.111 See Judgment, para. 428.112 See Judgment, para. 429.113 See Decision, para. 1507.114 See Judgment, para. 439.115 Judgment, para. 436.116 Judgment, para. 436.117 Judgment, para. 439.118 See Judgment, para. 443–44, 448–53.119 See Judgment, para. 441.120 See Judgment, para. 445–47.121 See Judgment, para. 455–57.122 See (n 103).123 See Judgment, para. 452.124 In principle, the value of a warranty is closely related to its cost. For example, a warranty for a product that never fails will have a value of zero, but also a cost of zero (except for the warranty stamp, potential fraud, etc.). A warranty for a product worth $100 with a failure probability of 10% will have a value of $10. This value may exceed the cost of the warranty, e.g. where the production cost is only $35 per unit (and therefore less than the price of $100), so that the expected cost at a 10% failure probability is only $3.5. The reverse situation may occur where additional costs arise in the event of a failure, such as the costs for repair or replacement service and administrative costs. The warranty may also provide additional value if consumers are risk averse or the warranty serves as a signal, in the presence of information asymmetries, for a high-quality product. Which of these considerations, if any, played a role in the assessment of the value of the warranty extension, is not clear from the materials available in the public domain.125 The competitor may incur a lower total cost if, for example, providing a warranty extension is less costly when applied to less sales. But this is beside the point because then the measure of cost actually incurred by the as-efficient competitor could not fully compensate Lenovo for waiving exclusivity with Intel, which grants a warranty extension for a much higher sales basis.126 See Decision, section 3.3.1.3.127 See Judgment, para. 442.128 See also Germain Gaudin and Despoina Mantzari, ‘Google Shopping and the As-Efficient-Competitor Test: Taking Stock and Looking Ahead’ (2022) 13(2) Journal of European Competition Law & Practice 125–35, who make a similar point.129 See EC Qualcomm: Decision of the European Commission of 24 January 2018 – Qualcomm (Exclusivity payments) (AT.40220).130 Relatedly, in the Broadcom case, the Commission did not carry out an AEC test, but neither did Broadcom. See Decision of the European Commission of 16 October 2019 – Broadcom (AT.40608).131 See EC Qualcomm, section 11.5.132 See GC Qualcomm: Judgment of the General Court of 15 June 2022 – Qualcomm v Commission (T-235/18, EU:T:2022:358).133 See GC Qualcomm, section III. B. 3 (pp. 42–48).134 See EC Google Android: Decision of the European Commission of 18 July 2018 – Google Android (AT.40099).135 See GC Google Android, Article 2.136 See GC Google Android, Article 1.137 See GC Google Android, para. 1032–114.138 See GC Google Android, para. 651–57.139 See GC Google Android, para. 679–99.140 See GC Google Android, para. 752, 798–99, referring to ‘the AEC test carried out by the Commission’. In the Commission’s Google Android decision, the AEC analysis is carried out in section 13.4.1.2 (‘Competing general search services could not have matched Google’s portfolio-based revenue share payments to OEMs and MNOs’). There, the Commission finds that for a competing general search service to match Google’s offer, it would have had to offer a revenue share payment greater than 100%, which is another way of saying there is pricing below cost.141 GC Google Android, para. 644.142 See GC Google Android, para. 735–52.143 See GC Google Android, para. 758–74.144 See GC Google Android, para. 789–97.145 GC Google Android, para. 799. The GC’s assessment of the Commission’s AEC test in Google Android has also attracted some critical policy remarks. See Cristina Caffarra, ‘The EU General Court Confirms Android Abuse of Dominance through Tying, with the Real Legacy of the Case Extending Far Beyond (Google Android)’ (2022) Concurrences (Preview) .146 See GC Google Android, para. 964–1005.147 See also Christian Burholt and Katrin Kurz, ‘Das Ende des ‘as efficient competitor’-Tests? – Die jüngste Entscheidung des EuG im Fall Intel’ (2022) 4 Wirtschaft und Wettbewerb 182–84.148 https://ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/13796-EU-competition-law-guidelines-on-exclusionary-abuses-by-dominant-undertakings_en