CJEU, 5th chamber, October 24, 2024, No C-240/22 P

THE COURT (Fifth Chamber),

1 By its appeal, the European Commission asks the Court of Justice to set aside the judgment of the General Court of the European Union of 26 January 2022, Intel Corporation v Commission (T‑286/09 RENV, ‘the judgment under appeal’, EU:T:2022:19), in so far as it annulled, first, Article 1(a) to (e) and Article 2 of Commission Decision C(2009) 3726 final of 13 May 2009 relating to a proceeding under Article [102 TFEU] and Article 54 of the EEA Agreement (Case COMP/C‑3/37.990 – Intel) (‘the decision at issue’) and, second, Article 3 of that decision in so far as it concerns Article 1(a) to (e) thereof.

Background to the dispute and the decision at issue

2 The background to the dispute and the content of the decision at issue are described in paragraphs 1 to 35 of the judgment under appeal. For the purposes of the present proceedings, they may be summarised as follows.

3 Intel Corporation Inc. (‘Intel’) is a company established under United States law which designs, develops, manufactures and markets central processing units (‘CPUs’), chipsets and other semiconductor components, as well as platform solutions for data processing and communications devices.

4 Following a complaint lodged by Advanced Micro Devices Inc. (‘AMD’) on 18 October 2000 and supplemented on 26 November 2003, the Commission launched an investigation into Intel on the basis of Council Regulation (EC) No 1/2003 of 16 December 2002 on the implementation of the rules on competition laid down in Articles [101] and [102 TFEU] (OJ 2003 L 1, p. 1).

5 On 26 July 2007, the Commission notified to Intel a statement of objections concerning its conduct vis-à-vis five major original equipment manufacturers (‘OEMs’), namely Dell, Hewlett-Packard Company (HP), Acer Inc., NEC Corp. and International Business Machines Corp. (IBM).

6 On 17 July 2008, the Commission issued to Intel a supplementary statement of objections concerning its conduct vis-à-vis Media-Saturn-Holding GmbH (‘MSH’), a European retailer of microelectronic devices and the largest desktop computer distributor in Europe. That statement of objections also covered Intel’s conduct vis-à-vis Lenovo Group Ltd (‘Lenovo’) and contained new evidence relating to Intel’s conduct vis-à-vis some of the OEMs covered by the statement of objections of 26 July 2007, which had been obtained by the Commission after that date.

7 On 13 May 2009, the Commission adopted the decision at issue, a summary of which appears in the Official Journal of the European Union (OJ 2009 C 227, p. 13).

8 According to the decision at issue, Intel committed a single and continuous infringement of Article 102 TFEU and of Article 54 of the Agreement on the European Economic Area (EEA) of 2 May 1992 (OJ 1994 L 1, p. 3), from October 2002 until December 2007, by implementing a strategy aimed at foreclosing a competitor, AMD, from the market for x86 CPU microprocessors (‘x86 CPUs’).

9 The goods at issue in the decision at issue are CPUs, which are key components of any computer, both in terms of the overall performance and cost of the system. They are often referred to as the computer’s ‘brain’. The process for the manufacture of CPUs thus requires expensive high-tech facilities.

10 CPUs used in computers can be subdivided into two categories, namely x86 CPUs and CPUs based on another architecture. The x86 architecture is a standard designed by Intel for its CPUs. It can run both the Windows and Linux operating systems. Windows is primarily linked to the instruction set corresponding to the x86 architecture. Prior to 2000, a number of manufacturers of x86 CPUs were present on the market. However, those manufacturers have since exited the market. Since then, Intel and AMD have been essentially the only two companies manufacturing x86 CPUs.

11 The geographical market was defined as worldwide.

12 On the basis, first, of a market share in excess of 70% held by Intel between 1997 and 2007 and, second, of the existence of significant barriers to entry to that market and to expansion resulting from sunk investments in research and development, intellectual property and the necessary production facilities to manufacture x86 CPUs, the Commission concluded that Intel had held a dominant position on that market at least in the period from October 2002 to December 2007.

13 The Commission imputed two types of abusive conduct to Intel. Those are, first, rebates granted to four OEMs – namely Dell, Lenovo, HP and NEC – on condition that those OEMs purchase all or almost all of their x86 CPUs from Intel and payments awarded to MSH, on condition that MSH sold exclusively computers containing x86 CPUs made by Intel (together, ‘the contested rebates’). According to the Commission, those rebates constituted loyalty rebates. Moreover, the conditional payments made to MSH constituted a mechanism equivalent to that of those rebates.

14 In the decision at issue, the Commission carried out an economic analysis concerning the capability of the contested rebates to foreclose a competitor as efficient as Intel, albeit not dominant (as efficient competitor test; ‘the AEC test’). That analysis makes it possible to establish at what price a competitor as efficient as Intel would have had to offer CPUs in order to ‘compensate’ an OEM for the loss of such a rebate, granted by Intel. The same kind of analysis was conducted for the payments granted by Intel to MSH.

15 The evidence gathered by the Commission led it to conclude that the contested rebates induced the loyalty of the ‘key’ OEMs and of MSH. The effects of those rebates were complementary, in that they significantly diminished the ability of Intel’s competitors to compete on the merits of their x86 CPUs. Intel’s anticompetitive conduct thereby resulted in a reduction of consumer choice and in lower incentives to innovate.

16 Second, the Commission imputed to Intel practices constituting naked restrictions consisting of awarding payments to three OEMs, namely HP, Acer and Lenovo, so that they postponed or cancelled the launch of AMD x86 CPU-based products or placed restrictions on the distribution of those products. The Commission concluded that those practices, which did not constitute normal competition on the merits, also directly harmed competition.

17 The Commission concluded in the decision at issue that, in each instance, the conduct imputed to Intel vis-à-vis the OEMs set out above and MSH constituted an abuse under Article 102 TFEU, but that each of those individual abuses was also part of a single strategy aimed at foreclosing AMD, Intel’s only significant competitor, from the market for x86 CPUs. Those individual abuses, it found, were therefore part of a single infringement of Article 102 TFEU.

18 By applying the Guidelines on the method of setting fines imposed pursuant to Article 23(2)(a) of Regulation (EC) No 1/2003 (OJ 2006 C 210, p. 2), the Commission imposed on Intel a fine of EUR 1.06 billion.

19 The operative part of the decision at issue, reproduced in paragraph 35 of the judgment under appeal, reads as follows:

‘Article 1

Intel … has committed a single and continuous infringement of Article [102 TFEU] and Article 54 of the EEA Agreement from October 2002 until December 2007 by implementing a strategy aimed at foreclosing competitors from the market of x86 CPUs which consisted of the following elements:

(a) Granting rebates to Dell between December 2002 and December 2005 at a level that was conditional on Dell obtaining all of its x86 CPU supplies from Intel;

(b) Granting rebates to HP between November 2002 and May 2005 at a level that was conditional on HP obtaining at least 95% of its corporate desktop x86 CPU supplies from Intel;

(c) Granting rebates to NEC between October 2002 and November 2005 at a level that was conditional on NEC obtaining at least 80% of its client PC x86 CPU supplies from Intel;

(d) Granting rebates to Lenovo between January 2007 and December 2007 at a level that was conditional on Lenovo obtaining all of its notebook x86 CPU supplies from Intel;

(e) Granting payments to [MSH] between October 2002 and December 2007 at a level that was conditional on [MSH] selling only computers incorporating Intel x86 CPUs;

(f) Granting payments to HP between November 2002 and May 2005 conditional on: (i) HP directing HP’s AMD-based x86 CPU business desktops to Small and Medium Business and Government, and Educational and Medical customers rather than to enterprise business customers; (ii) precluding HP’s channel partners from stocking HP’s AMD-based x86 CPU business desktops such that such desktops would only be available to customers by ordering them from HP (either directly or via HP channel partners acting as sales agent); and (iii) HP delaying the launch of its AMD-based x86 CPU business desktop in the [Europe, Middle East and Africa] region by six months;

(g) Granting payments to Acer between September 2003 and January 2004 conditional on Acer delaying an AMD-based x86 CPU notebook;

(h) Granting payments to Lenovo between June 2006 and December 2006 conditional on Lenovo delaying and finally cancelling its AMD-based x86 CPU notebooks.

Article 2

For the infringement referred to in Article 1, a fine of EUR 1 060 000 000 is imposed on Intel …

Article 3

Intel … shall immediately bring to an end the infringement referred to in Article 1 in so far as it has not already done so.

Intel … shall refrain from repeating any act or conduct described in Article 1, and from any act or conduct having the same or equivalent object or effect.

…’

The first set of proceedings before the General Court and the Court of Justice

20 By application lodged at the Registry of the General Court on 22 July 2009, Intel brought an action for the annulment of the decision at issue, relying on nine pleas in law.

21 Association for Competitive Technology Inc. (‘ACT’) and the Union fédérale des consommateurs – Que choisir (UFC – Que choisir) were granted leave to intervene in the proceedings in support of Intel and of the Commission respectively.

22 By judgment of 12 June 2014, Intel v Commission (T‑286/09, ‘the initial judgment’, EU:T:2014:547), the General Court dismissed the action in its entirety. The dismissal of that action was based, inter alia, on two premisses concerning the role of the AEC test used by the Commission to analyse the impact of the contested rebates. According to the first of those premisses, a finding that those rebates were illegal did not necessitate an examination of the circumstances of the case, so that the Commission was not required to demonstrate the foreclosure capability of those rebates on a case-by-case basis. According to the second of those premisses, even if an assessment of the circumstances of the case were necessary to demonstrate the potential anticompetitive effects of the loyalty rebates, it would still not have been necessary to demonstrate those effects by means of an AEC test. As part of an examination carried out in the alternative, the General Court held that the Commission had demonstrated, to the requisite legal standard and on the basis of an analysis of the circumstances of the case, that the contested rebates were capable of restricting competition. It followed that the General Court was not required to consider whether the Commission had carried out the AEC test in accordance with the applicable rules and without making errors or to verify that the alternative calculations proposed by Intel for that test had been carried out correctly.

23 By document lodged at the Registry of the Court of Justice on 26 August 2014, Intel brought an appeal against the initial judgment. In support of that appeal, Intel put forward six grounds of appeal.

24 By judgment of 6 September 2017, Intel v Commission (C‑413/14 P, ‘the judgment on the appeal’, EU:C:2017:632), the Court of Justice set aside the initial judgment and referred the case back to the General Court.

25 In particular, after rejecting the fifth and fourth grounds of appeal, alleging, respectively, the misapplication by the General Court of the criteria relating to the Commission’s jurisdiction with regard to the agreements concluded between Intel and Lenovo, and a procedural irregularity, the Court of Justice examined and upheld the first ground of appeal, which alleged an error of law on account of the failure to examine the contested rebates in the light of all the relevant circumstances.

26 In that regard, in paragraphs 133 and 134 of the judgment on the appeal, the Court of Justice emphasised that it is in no way the purpose of Article 102 TFEU to prevent an undertaking from acquiring, on its own merits, a 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 than the dominant undertaking and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation.

27 In paragraphs 135 to 137 of the judgment on the appeal, the Court of Justice nevertheless recalled that, given the special responsibility of an undertaking in a dominant position not to allow its behaviour to impair genuine, undistorted competition on the internal market, it cannot foreclose its competitors considered to be as efficient as it is itself and strengthen its dominant position by using methods other than those that are part of competition on the merits, such as a system of discounts conditional on the customer’s obtaining all or most of its requirements exclusively from that undertaking.

28 That being so, in paragraphs 138 and 139 of the judgment on the appeal, the Court of Justice stated that, where the undertaking concerned submits, during the administrative procedure, on the basis of supporting evidence, that its conduct, consisting of the application of loyalty rebates, was not capable of restricting competition and, in particular, of producing the alleged foreclosure effects, the Commission is not only required to analyse, first, the extent of the undertaking’s dominant position on the relevant market and, second, the share of the market covered by the contested rebates, as well as the conditions and arrangements for granting those rebates, their duration and their amount; it is also required to assess the possible existence of a strategy aiming to exclude competitors that are at least as efficient as the dominant undertaking from the market.

29 Thus, the Court of Justice found, in paragraphs 143 and 144 of the judgment on the appeal, that the AEC test had played an important role in the Commission’s assessment of whether the contested rebates were capable of having foreclosure effects on competitors as efficient as Intel, with the result that the General Court was required to examine all of Intel’s arguments concerning that test.

30 However, as set out in paragraphs 145 and 146 of the judgment on the appeal, the General Court had held, in the initial judgment, that it was not necessary to consider whether the Commission had carried out the AEC test in accordance with the applicable rules and without making any errors, and that it was also not necessary to examine the question whether the alternative calculations proposed by Intel had been carried out correctly. Therefore, in its examination of the circumstances of the case, carried out for the sake of completeness, the General Court had, in the initial judgment, attached no importance to the AEC test carried out by the Commission and, accordingly, had not addressed Intel’s criticisms of that test.

31 In those circumstances, the Court of Justice held, in paragraph 147 of the judgment on the appeal, that it was necessary to set aside the initial judgment, since, in its analysis of whether the contested rebates were capable of restricting competition, the General Court had wrongly failed to take into consideration Intel’s line of argument seeking to expose alleged errors made by the Commission in the AEC test.

32 Thus, in paragraphs 149 and 150 of the judgment on the appeal, the Court of Justice held that it was for the General Court, to which the case had to be referred back, to examine, in the light of the arguments put forward by Intel, whether the contested rebates were capable of restricting competition.

The procedure before the General Court following the referral of the case back to it and the judgment under appeal

33 On 14, 15 and 16 November 2017 respectively, ACT, Intel and the Commission submitted their written observations on the conclusions to be drawn from the judgment on the appeal for the outcome of the proceedings, pursuant to Article 217(1) of the Rules of Procedure of the General Court (‘the main observations on the referral of the case back to the General Court’).

34 On 20 February 2018, ACT and, on 5 March 2018, Intel and the Commission, respectively, lodged supplementary written observations in accordance with Article 217(3) of the Rules of Procedure of the General Court (‘the supplementary observations on the referral of the case back to the General Court’).

35 In its main observations on the referral of the case back to the General Court, Intel, supported by ACT, reiterated, inter alia, its application for annulment of the decision at issue. The Commission, for its part, contended that the General Court should dismiss the action.

36 As a preliminary point, in paragraphs 74 to 102 of the judgment under appeal, the General Court ruled on the subject matter of the proceedings following the referral of the case back to it and, in paragraphs 103 to 111 of that judgment, on the admissibility of Intel’s and ACT’s main and supplementary observations on the referral of the case back to the General Court.

37 As regards the subject matter of the proceedings before it after the case had been referred back to it, the General Court concluded that the proceedings concerned, in essence, the analysis of whether the contested rebates were capable of restricting competition in the light, first, of the clarifications relating to the principles laid down in the judgment of 13 February 1979, Hoffmann-La Roche v Commission (85/76, EU:C:1979:36), set out in paragraph 133 et seq. of the judgment on the appeal and, second, of the main and supplementary observations on the referral of the case back to the General Court regarding the conclusions to be drawn from those clarifications.

38 In that regard, the General Court endorsed the findings in the initial judgment concerning the naked restrictions and their unlawfulness under Article 102 TFEU and the assessments in that judgment concerning the characterisation of the contested rebates as ‘exclusivity rebates’. However, the General Court held that, in accordance with the points of law decided by the judgment on the appeal, characterising the contested rebates as ‘exclusivity rebates’ did not mean that an AEC test was not necessary in order to assess whether those rebates were capable of restricting competition or that it was sufficient for them also to be characterised as abusive for the purposes of Article 102 TFEU.

39 As regards the admissibility of Intel’s and ACT’s main and supplementary observations on the referral of the case back to the General Court, the General Court pointed out that Article 84(1) of its Rules of Procedure does not preclude Intel from devoting a substantial part of those observations to restating the arguments in the application concerning the AEC test, or even to amplifying those arguments. That being so, according to the General Court, the judgment on the appeal does not, as such, constitute a new factor justifying an amendment to or an enlargement of Intel’s complaints against the decision at issue.

40 Next, after recalling the method defined by the Court of Justice for assessing whether a rebate scheme has the capability to restrict competition and the principles arising from the judgment on the appeal, the General Court assessed the merits of the arguments raised by Intel and by ACT.

41 In that context, the General Court examined, as a first step, the argument that the decision at issue is based on an incorrect legal assessment. In that regard, the General Court held that, in taking as a starting point the premiss that it could validly find that the contested rebates infringed Article 102 TFEU on the ground that they were by their very nature abusive, without necessarily having to take into account, for that purpose, the capability of those rebates to restrict competition, the Commission had vitiated the decision at issue by an error of law. That being so, relying on the ‘important role’ that the AEC test had played in the Commission’s assessment of whether the contested rebates were capable of having foreclosure effects on competitors as efficient as Intel, the General Court considered itself bound to examine, as a second step, Intel’s arguments by which it claimed that the AEC test was vitiated by numerous errors.

42 The section of the judgment under appeal devoted to those arguments consists of four subsections. The grounds set out in paragraphs 150 and 151 of the judgment under appeal concern the scope of the General Court’s review.

43 The grounds set out in paragraphs 152 to 159 of that judgment are devoted to general considerations on the AEC test carried out in the decision at issue. Those paragraphs of the judgment under appeal are worded as follows:

‘152 The starting point for the AEC test, as defined in recital 1003 et seq. of the [decision at issue] and applied by the Commission in the present case, is that, given, in particular, the nature of its product, its brand image and its profile, Intel was an unavoidable trading partner and that the OEMs would always have purchased at least part of their CPU requirements from Intel, regardless of the quality of products offered by alternative suppliers. Therefore, customers were willing and able to switch only a certain share of their demand to that alternative supplier (“the contestable share”). Intel’s power to use the non-contestable share as leverage to reduce the price on the contestable share of the market was the result of that status as an unavoidable trading partner.

153 As the [General] Court observed in paragraph 141 of the initial judgment, the AEC test carried out in the [decision at issue] starts from the premiss that [a competitor as efficient as Intel], which seeks to obtain the contestable share of the orders hitherto satisfied by a dominant undertaking, must compensate the customer for the exclusivity rebate which it would lose if it purchased a smaller portion than that stipulated by the exclusivity or quasi-exclusivity condition. The AEC test is designed to determine whether the competitor which is as efficient as the undertaking in a dominant position, which faces the same costs as the latter, can still cover its costs in that case.

154 The AEC test, as applied in the present case, establishes the price at which a competitor as efficient as Intel would have had to offer its x86 CPUs in order to compensate an OEM for the loss of any exclusivity payment granted by Intel. In the AEC test, that price is called the “effective price” or “EP”.

155 In principle, the portion of the total rebates for which [a competitor as efficient as Intel] must offer compensation comprises only the amount of the rebates which is subject to the exclusive supply condition, excluding quantity rebates (“the conditional portion of the rebates”). As is apparent in particular from recital 1460 of the [decision at issue], in order to take into consideration only the conditional part of a payment, the AEC test refers, in the present case, to the average sales price …, namely the list price, after deduction of conditional discounts.

156 The smaller the contestable share and, therefore, the smaller the quantity of products for which the alternative supplier may compete, the greater the likelihood that the exclusivity payment will be capable of foreclosing [a competitor as efficient as the dominant undertaking]. If the loss of payments granted by Intel to its customer must be spread over a small quantity of products offered by the alternative supplier in the contestable share, that leads to an appreciable reduction in the effective price. It is therefore more probable that the effective price will be below Intel’s measure of viable cost.

157 The effective price must be compared with Intel’s measure of viable cost. Intel’s measure of viable cost adopted in the [decision at issue] is that of the average avoidable cost …

158 As is apparent in particular from recital 1006 of the [decision at issue], it may be concluded that a system of exclusivity payments is capable of foreclosing market access for equally efficient competitors [as Intel] if the effective price is below Intel’s [average avoidable cost]. In that situation, the AEC test has a negative result. If, by contrast, the effective price is above [Intel’s average avoidable cost], [a competitor as efficient as Intel] is deemed to be able to cover its costs and to therefore be in a position to enter the market. In that situation, the AEC test has a positive result.

159 It is in the light of those general considerations that the [General] Court must examine the substance of [Intel’s] arguments that the AEC [test] is vitiated by numerous errors.’

44 It is apparent, in particular, from paragraphs 175, 258, 260, 283, 285, 286, 297 to 299 and 334 of the judgment under appeal that, according to the methodology adopted by the Commission, the result – positive or negative – of the AEC test, within the meaning set out in paragraph 158 of that judgment, is ultimately determined by means of a comparison of the contestable share and the required share, the latter being the proportion of the customer’s requirements that a competitor as efficient as Intel must obtain in order for it to be able to enter the market without incurring losses. If the contestable share is higher than the required share, the AEC test result is positive for Intel, while the opposite situation leads to a negative result and means that the contested rebates are capable of foreclosing a competitor as efficient as Intel.

45 The grounds set out in paragraphs 160 to 166 of the judgment under appeal relate to the burden of proof and the standard of proof required.

46 In the grounds set out in paragraphs 167 to 482 of that judgment, the General Court examined the merits of Intel’s arguments that the decision at issue was vitiated by numerous errors as regards the AEC test.

47 Those grounds are divided into five parts, each of which deals with the arguments which Intel raised concerning the AEC test contained in the decision at issue as regards the four OEMs concerned, namely Dell, HP, NEC and Lenovo, on the one hand, and the retailer MSH, on the other.

48 The General Court’s analysis of Dell’s situation is set out in paragraphs 202 to 282 of the judgment under appeal. Following that analysis, the General Court held, first, in paragraph 283 of that judgment, that the Commission had not established to the requisite legal standard that its assumption that Dell’s contestable share for the period in question was 7.1% was well founded. Since that assumption formed the basis for demonstrating, by comparing the contestable share and the required share, the capability of Intel’s rebates to Dell to have a foreclosure effect, the comparison in question did not, according to the General Court, demonstrate that capability to the requisite legal standard.

49 Second, in paragraph 284 of the judgment under appeal, the General Court held that the factors which, according to the Commission, reinforced the estimated foreclosure capability of the contested rebates, those factors being, in essence, that any loss of Intel’s rebates would result in increased rebates from Intel to Dell’s OEM competitors and that the estimate of the contestable share did not take account of the fact that Dell also purchased from Intel products other than x86 CPUs, in particular chipsets, had not been analysed sufficiently in the decision at issue. Moreover, an alternative calculation method used by the Commission in that decision did not demonstrate that the contested rebates were capable of having an anticompetitive foreclosure effect throughout the relevant period.

50 In paragraph 285 of the judgment under appeal, the General Court recalled that, according to recital 1281 of the decision at issue, the conclusions which the Commission had reached as regards the capability of the contested rebates granted to Dell to have an anticompetitive foreclosure effect were inferred from the comparison of the contestable share and the required share, the reinforcing factors and that alternative calculation method.

51 In those circumstances, the General Court held, in paragraph 286 of the judgment under appeal, that the Commission had not been in a position to establish, on the basis of the comparison of the contestable share and the required share and the reinforcing factors, the capability of the contested rebates granted to Dell to have such a foreclosure effect. In addition, the third element relied on by the Commission, namely an alternative calculation method, which, according to the wording of recital 1281 of the decision at issue, served to confirm those first two elements, could not, according to the General Court, in itself substantiate the Commission’s findings, a fortiori since it did not demonstrate that the contested rebates were capable of having an anticompetitive foreclosure effect throughout the relevant period.

52 Accordingly, in paragraph 287 of the judgment under appeal, the General Court upheld Intel’s complaint that the Commission had not established to the requisite legal standard the validity of the conclusion set out in recital 1281 of the decision at issue that, over the period between December 2002 and December 2005, the contested rebates were capable of having or were likely to have an anticompetitive foreclosure effect on the ground that even a competitor as efficient as Intel would have been prevented from supplying Dell for its x86 CPU requirements.

53 The General Court’s analysis of HP’s situation is set out in paragraphs 288 to 335 of the judgment under appeal. After examining a series of factors relating to the volume of purchases, the amount of the rebates and the average sales price of the x86 CPU, the General Court held, in paragraph 319 of that judgment, that the Commission had erred in finding that its calculation of the required share allowed it to draw conclusions concerning the anticompetitive foreclosure effect of the rebates which Intel granted to HP for the entire period between November 2002 and May 2005. In particular, the Commission had not demonstrated that that effect was present for the period between November 2002 and September 2003.

54 In addition, in paragraph 320 of the judgment under appeal, the General Court held that the fact that the Commission had carried out an alternative calculation of the required share in recital 1389 of the decision at issue, using the figures set out in recital 1338 of that decision, could not make up for that error, since that alternative calculation also did not cover the whole of the period between November 2002 and May 2005.

55 Moreover, in paragraphs 328 to 331 of the judgment under appeal, the General Court held, with regard to the reinforcing factor consisting of a transfer of the contested rebates initially granted by Intel to HP to one of HP’s competitors, that the Commission should have specified which of the factors taken into account in the AEC test would have been influenced and how. Therefore, the decision at issue was vitiated by a failure to state reasons.

56 In those circumstances, the General Court concluded, in paragraph 335 of the judgment under appeal, that the Commission had not established to the requisite legal standard the validity of the conclusion set out in recital 1406 of the decision at issue that, during the period from November 2002 to May 2005, the rebates granted by Intel to HP were capable of having an anticompetitive foreclosure effect or were likely to have such an effect, since it had not demonstrated that there were such foreclosure effects in the period between 1 November 2002 and 30 September 2003.

57 The General Court’s analysis of NEC’s situation is set out in paragraphs 336 to 411 of the judgment under appeal. In paragraph 411 of the judgment under appeal, the General Court held that, by attributing an exaggerated value to the conditional portion of the rebates and by extrapolating the results which it had obtained for the fourth quarter of 2002 to the entire period of the infringement, the Commission had made two errors of assessment. The General Court inferred from those findings that the basic parameters of the Commission’s AEC test were incorrect. Accordingly, the Commission had not established to the requisite legal standard the validity of the conclusion set out in recital 1456 of the decision at issue, according to which the payments granted by Intel to NEC were capable of foreclosing or likely to foreclose a competitor as efficient as Intel.

58 The General Court’s analysis of Lenovo’s situation is set out in paragraphs 412 to 457 of the judgment under appeal. In paragraph 439 of that judgment, the General Court held that the calculations made by the Commission in order to establish the conditional portion of the rebates and, consequently, the required share that a competitor as efficient as Intel had to obtain in order for it to be able to enter the market without incurring losses were based on a premiss that was contrary to the foundation of the AEC test set out in recitals 1003 and 1004 of the decision at issue. In particular, that foundation includes the principle that the hypothetical competitor is as efficient as Intel. In its analysis relating to the assessment of the amount of the contested rebates granted by Intel to Lenovo in the form of two non-cash advantages, namely the extension of Intel’s standard one-year warranty and the offer of improved use of a supply hub in China, the Commission, according to the General Court, in fact conducted its reasoning by reference to a less efficient competitor, which is not, however, the relevant economic operator for assessing whether the contested rebates are capable of having an anticompetitive foreclosure effect.

59 After finding, in paragraphs 440 to 456 of the judgment under appeal, that that finding could not be invalidated by the Commission’s arguments, the General Court held, in paragraph 457 of that judgment, that the Commission had not established to the requisite legal standard the validity of the conclusion set out in recital 1507 of the decision at issue that in 2007 the contested rebates granted by Intel were capable of having or likely to have such foreclosure effects to the detriment of a competitor as efficient as Intel.

60 The General Court’s analysis of MSH’s situation is set out in paragraphs 458 to 481 of the judgment under appeal. In paragraphs 466 to 480 of that judgment, the General Court focused on the ‘double conditional rebate’ method. That method is based on the idea that, in order to be able to sell computers of a particular brand to MSH, a competitor as efficient as Intel would have had to ensure not only that MSH was ready to purchase computers based on that competitor’s CPUs, but also, and most importantly, that OEMs were ready to manufacture those computers. That competitor would therefore have had to grant two payments: one to obtain the contestable share of the OEM and the other to obtain the contestable share of MSH. Intel’s practices at different levels of the supply chain could therefore have had a cumulative effect. Without calling into question the principle of that reasoning, the General Court found that its implementation, based on the assumption that each OEM which supplied MSH benefited from a conditional rebate equivalent to the total contested rebate offered to NEC in the fourth quarter of 2002, was vitiated by two flaws, each of which was capable of invalidating the results of the AEC test for MSH. First, the Commission did not claim, nor did it demonstrate, that in the segment of computers for private individuals Intel granted conditional rebates to any of the other OEMs from which MSH made its purchases, namely Fujitsu, Acer, HP, Compaq, Toshiba and Medion, on conditions comparable to the contested rebates relating to computers purchased from NEC. Second, and in any event, the Commission had not shown that the contested rebates granted to NEC for the fourth quarter of 2002, even if they were representative for all OEMs, had been stable over a ten-year period.

61 In those circumstances, the General Court held, in paragraph 482 of the judgment under appeal, that it had to accept Intel’s argument that the AEC test carried out by the Commission in the decision at issue with regard to Dell, HP, NEC, Lenovo and MSH was vitiated by errors.

62 Next, the General Court examined, in paragraphs 483 to 521 of the judgment under appeal, the arguments made by Intel and ACT to the effect that the Commission had not analysed properly, as criteria referred to in paragraph 139 of the judgment on the appeal, the share of the market covered by the contested rebates, on the one hand, and the duration and amount of those rebates, on the other.

63 As regards the share of the market covered by the contested rebates, in paragraph 485 of the judgment under appeal, the General Court identified Section 4.2.4 of the decision at issue, which contains recitals 1577 to 1596 of that decision, as being relevant in respect of the strategic importance of the OEMs which benefited from the contested rebates. According to the General Court, in recital 1577 of that decision, the Commission stated, in essence, that, because of their market share, their strong presence in the more profitable segment of the market in question and their ability to legitimise a new processor in that market, some OEMs, in the present case Dell and HP, were strategically more important than others in providing x86 CPU manufacturers with access to that market. The General Court also identified as being relevant recital 1597 of that decision, according to which the OEMs targeted by Intel’s conduct held a significant part of the market and were strategically more important than other OEMs, which had had a more significant impact on the overall market than would have corresponded to their aggregate market share alone. The Commission concluded that the market coverage of the contested rebates had to be regarded as significant.

64 In paragraph 493 of the judgment under appeal, the General Court held that it could not be ruled out that Section 4.2.4 of the decision at issue might have been relevant in examining the share of the market covered by the contested rebates, in that it deals with certain factors which are a priori relevant to examining the foreclosure capability of a rebate scheme. That being so, in paragraph 494 of that judgment, the General Court noted that Section 4.2.4 and recital 1597 of that decision, on which the Commission relied in order to find that the share of the market covered had been examined, could not be interpreted as constituting in themselves sufficient examination, in the circumstances of the present case, of the share of the market covered by the contested rebates, within the meaning of paragraph 139 of the judgment on the appeal.

65 In that regard, in paragraph 495 of the judgment under appeal, the General Court observed that, on the assumption that the Commission could legitimately rely on the market shares of certain OEMs rather than examining the share of the market covered by the contested rebates, the Commission had taken into account, in recitals 1578 to 1580 of the decision at issue, solely Dell’s and HP’s market shares, and excluded the other OEMs concerned by those rebates. In addition, the market shares thus taken into account by the Commission covered only the period from the first quarter of 2003 to the final quarter of 2005, that is to say, only part of the period from October 2002 to December 2007 covered by that decision, and therefore also disregarded the period from 2006 to 2007, during which Lenovo and MSH were concerned. Last, the figures for market shares relied on by the Commission take account of Dell’s and HP’s worldwide market shares in all segments, despite the fact that the sole contested rebates as regards HP related to corporate desktops, as stated in Article 1(b) of that decision.

66 In those circumstances, the General Court noted, in paragraphs 499 and 500 of the judgment under appeal, that the Commission had failed to determine the share of the market covered by the contested rebates, contrary to the requirement placed on it pursuant to paragraph 139 of the judgment on the appeal, and concluded that it had not considered properly the criterion relating to that share of the market covered.

67 As regards the duration of the contested rebates, the General Court identified, in paragraphs 508 to 515 of the judgment under appeal, two groups of recitals of the decision at issue in which the Commission had examined factors relating to that duration. Those are, first, recitals 1013 to 1035 of that decision and, second, recitals 201, 202, 965 to 968 and 1227 thereof.

68 In that regard, the General Court held, first, that the purpose of recitals 1013 to 1035 of the decision at issue was solely to define the time horizon, in the present case one year, on which the OEMs based their decisions regarding their supply requirements for x86 CPUs as an assumption underlying the calculation of the contestable share of the contested rebates granted by Intel to each of the OEMs concerned.

69 Therefore, according to the General Court, the Commission used such a time horizon to determine the methodology for calculating an OEM’s contestable share, which then had to be compared with other factors in the AEC test in order to assess whether the contested rebates could have an anticompetitive foreclosure effect. Such an examination does not therefore constitute, according to the General Court, an analysis of the duration of those rebates as a factor which is capable, in itself, of demonstrating their ability to have such a foreclosure effect.

70 Second, it is true that, in recitals 201, 202, 965 to 968 and 1227 of the decision at issue, the Commission examined the duration and form of the OEMs’ commitments to Intel giving them a right to rebates as factors capable of encouraging or hindering a new competitor’s entry into the market, having regard in particular to the temporal scope of those commitments or Intel’s ability to pay or adjust its rebates within a short time frame.

71 However, according to the General Court, although those aspects of the time horizon appeared to it to be relevant, the Commission examined them only in a haphazard and limited manner. It did not carry out an examination of those aspects for all OEMs as a relevant factor for determining the capability of the contested rebates to have an anticompetitive foreclosure effect.

72 In addition, in paragraphs 516 to 518 of the judgment under appeal, the General Court rejected an argument put forward by the Commission to the effect that, even if the AEC test did not demonstrate the capability of the contested rebates to have a foreclosure effect, it was the total period during which Intel applied the rebates and exclusivity payments to OEMs which had to be considered and that, in so far as the rebates lasted one year for Lenovo and several years for the other OEMs and for MSH, the conclusion had to be drawn that a competitor of Intel in the x86 CPU market would have had to accept a drop in profitability and a much lower level of profitability on those sales as compared with Intel. According to the Commission, those considerations were apparent from paragraphs 93 and 195 of the initial judgment and were therefore definitive.

73 In that regard, the General Court noted, first, that the initial judgment had been set aside in its entirety by the Court of Justice, so that it was required, following the referral of the case back to it, to carry out a fresh examination of the parties’ arguments concerning the duration of the rebates, without being bound by paragraphs 93 and 195 of that judgment.

74 Second, the General Court considered that, in the light of all of the criteria referred to in paragraph 139 of the judgment on the appeal, which must be examined in the case where the undertaking concerned submits, on the basis of supporting evidence, that its conduct was not capable of restricting competition, the sole reference, notwithstanding the conclusions that may be drawn from the AEC test, to the period during which the contested rebates were granted by Intel to the OEMs and MSH was not sufficient, in itself, to justify definitive assessments as to the anticompetitive foreclosure effects thereby produced.

75 For the same reasons, the General Court rejected, in paragraph 519 of the judgment under appeal, the Commission’s argument that the duration of the contested rebates could not be isolated from their timing, since they had been necessary to overcome Intel’s inability to produce a timely technical response to AMD’s 64-bit x86 CPUs.

76 On the basis of those considerations, the General Court held, in paragraph 520 of the judgment under appeal, that the Commission had erred in failing to examine, in the decision at issue, the duration of the rebates as evidence making it possible to determine the capability of the contested rebates to have an anticompetitive foreclosure effect.

77 Thus, in paragraph 521 of the judgment under appeal, the General Court held that Intel was justified in maintaining that the analysis in the decision at issue of the criteria mentioned in paragraph 139 of the judgment on the appeal was vitiated by a number of errors, in that the Commission had not considered properly, in that decision, the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates.

78 In the concluding paragraphs of the judgment under appeal, the General Court summarised the grounds underlying the operative part of that judgment.

79 In that regard, in paragraph 523 of the judgment under appeal, the General Court recalled that Intel had submitted, during the administrative procedure, on the basis of supporting evidence, that the contested rebates had not been capable of producing the alleged foreclosure effects. In recitals 1002 to 1573 of the decision at issue, the Commission carried out an AEC test and, in the light of the results of that test, concluded, in recitals 1574 and 1575 of that decision, that the contested rebates and the payments at issue made by Intel were capable of having or likely to have anticompetitive foreclosure effects, since even a competitor as efficient as Intel would have been prevented from supplying Dell, HP, NEC and Lenovo for their x86 CPU requirements or from selling, through the computers marketed by MSH, the x86 CPUs it produced.

80 In paragraph 524 of the judgment under appeal, the General Court stated that it followed, however, from the assessments relating to the pleas and arguments examined, first, that the AEC test carried out in the decision at issue was vitiated by errors and, second, as regards the criteria mentioned in paragraph 139 of the judgment on the appeal, that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates.

81 As regards, in particular, the rebates granted to HP, in paragraph 525 of the judgment under appeal, the General Court recalled that the Commission had not established to the requisite legal standard its finding that, during the period from November 2002 to May 2005, the contested rebates granted to HP were capable of having or likely to have an anticompetitive foreclosure effect, since it had not demonstrated that those effects were present for the period between 1 November 2002 and 30 September 2003. Even if it were necessary to infer that the AEC test could be regarded as conclusive for part of the period from November 2002 to May 2005, the anticompetitive foreclosure effect of those rebates could not, according to the General Court, be demonstrated to the requisite legal standard, since the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates.

82 Thus, in paragraphs 526 and 527 of the judgment under appeal, the General Court held that the Commission had not been in a position to determine that the contested rebates and the payments at issue made by Intel were capable of having or likely to have anticompetitive foreclosure effects and that they therefore constituted an infringement of Article 102 TFEU, so that the grounds of the decision at issue were not capable of serving as a basis for Article 1(a) to (e) of that decision.

83 In those circumstances, the General Court annulled, first, Article 1(a) to (e) and Article 2 of the decision at issue and, second, Article 3 of that decision solely with regard to the conduct described in Article 1(a) to (e) thereof. By contrast, having accepted the findings in the initial judgment concerning the naked restrictions and their unlawfulness under Article 102 TFEU, the General Court dismissed the action as to the remainder.

Procedure before the Court of Justice and forms of order sought

84 By decision of the President of the Court of Justice of 5 August 2022, the Federal Republic of Germany was granted leave to intervene in the present proceedings in support of the form of order sought by the Commission.

85 The Commission claims that the Court of Justice should:

– set aside the judgment under appeal, with the exception of paragraph 3 of the operative part, pursuant to which the General Court dismissed the action before it as to the remainder;

– refer the case back to the General Court, and

– reserve the costs.

86 Intel and ACT contend that the Court of Justice should:

– dismiss the appeal, and

– order the Commission to pay the costs relating to the appeal and the Federal Republic of Germany to pay the costs relating to its intervention.

87 The Federal Republic of Germany claims that the Court of Justice should:

– set aside the judgment under appeal, save for paragraph 3 of its operative part;

– refer the case back to the General Court, and

– reserve the costs.

The appeal

88 The Commission puts forward six grounds of appeal:

– the first alleges infringement of the principle of ne ultra petita, together with a misinterpretation and misapplication of the criteria established by the Court of Justice in the judgment on the appeal, as regards the capability of the contested rebates to foreclose competition;

– the second alleges infringement of the Commission’s rights of defence in the examination of the AEC test;

– the third alleges an error of law as regards the evidence required, infringement of the rights of the defence and distortion of the evidence in the examination of the AEC test in relation to Dell;

– the fourth alleges several errors of law and infringement of the Commission’s rights of defence in the examination of the AEC test in relation to HP;

– the fifth alleges misinterpretation of the AEC test carried out by the Commission and of Article 102 TFEU, distortion of the evidence and infringement of the Commission’s rights of defence in the examination of the AEC test in relation to Lenovo, and

– the sixth alleges an incorrect assessment of the consequences to be drawn from the errors found in the AEC test.

89 Before setting out its response to each of the grounds of appeal, Intel, supported by ACT, observes that those grounds of appeal are ineffective on the ground that the Commission did not call into question the General Court’s analysis in paragraphs 144 to 149 of the judgment under appeal. According to that analysis, the characterisation of the contested rebates as abusive by their very nature and irrespective of whether they are capable of restricting competition constitutes an error of law. According to Intel, the Commission thus acknowledged that the AEC tests was not part of the reasoning underlying the operative part of the decision at issue. In those circumstances, the finding of that error constitutes a sufficient basis, in itself, for the operative part of the judgment under appeal, and any alleged error made by the General Court in its assessments following that finding is irrelevant.

90 That argument cannot be accepted. As the General Court observed, in essence, in paragraph 149 of the judgment under appeal, the error relating to the characterisation of the contested rebates as, by their very nature, constituting an infringement of Article 102 TFEU was not sufficient to annul the decision at issue, since the Commission had carried out an AEC test that played an important role in the assessment of whether those rebates were capable of foreclosing a competitor as efficient as Intel. Consequently, the General Court’s annulment of that decision is based, as is also apparent from paragraph 524 of the judgment under appeal, on the errors found in the AEC test, on the failure to consider the share of the market covered by the contested rebates and on the incorrectness of the Commission’s analysis of the duration of those rebates. It follows that the grounds raised in support of the present appeal are capable of calling into question the General Court’s findings underlying the operative part of the judgment under appeal and of leading, if they are well founded, to that judgment being set aside. Consequently, Intel’s objection that those grounds of appeal are ineffective must be rejected and, accordingly, those grounds of appeal must be examined.

The first ground of appeal, alleging that the General Court ruled ultra petita and that it misinterpreted and misapplied the capability of the contested rebates to foreclose competition

91 The first ground of appeal consists of two parts, alleging, respectively, that the General Court ruled ultra petita and, in the alternative, that it misinterpreted and misapplied the criteria for recognising the capability of the contested rebates to foreclose competition.

The first part, alleging that the General Court ruled ultra petita

–  Arguments of the parties

92 The Commission claims that Intel had not raised, in its application at first instance, any argument to the effect that the scope of the examination, in the decision at issue, of the criteria relating to the share of the market covered and to the duration of the contested rebates was insufficient. Intel argued to that effect only in the main observations on the referral of the case back to the General Court.

93 First, according to the Commission, that application contains no reference to the recitals of the decision at issue relating to the strategic significance of the OEMs that benefited from the contested rebates. In its appeal against the initial judgment, Intel challenged only the General Court’s assessment of the significance of the share of the market covered by those rebates. Intel’s argument in the application concerned the allegedly very limited share of the market actually affected by the anticompetitive foreclosure effect stemming from those rebates, not whether the Commission had adequately analysed the share of the market covered by those rebates.

94 Second, and similarly, Intel did not call into question the scope of the Commission’s analysis of the duration of the contested rebates, but merely challenged the validity of the Commission’s approach of cumulating the short-term agreements concluded with the OEMs and MSH, which is a subsequent issue. Thus, in support of its appeal against the initial judgment, Intel merely called into question the cumulation of those agreements, relied on by the General Court in order to substantiate the capability of the contested rebates to foreclose competition. Contrary to what the General Court states in paragraph 506 of the judgment under appeal, the complaints relating to the duration of the contested rebates are not related to those put forward in paragraphs 102 and 111 to 114 of the application.

95 Intel, supported by ACT, disputes the merits of this part of the first ground of appeal.

–  Findings of the Court

96 It follows from the rules governing the procedure before the Courts of the European Union, in particular Article 21 of the Statute of the Court of Justice of the European Union and Article 76 and Article 84(1) of the Rules of Procedure of the General Court, that the dispute is in principle determined and circumscribed by the parties and that the Courts of the European Union may not rule ultra petita (judgment of 13 July 2023, Commission v CK Telecoms UK Investments, C‑376/20 P, EU:C:2023:561, paragraph 324), and that the annulment granted may not go beyond that sought by the applicant (see, to that effect, judgment of 14 November 2017, British Airways v Commission, C‑122/16 P, EU:C:2017:861, paragraph 81 and the case-law cited).

97 In the present case, as stated in paragraph 83 of the present judgment, the General Court annulled, first, Article 1(a) to (e) and Article 2 of the decision at issue and, second, Article 3 of that decision solely with regard to the conduct described in Article 1(a) to (e) thereof. Furthermore, pursuant to paragraph 3 of the operative part of the judgment under appeal, the General Court dismissed the action as to the remainder, namely as regards the naked restrictions and their unlawfulness under Article 102 TFEU.

98 It should be noted in that regard that, in paragraph 674 of the application, Intel sought, inter alia, the annulment of the decision at issue in its entirety, so that the annulment granted by the General Court falls short of that sought by Intel. It follows that, in so far as the Commission formally objects that the General Court ruled ultra petita, that complaint must be rejected.

99 The Commission also objects that the General Court reviewed, in paragraphs 485 to 500 of the judgment under appeal, the lawfulness of the assessments made by the Commission in the decision at issue concerning the share of the market covered by the contested rebates in the light of a plea raised for the first time by Intel in the main observations on the referral of the case back to the General Court. In so doing, according to the Commission, the General Court exceeded the scope of the dispute as determined and circumscribed by the application and, as the case may be, by the reply, thereby infringing Article 84 of its Rules of Procedure.

100 In that regard, it must be borne in mind that, according to Article 84(1) of the Rules of Procedure of the General Court, no new plea in law may be introduced in the course of proceedings unless it is based on matters of law or of fact which come to light in the course of the procedure.

101 It follows that, as the General Court correctly pointed out in paragraph 106 of the judgment under appeal, following the referral judgment of the Court of Justice, the parties are not entitled, in principle, to rely on pleas which were not raised in the procedure which gave rise to the judgment of the General Court set aside by the Court of Justice (judgment of 1 July 2008, Chronopost and La Poste v UFEX and Others, C‑341/06 P and C‑342/06 P, EU:C:2008:375, paragraph 71). Only a plea which may be regarded as amplifying a plea put forward previously, whether directly or by implication, in the original application, and which is closely connected therewith must be declared admissible (judgment of 11 March 2020, Commission v Gmina Miasto Gdynia and Port Lotniczy Gdynia Kosakowo, C‑56/18 P, EU:C:2020:192, paragraph 66).

102 In the present case, after stating, in paragraph 12 of the application, that, contrary to Article 102 TFEU, the Commission had failed to assess the capability of the contested rebates to restrict competition in the light of all the circumstances of the case, Intel stated, in paragraphs 115 to 117 of the application, that the Commission should have taken into consideration the negligible share of the x86 CPU market concerned by those rebates, which was between 0.3% and 2% per year. In those paragraphs of the application, Intel also criticised the Commission’s approach of assessing the anticompetitive foreclosure effects of those rebates in isolation by segment of the market, whereas AMD could compete in segments not affected by the contested rebates, and emphasised that the Commission had not found that rebates had been granted to several other leading OEMs. Intel further developed those complaints in paragraphs 12, 37, 38, 52, 155, 185, 208 and 257 of the reply, the latter four paragraphs concerning, respectively, the contested rebates granted to HP, Lenovo, NEC and MSH. To document its calculations, Intel repeatedly referred to the report of Professor Salop and Dr Hayes of 22 July 2009 (‘the Salop-Hayes report’).

103 It follows that, by those complaints, Intel argued, in essence, before the General Court that the share of the market covered by the contested rebates was so low that any anticompetitive foreclosure capability could be ruled out. Accordingly, in assessing, in paragraphs 492 to 500 of the judgment under appeal, the lawfulness of the assessments made in that regard by the Commission in the decision at issue, the General Court did not go beyond the scope of the dispute as determined and circumscribed by Intel in its action at first instance.

104 The Commission also submits that the examination, in the judgment under appeal, of the analysis of the duration of the contested rebates carried out in the decision at issue does not correspond to any plea raised in the application, but corresponds only to a new plea raised in Intel’s main and supplementary observations on the referral of the case back to the General Court, which is therefore inadmissible.

105 In that regard, in paragraphs 101 and 102 of the application, Intel emphasised the importance of the duration of a practice when assessing whether it was capable of restricting competition. In that context, in paragraphs 111 and 112 of the application, relating to the anticompetitive foreclosure capability of the contested rebates, Intel highlighted the fact that the contested rebates were granted only for a few months and that OEMs could terminate some of the agreements providing for those rebates with 30 days’ notice, so that such a foreclosure effect could not be assumed. Intel reiterated that argument in paragraph 39 of the reply.

106 It follows that, by those complaints, Intel claimed, in essence, that the duration of the contested rebates was such that it could not be assumed that they were capable of having an anticompetitive foreclosure effect. The General Court was therefore fully entitled to hold, in paragraph 506 of the judgment under appeal, rejecting the plea of inadmissibility raised in that regard by the Commission, that Intel’s arguments put forward in its main and supplementary observations on the referral of the case back to the General Court regarding the duration of the contested rebates clearly related to those put forward in its application and that they were therefore admissible.

107 It follows that the first part of the first ground of appeal must be rejected.

The second part, alleging misinterpretation and misapplication of the criteria relating to the capability of the contested rebates to foreclose competition

–  Arguments of the parties

108 In the alternative, the Commission submits that the General Court confined its examination of the criteria established in paragraph 139 of the judgment on the appeal to the sufficiency of the analysis of the share of the market covered by the contested rebates and of their duration, without taking into account any of the other relevant criteria for assessing the capability of those rebates to foreclose competition. That approach is contrary to the requirement of an overall assessment of all the relevant circumstances of the case.

109 In that regard, according to the Commission, although they are not exhaustive, the criteria established by the Court of Justice in paragraph 139 of the judgment on the appeal form part of the relevant circumstances for the purpose of assessing the capability of a rebate scheme to foreclose competition. The criteria in question are therefore not cumulative, in the sense that the Commission is not required to analyse each of them separately or to give each one the same weight. It follows that the capability of a rebate scheme to foreclose competition cannot be invalidated on the basis of one or two circumstances examined in isolation. However, in paragraphs 519 to 521 and 525 to 527 of the judgment under appeal, the General Court held that the fact that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of the contested rebates was sufficient to lead to the annulment of the decision at issue. The extent to which the General Court based its assessment on the errors found in the AEC test is not clear from the judgment under appeal. Moreover, in paragraph 525 of the judgment under appeal, the General Court held that anticompetitive foreclosure capability could not be considered to be demonstrated if the Commission had not considered properly the criteria relating to the share of the market covered and to the duration of the rebates.

110 The General Court therefore appears to have adopted a formalistic approach, according to which the Commission should have examined all the criteria established in paragraph 139 of the judgment on the appeal and given each of them the same weight, irrespective of the relative importance of each criterion in the specific context of the case before it and of an overall assessment of all the relevant circumstances that may demonstrate the capability of the contested rebates to foreclose competition.

111 According to the Commission, the General Court could not reach a conclusion on the capability of the conduct complained of to foreclose competition without examining (i) the extent of Intel’s dominant position on the relevant market; (ii) the conditions and arrangements for granting the contested rebates and their influence on the OEMs’ and MSH’s procurement decisions; (iii) the amount of those rebates; (iv) their timing and (v) the existence of an anticompetitive strategy.

112 In particular, according to the Commission, as regards the first two criteria mentioned in the preceding paragraph, Intel did not dispute the fact that the exclusive supply condition, on which the contested rebates depended, influenced the decisions of the OEMs and of MSH regarding their sources of supply. Similarly, in view of the duopoly characterising the x86 CPU market, the existence of a strategy aiming to exclude AMD from the market entails a risk of all competition being eliminated on that market. However, the General Court failed to take into account the existence of such a strategy not only as a relevant criterion in itself, but also as part of the AEC test. In addition, according to the Commission, the timing of the contested rebates is of particular importance, since it shows that they were implemented to overcome Intel’s inability to produce a timely technical response to AMD’s 64-bit x86 CPUs.

113 The Commission submits that the General Court therefore annulled the decision at issue on the sole ground that the Commission had insufficiently analysed two of the criteria established in paragraph 139 of the judgment on the appeal. That fact is not sufficient to annul a measure that is based on a series of complex assessments, if other aspects of those assessments, which have not been challenged or have been validated, are sufficient to support the conclusion reached by that measure. Thus, the General Court should have examined whether, despite the shortcomings identified in the scope of the examination of the two criteria in question carried out in the decision at issue, that decision contained an analysis with regard to the other criteria justifying the finding that the contested rebates were capable of foreclosing competition.

114 The Federal Republic of Germany submits that the General Court should have first examined whether, during the administrative procedure, Intel had raised sufficiently specific arguments concerning the share of the market covered by the contested rebates and the duration of those rebates to rebut the presumption relating to the capability of those rebates to foreclose competition. Only in that situation would the Commission have been obliged, based on the right to be heard, expressly to address those two criteria in the decision at issue. In addition, the General Court should have addressed whether a different analysis or weighing up of those two criteria by the Commission would have led to the conclusion that there was no possibility of anticompetitive foreclosure. Thus, the General Court should have assessed the significance of those criteria in the light of the existence of an overall strategy implemented by Intel in order to foreclose AMD, its only competitor on the x86 CPU market. The existence of an intention to exclude competitors may offset possible doubts as to the share of the market covered by the contested rebates and their duration.

115 Intel and ACT dispute the merits of this part of the first ground of appeal.

–  Findings of the Court

116 It should be recalled that a party who, pursuant to Article 40 of the Statute of the Court of Justice of the European Union, is granted leave to intervene in a case submitted to the Court of Justice may not alter the subject matter of the dispute as defined by the forms of order sought and the grounds of appeal raised by the main parties. It follows that only arguments submitted by an intervener which fall within the framework provided by those forms of order sought and grounds of appeal are admissible, since the intervener may not rely on new grounds of appeal, distinct from those relied on by the appellant (see, to that effect, judgment of 10 November 2016, DTS Distribuidora de Televisión Digital v Commission, C‑449/14 P, EU:C:2016:848, paragraphs 114 and 121).

117 The complaint raised by the Federal Republic of Germany, alleging that the General Court should have first examined whether, during the administrative procedure, Intel had raised specific arguments concerning the share of the market covered by the contested rebates and the duration of those rebates, must be rejected as inadmissible. That complaint does not relate to any ground raised by the Commission in support of the appeal. Moreover, it in fact contradicts the assertion made by the Commission in the second ground of appeal, according to which Intel submitted, during the administrative procedure, a vast volume of evidence concerning the capability of the contested rebates to produce the alleged anticompetitive foreclosure effects, which, according to paragraphs 138 and 139 of the judgment on the appeal, means that the Commission was required to analyse, inter alia, the share of the market covered by the contested rebates and their duration.

118 As regards the complaints on which the Commission relies in the second part of the first ground of appeal, it should be noted that, in its appeal, the Commission states that that part is based on the premiss that the General Court annulled the decision at issue on the sole basis that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates. The Commission also states that, if the Court of Justice considers that premiss to be incorrect, it should proceed to examine the other grounds of appeal.

119 In order to justify that premiss, the Commission relies on paragraph 525 of the judgment under appeal, the content of which is reproduced in paragraph 81 of the present judgment.

120 The Commission infers from that paragraph 525, which concerns in particular the contested rebates granted by Intel to HP, that, despite the express reference, made in paragraph 524 of the judgment under appeal, to three grounds underpinning the annulment of the decision at issue, namely, first, the errors vitiating the AEC test with regard to the OEMs and MSH, second, the inadequate examination of the share of the market covered by the contested rebates and, third, the incorrect analysis of the duration of those rebates, the General Court in fact based that annulment solely on the assessments relating to the last two grounds. In adopting that approach, the General Court, according to the Commission, assessed the sufficiency of the Commission’s examination of those two criteria in an abstract and formalistic way, without taking account of the relative importance of each of them and without taking into consideration other circumstances analysed in the decision at issue.

121 In that regard, it should be recalled that, after finding, in paragraphs 145 and 147 of the judgment under appeal, that the approach adopted in the decision at issue, according to which the contested rebates were by their very nature abusive irrespective of whether they were capable of restricting competition by foreclosing an as-efficient competitor, so that an AEC test was not necessary for the purpose of assessing that capability, was vitiated by an error of law, the General Court found, in paragraph 149 of that judgment, that the Commission had nevertheless carried out such a test and that that test had played an important role in the assessment of that capability, so that it was necessary still to examine Intel’s arguments concerning that test. In line with that assessment, the General Court examined, in paragraphs 150 to 482 of that judgment, Intel’s arguments that the decision at issue should be annulled on the ground that the AEC test carried out in relation to Dell, HP, NEC, Lenovo and MSH was vitiated by several errors (see paragraphs 42 to 61 of the present judgment).

122 It was only after finding, in paragraph 482 of the judgment under appeal, that a large part of those arguments had to be upheld, that the General Court examined, in paragraphs 483 to 520 of that judgment, the arguments put forward by Intel and ACT to the effect that the Commission had not analysed properly, as criteria referred to in paragraph 139 of the judgment on the appeal, the share of the market covered by the contested rebates, on the one hand, and the duration and amount of those rebates, on the other (see paragraphs 62 to 77 of the present judgment).

123 At the end of that examination, the General Court held, in paragraph 521 of the judgment under appeal, that Intel was justified in maintaining that the analysis in the decision at issue of the criteria mentioned in paragraph 139 of the judgment on the appeal was vitiated by a number of errors, in that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates.

124 Thus, in paragraphs 523 and 524 of the judgment under appeal, the General Court stated that, although the Commission had found, on the basis of an AEC test carried out in recitals 1002 to 1575 of the decision at issue, that the contested rebates were capable of foreclosing a competitor as efficient as Intel and thereby restricting competition, the fact remained, first, that that test was vitiated by errors, second, that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and, third, that it had not analysed correctly the duration of those rebates.

125 In that context, the General Court recalled, in paragraph 525 of the judgment under appeal, that the error found in the AEC test with regard, in particular, to the contested rebates granted to HP related to the period from 1 November 2002 to 30 September 2003, whereas the infringement period relating to that OEM was from November 2002 to May 2005. However, in view of the fact that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates, the General Court held that the capability of the practice implemented with regard to HP to create an anticompetitive foreclosure effect could not be established to the requisite legal standard for the entire infringement period, even if it were necessary to consider that the AEC test was conclusive for part of that period.

126 It follows from the foregoing that the Commission has misread paragraph 525 of the judgment under appeal.

127 First, that paragraph 525, in so far as it relates solely to the rebates granted to HP, cannot call into question the fact that, as is apparent from paragraph 524 of the judgment under appeal, the capability of the contested rebates to foreclose a competitor as efficient as Intel from the market made up of Dell, Lenovo, NEC and MSH is based on the examination of several criteria, namely the AEC test, the share of the market covered and the duration of those rebates. Second, in so far as that capability should be assessed solely in the light of the contested rebates granted to HP, it is apparent from paragraph 525 of that judgment that the share of the market covered by those rebates and their duration are of even more fundamental importance for the finding of such capability and therefore of an infringement for the entire infringement period. The Commission does not claim that the decision at issue contains an analysis capable of demonstrating the existence of anticompetitive foreclosure capability solely on the basis of the rebates granted to HP, for the period from October 2003 to May 2005. Moreover, it is not apparent from paragraphs 483 to 521 of that judgment that the decision at issue contains such an analysis.

128 It follows that paragraph 525 of the judgment under appeal cannot be understood, in isolation from its context, to mean that the General Court’s analysis relating to the AEC test, carried out in paragraphs 150 to 482 of that judgment, should be disregarded, even though it is expressly stated, in paragraph 524 of that judgment, that that analysis underpins the annulment of the decision at issue in the same way as the assessments of the share of the market covered by the contested rebates and the duration of those rebates. The premiss set out in paragraph 118 of the present judgment, on which the Commission based the second part of the first ground of appeal, is therefore inaccurate.

129 The inaccuracy of that premiss also means that the Commission cannot validly complain that the General Court annulled the decision at issue on purely formal grounds, according to which the Commission should have examined each of the criteria set out in paragraph 139 of the judgment on the appeal separately and with the same degree of detail, irrespective of the relative weight of each criterion in the light of the circumstances of the case.

130 In that regard, as is apparent from paragraphs 137 to 139 of the judgment on the appeal, the share of the market covered by the contested rebates and their duration are among the factors which the Commission must assess in order to establish that the undertaking concerned committed an abuse of a dominant position, and the fact that that undertaking submits, during the administrative procedure, on the basis of supporting evidence, that its conduct was not capable of producing an anticompetitive foreclosure effect means that the Commission is under a specific obligation to assess the possible existence of a strategy aiming to exclude competitors that are at least as efficient as that undertaking from the market.

131 In that context, first, as is apparent from paragraphs 485, 493 to 495, 509 and 510 of the judgment under appeal, and contrary to what the Commission alleges, the General Court sought to identify the recitals of the decision at issue that might be relevant to the share of the market covered by the contested rebates and the duration of those rebates.

132 Second, it is apparent from paragraphs 493 to 500 and 506 to 520 of the judgment under appeal that the General Court identified a series of shortcomings vitiating the relevant recitals of the decision at issue that led it to find that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates or the duration of those rebates as evidence making it possible to determine the capability of those rebates to have an anticompetitive foreclosure effect. The Commission does not call into question, in the present appeal, those findings of the General Court.

133 Moreover, the Commission’s complaint, shared by the Federal Republic of Germany, that the General Court could not annul the decision at issue without examining the extent of Intel’s dominant position on the relevant market, the conditions and arrangements for granting the contested rebates and the payments at issue and their influence on the OEMs’ and MSH’s procurement decisions, the amount of those rebates, their timing and the existence of an anticompetitive strategy cannot succeed either.

134 In that regard, first, in so far as it forms an integral part of the second part of the first ground of appeal, that complaint is also based on the incorrect premiss that the annulment of the decision at issue is based exclusively on the General Court’s findings relating to the share of the market covered by the contested rebates and the duration of those rebates.

135 Second, according to the judgment under appeal, the only analysis in the decision at issue seeking to demonstrate that the contested rebates constitute an abuse irrespective of their capability to foreclose a competitor as efficient as Intel is that set out, inter alia, in recitals 920 to 926, 950, 972, 981, 989, 1000 and 1001 of that decision, examined in paragraphs 133 to 144 of the judgment under appeal. As the General Court observed in paragraphs 145 to 147 of that judgment, it is apparent from paragraphs 137 to 139 and 141 of the judgment on the appeal that that analysis is vitiated by an error of law.

136 In so far as the Commission relies on Intel’s dominant position, on the conditional nature of the rebates and on the existence of a strategy aiming to exclude a competitor of Intel from the market, irrespective of whether that competitor is as efficient as Intel, the arguments thus relied on in support of that complaint are based, implicitly but necessarily, on the idea that the contested rebates are abusive per se.

137 In so far as the Commission relies on the amount of the contested rebates, the reinforcing factors for the foreclosure effects and the strategic nature of the OEMs benefiting from those rebates, it must be noted that, contrary to what the Commission claims, the General Court examined those factors in its assessments regarding the AEC test, the share of the market covered by the contested rebates or the duration of those rebates.

138 The Commission thus claims, in reality, that the General Court failed to examine whether the various chapters of the decision at issue contain material that makes it possible to construct a line of reasoning that demonstrates the capability of the contested rebates to have an anticompetitive foreclosure effect, despite the General Court’s findings concerning the share of the market covered by the contested rebates and their duration, even though the Commission does not dispute those findings in the present appeal. However, irrespective of the fact that, in themselves, the criteria relied on by the Commission do not appear to be sufficient to find an infringement of Article 102 TFEU, the General Court could not carry out such an examination, since, as it recalled, in essence, in paragraph 150 of the judgment under appeal, it cannot alter the constituent elements of the infringement found by the Commission by substituting its own reasoning for that of the author of the act the legality of which it is reviewing under Article 263 TFEU (see, to that effect, judgment of 16 June 2022, Sony Corporation and Sony Electronics v Commission, C‑697/19 P, EU:C:2022:478, paragraph 95 and the case-law cited).

139 It follows that the General Court did not disregard the criteria relating to the capability of the contested rebates to foreclose competition in identifying, in paragraphs 150 to 527 of the judgment under appeal, the errors which, in its view, rendered unlawful the grounds of the decision at issue with regard to the way in which the AEC test was conducted, the share of the market covered by the contested rebates and the duration of those rebates, so that those grounds were not, in accordance with its analysis, appropriate for the purpose of establishing an infringement of Article 102 TFEU and, therefore, of serving as a basis for Article 1(a) to (e) of that decision.

140 Consequently, the second part of the first ground of appeal must be rejected, so that the first ground of appeal must be rejected in its entirety.

The second ground of appeal, alleging infringement of the Commission’s rights of defence

Arguments of the parties

141 According to the Commission, since the contested rebates granted by Intel are presumed to be abusive, the Commission cannot be required to state reasons for the decision at issue in such a way as to respond not only to the vast volume of evidence submitted by Intel during the administrative procedure, but also to arguments that Intel did not raise during that procedure. During the proceedings before the General Court, Intel submitted additional arguments concerning the AEC test in the form of economic expert reports. Despite the fact that the General Court took account of that extensive new evidence, it refused, in paragraphs 235, 236, 252, 253, 316, 317, 443 and 444 of the judgment under appeal, to take account of the rebuttals of that evidence submitted by the Commission, on the ground that such an examination would lead it to substitute its own reasoning for that set out in the decision at issue. The purpose of those rebuttals is not to fill any gaps in the reasoning of that decision but to respond to the new economic analyses provided by Intel for the first time during the proceedings. Taking account of the Commission’s arguments would not therefore have led the General Court to substitute its own assessment for the analyses contained in the decision at issue.

142 In those circumstances, the General Court infringed the Commission’s rights of defence as regards (i) the calculation of Dell’s contestable share; (ii) the foreclosure capability of the contested rebates granted to HP between November 2002 and September 2003 and (iii) the value of the non-cash advantages granted to Lenovo.

143 Intel disputes the merits of the second ground of appeal.

Findings of the Court

144 It should be borne in mind, first, that, as is apparent from paragraphs 137 to 139 of the judgment on the appeal, the Commission is under an obligation to assess the possible existence of a strategy of the dominant undertaking aiming to exclude competitors that are at least as efficient as that undertaking from the market only if that undertaking submits, during the administrative procedure, on the basis of supporting evidence, that its conduct was not capable of restricting competition and, in particular, of producing the alleged foreclosure effects.

145 It follows that, where that capability is demonstrated by the Commission by means of an AEC test taking into account the evidence provided by the undertaking concerned, the General Court cannot conclude that that test is invalid on the basis of evidence that that undertaking produced for the first time before it for that purpose, when that evidence was not known to the Commission and could have been submitted during the administrative procedure.

146 Second, according to the methodology adopted by the Commission to assess the capability of the contested rebates to foreclose a competitor as efficient as Intel, the positive or negative result of the AEC test, within the meaning set out in paragraph 158 of the judgment under appeal, is ultimately determined by means of a comparison of the contestable share and the required share. The latter is the proportion of the customer’s requirements that a competitor as efficient as Intel must obtain in order for it to be able to enter the market without incurring losses. If the contestable share is higher than the required share, the AEC test result is positive for Intel, while the opposite situation leads to a negative result and to a finding that the contested rebates are capable of foreclosing a competitor as efficient as Intel.

147 In the first place, as regards Dell’s situation, in paragraphs 204 to 206, 215 and 220 of the judgment under appeal, the General Court presented the evidence relied on by Intel to dispute the calculation of the contestable share of 7.1% that the Commission had taken into account with regard to that undertaking in the decision at issue.

148 First, it presented an email dated 10 November 2005 and written by a Dell executive, referred to as ‘D 1’ (‘the D 1 email’), from which it is apparent that the switch of Dell’s requirements to AMD could concern up to 25% of the volume of x86 CPUs, that executive having, moreover, confirmed the content of that email under oath in the course of private litigation between Intel and AMD in the State of Delaware (United States of America). In paragraph 204 of the judgment under appeal, the General Court also states that, according to the Salop-Hayes report, produced by Intel for the first time before the General Court, the volume projection of 25% of Dell’s requirements translated to a contestable share of 17.5% for the first year, or 12.5% using the Commission’s approach, which Intel considered to be unreasonable.

149 Second, an email dated 9 March 2004 and written by another Dell executive, referred to as ‘D 5’, made reference to a possible switch of Dell supplies to AMD for 25% of the total volume of Dell’s x86 CPU requirements inside 90 days (‘the D 5 email’).

150 Third, a declaration made on 21 December 2007 by, as is apparent from paragraph 194 of the judgment under appeal, an Intel employee responsible for the relationship with Dell at the material time, referred to as ‘I1’, indicates that Intel internally estimated that Dell’s contestable share of x86 CPU requirements was between 15% and 25% in the first year (‘the I1 declaration’).

151 Fourth and last, a statement made by D 1 in the course of private litigation between Intel and AMD in the State of Delaware confirms the content of his email of 10 November 2005, while a statement made by another Dell executive, referred to as ‘D 3’, in the course of that litigation indicates that he had no reason to question the accuracy of D 1’s remarks.

152 It is apparent from paragraphs 174, 196, 209 to 211 and 222 of the judgment under appeal that, with the exception of the Salop-Hayes report, all the documents relied on by Intel formed part of the administrative file and that they were discussed in the decision at issue, which the Commission does not dispute. The Commission focuses its arguments on the General Court’s use of the Salop-Hayes report to cast doubt on the contestable share of 7.1% calculated by the Commission in that decision, and criticises the General Court’s refusal to take into account analyses that did not form part of the statement of reasons for that decision, which the Commission annexed to the defence and to the rejoinder lodged in the proceedings at first instance.

153 In that regard, it is apparent from paragraphs 171, 229, 233 and 234 of the judgment under appeal that, as regards the AEC test relating to Dell, the Commission used a contestable share of 7.1% on the basis of a spreadsheet dating from January 2004 (‘the 2004 spreadsheet’) that Dell had provided to it during the administrative procedure. In particular, the Commission inferred the contestable share of 7.1% on the basis of a switch of Dell’s requirements to AMD concerning 7% of the volume of x86 CPUs that, according to the 2004 spreadsheet, Dell intended to purchase from AMD for 2005.

154 In paragraph 175 of the judgment under appeal, the General Court referred to the comparison of the required share with the contestable share, as carried out by the Commission in the decision at issue. From that comparison, the Commission concluded that, for 9 of the 13 quarters comprising the infringement period concerning Dell, the required share was higher than the contestable share.

155 It is true that, as is apparent from paragraph 204 of the judgment under appeal, which sets out Intel’s arguments, the Salop-Hayes report made it possible to translate the volume projection of 25% of Dell’s requirements evidenced by the D 1 email and the D 5 email into a contestable share, establishing that share at 17.5% for the first year, a percentage that would be reduced to 12.5% if the Commission’s approach were to be used, which Intel rejects.

156 However, first, it is apparent from paragraphs 220 to 234 of the judgment under appeal, which set out the assessment of the evidence relied on by Intel before the General Court, that the General Court called into question the calculation of Dell’s contestable share in the decision at issue on the basis of the D 1 email and the D 5 email, the statements made by the Dell executives in private litigation between Intel and AMD in the State of Delaware, and the I1 declaration. By contrast, the General Court did not rely, for the purposes of its assessment, on the Salop-Hayes report, which, moreover, is not mentioned in those paragraphs.

157 Consequently, contrary to the premiss underpinning the Commission’s reasoning, the General Court’s finding is not based on that report, so that the Commission’s argument must, in any event, be rejected.

158 Second, it is true that, in paragraph 236 of the judgment under appeal, the General Court recalled that taking into account the additional analyses submitted by the Commission for the first time before it in order to substantiate the AEC test contained in the decision at issue would lead it to substitute its own reasoning for that contained in that decision, which it would not be permitted to do in the review of legality referred to in Article 263 TFEU. However, in paragraphs 237 to 239 of the judgment under appeal, the General Court nonetheless examined those matters, for the sake of completeness, and concluded that they did not call into question the consideration set out in paragraph 234 of that judgment, so that the complaint put forward by the Commission in support of that ground of appeal is, also for that reason, based on an incorrect premiss.

159 In addition, in paragraphs 240 to 255 of the judgment under appeal, the General Court examined the evidence relating to the actual switch of Dell’s requirements to AMD in 2006 and in 2007, taking the view that that evidence supported the existence of doubt as to the establishment of Dell’s contestable share at 7.1%. In that regard, in paragraph 243 of the judgment under appeal, the General Court stated that, according to recital 1245 of the decision at issue, on the basis of the observations drawn from the actual switch of part of Dell’s requirements to AMD, it was possible to calculate a contestable share of more than 7.1%, being between 8.2% and 10.1%.

160 Contrary to what is claimed by the Commission, the General Court’s findings relating to the actual switch of Dell’s requirements to AMD in 2006 and in 2007 are based exclusively on evidence that is apparent from the decision at issue. The General Court does not rely on any evidence submitted by Intel for the first time during the proceedings and the Commission does not indicate in which paragraph of the judgment under appeal the General Court relied on such evidence.

161 It follows that the Commission’s complaint that the General Court based its assessment relating to Dell’s contestable share on evidence produced for the first time before it without taking into account the analyses that the Commission submitted in order to refute the merits of that evidence is, in any event, based on incorrect premisses.

162 In the second place, with regard to HP, the General Court noted that the Commission had not demonstrated the existence of anticompetitive foreclosure effects between 1 November 2002 and 30 September 2003. The Commission submits that, if Intel was entitled to claim, for the first time before the General Court, that evidence relating to a certain period had not been included in the AEC test, the Commission should be allowed to supplement that test during the proceedings.

163 It must be pointed out, in that regard, that Intel’s line of argument before the General Court in relation to HP sought to challenge the legality of the AEC test contained in the decision at issue without relying on any evidence that had not been produced during the administrative procedure. It follows that the General Court did not, in any event, err in rejecting, in paragraphs 300 and 301 of the judgment under appeal, on the basis of the judgment of 1 July 2010, Knauf Gips v Commission (C‑407/08 P, EU:C:2010:389, paragraph 89), the Commission’s argument alleging that that line of argument was inadmissible on the ground that Intel did not challenge, during the administrative procedure, the periods used by the Commission for its calculations. Similarly, as noted in paragraph 138 of the present judgment, the General Court was correct in holding, in paragraph 317 of the judgment under appeal, that it could not take account of the additional calculations, which were not apparent from the decision at issue, produced by the Commission for the first time as an annex to the rejoinder to substantiate the AEC test contained in that decision without substituting its own reasoning for that underlying that decision.

164 In the third place, the General Court considered that, in its analysis relating to the assessment of the amount of the contested rebates granted to Lenovo in the form of two non-cash advantages, namely the extension of Intel’s standard one-year warranty and improved use of a supply hub in China, on the basis of the value of those advantages to Lenovo rather than their cost to Intel, the Commission had in fact conducted its reasoning by reference to a less efficient competitor than Intel, which was not, however, the relevant economic operator for assessing whether the contested rebates were capable of having an anticompetitive foreclosure effect.

165 As is apparent from paragraphs 433 to 439 of the judgment under appeal, in support of that assessment, which concerns the principles of the AEC test, the General Court relied on the nature of that test as described in the decision at issue. Contrary to what the Commission argues in order to establish an infringement of its rights of defence by the General Court, the General Court did not rely, for that purpose, on the supplementary Shapiro-Hayes report of 28 January 2009 (‘the supplementary Shapiro-Hayes report’), produced by Intel for the first time at first instance, which, as is apparent from paragraphs 423 and 452 of the judgment under appeal, concerns a separate issue, namely the calculation of the cost to Intel of the two non-cash advantages.

166 In those circumstances, the General Court’s refusal, apparent from paragraphs 443 and 444 of the judgment under appeal, to take into account the calculations submitted by the Commission relating to the actual costs for Intel of the non-cash advantages concerned, on the ground that taking them into account would lead it to substitute its own assessment for that contained in the decision at issue, does not constitute an infringement of the Commission’s rights of defence.

167 It was only as part of an analysis of the evidence produced by the Commission for the first time before it, carried out for the sake of completeness in paragraphs 451 to 453 of the judgment under appeal, that, in paragraph 452 of that judgment, the General Court referred to the supplementary Shapiro-Hayes report.

168 It follows from the foregoing that the second ground of appeal must be rejected.

The third ground of appeal, alleging an error of law concerning the evidence required, infringement of the Commission’s rights of defence and distortion of the evidence in the examination of the AEC test in relation to Dell

169 The third ground of appeal consists of three parts.

The first part, alleging an error of law concerning the standard of proof required

–  Arguments of the parties

170 The Commission submits that the General Court relied on incorrect legal criteria in order to appraise the evidence adduced by the Commission in support of its finding that there was an abuse of a dominant position. In particular, first, the AEC test does not rely on suppositions about conduct derived from observed facts, but on the application of an econometric model to assessments of the input values for that model. Consequently, contrary to what is stated by the General Court in paragraph 165 of the judgment under appeal, the fact that Intel provided a plausible explanation of the facts from which it could be concluded that there was no infringement cannot lead to the annulment of the decision at issue. Second, the AEC test is, in essence, an evaluative exercise which, as regards the contestable share of an OEM, relies on assumptions for which, by definition, there is no certain and definitive answer, only a best or reasonable assessment. Thus, contrary to what the General Court held in paragraph 166 of the judgment under appeal, it is not sufficient to cast doubt on the assessment made by the Commission at the end of such a test or, a fortiori, merely to suggest that another result would be conceivable in order to obtain the annulment of the decision at issue.

171 The Commission submits that the correct approach to the standard of proof and judicial review for an AEC test is therefore the approach which, without exempting the AEC test from judicial review, consists not only in establishing whether the evidence relied on is factually accurate, reliable and consistent but also whether that evidence contains all the relevant information that must be taken into account in order to assess a complex situation and whether it is capable of substantiating the conclusions drawn from it. The capability of a dominant undertaking’s practice to restrict competition is assessed on the basis of contextual elements that are not exclusively linked to the conduct of that undertaking.

172 According to the Commission, it was therefore on the basis of an incorrect standard of proof that the General Court rejected the assessment of Dell’s contestable share in the decision at issue by merely stating, in paragraph 244 of the judgment under appeal, that ‘the assumption of a 7.1% contestable share was not the only conceivable assumption’ and that that fact ‘[cast] doubt on the substance of the assessment made by the Commission in the [decision at issue]’. The presumption of innocence does not mean that the assessments relating to contextual elements, not relating to the conduct of the dominant undertaking, must be regarded as invalid where the General Court has a ‘doubt’ about one of them.

173 Intel disputes the merits of this part of the third ground of appeal.

–  Findings of the Court

174 In order to assess the merits of the Commission’s arguments, it is necessary to recall the context of the AEC test carried out in the present case.

175 In paragraphs 133 and 134 of the judgment on the appeal, the Court of Justice recalled that it is not the purpose of Article 102 TFEU to prevent an undertaking from acquiring, on its own merits, a dominant position on one or more markets or to ensure that competitors less efficient than the undertaking with such a position should remain on the market. Thus, not every exclusionary effect is necessarily detrimental to competition. Competition on the merits may lead to the departure from the market or the marginalisation of competitors which are less efficient and so less attractive to consumers from the point of view of, among other things, price, output, choice, quality or innovation (judgment of 21 December 2023, European Superleague Company, C‑333/21, EU:C:2023:1011, paragraphs 126 and 127 and the case-law cited).

176 Consequently, in order to find, in a given case, that conduct must be categorised as ‘abuse of a dominant position’, it is necessary, as a rule, to demonstrate, through the use of methods other than those which are part of competition on the merits between undertakings, that that conduct has the actual or potential effect of restricting that competition by excluding equally efficient competing undertakings from the market or markets concerned or by hindering their growth on those markets (judgment of 21 December 2023, European Superleague Company, C‑333/21, EU:C:2023:1011, paragraph 129 and the case-law cited).

177 In that context, in view of the special responsibility of an undertaking in a dominant position not to impair, by its conduct, genuine undistorted competition in the internal market, Article 102 TFEU prohibits it from engaging in practices, including pricing practices, which have an exclusionary effect on competitors considered to be as efficient as the dominant undertaking itself, thereby strengthening its dominant position by using methods other than those which come within the scope of competition on the merits (see, to that effect, judgment on the appeal, paragraphs 135 and 136, and judgment of 12 May 2022, Servizio Elettrico Nazionale and Others, C‑377/20, EU:C:2022:379, paragraph 76).

178 Such a practice may take the form, as the Court of Justice points out in paragraph 137 of the judgment on the appeal, of a system of loyalty rebates, that is to say, discounts conditional on the customer’s obtaining all or most of its requirements – whether the quantity of its purchases from the undertaking in a dominant position be large or small – from that undertaking.

179 That being so, the demonstration that conduct has the actual or potential effect of restricting competition, which may entail the use of different analytical templates depending on the type of conduct at issue in a given case, must be made, in all cases, in the light of all the relevant factual circumstances, irrespective of whether they concern the conduct itself, the market or markets in question or the functioning of competition on that market or those markets. That demonstration must, moreover, be aimed at establishing, on the basis of specific, tangible points of analysis and evidence, that that conduct, at the very least, is capable of producing exclusionary effects (judgment of 21 December 2023, European Superleague Company, C‑333/21, EU:C:2023:1011, paragraphs 129 and 130 and the case-law cited).

180 Thus, as regards a practice consisting in the grant of loyalty rebates, in respect of which the undertaking in a dominant position submits, during the administrative procedure, on the basis of supporting evidence, that it was not capable of producing the alleged foreclosure effects, the Commission is required to analyse not only factors such as the extent of the dominant position of the undertaking in question, the share of market covered by the contested rebates and the conditions and arrangements for granting the rebates in question, their duration and their amount, but also the possible existence of a strategy aiming to exclude competitors that are at least as efficient as the dominant undertaking from the market (judgment on the appeal, paragraphs 138 and 139 and the case-law cited; see, to that effect, judgment of 19 January 2023, Unilever Italia Mkt. Operations, C‑680/20, EU:C:2023:33, paragraphs 47 to 49).

181 The capability of such rebates to foreclose a competitor as efficient as the dominant undertaking, which competitor is supposed to meet the same costs as those borne by that undertaking, must be assessed, as a general rule, using the AEC test. Even though that test is merely one of the ways of assessing whether an undertaking in a dominant position has used means other than those that come within the scope of ‘normal’ competition, it seeks specifically to assess whether such an as-efficient competitor, considered in abstracto, is capable of reproducing the conduct of the undertaking in a dominant position and, consequently, whether that conduct must be considered to come within the scope of normal competition, that is to say, competition on the merits (see, to that effect, judgment of 12 May 2022, Servizio Elettrico Nazionale and Others, C‑377/20, EU:C:2022:379, paragraphs 80 to 82 and the case-law cited).

182 In assessing the merits of the arguments raised by the Commission, it must also be borne in mind that, as set out in paragraphs 43 and 44 of the present judgment, the AEC test, as carried out in the present case, takes account, inter alia, of the possible loss of the contested rebates in the event of an OEM being supplied by AMD and is ultimately based on a comparison of the contestable share and the required share for each OEM and for MSH.

183 As regards the comparison of the contestable share and the required share, it is apparent from paragraph 175 of the judgment under appeal that the Commission set out the required share in Table 22 of the decision at issue (‘Table 22’).

184 In addition, in paragraph 171 of the judgment under appeal, the General Court states that the Commission used a contestable share of 7.1%, a figure that comes from the 2004 spreadsheet, referred to in paragraph 153 of the present judgment.

185 In paragraph 175 of the judgment under appeal, the General Court stated that, in essence, the Commission took that figure as the relevant percentage to be applied in order to determine the contestable share and compared it with the required share shown in Table 22 for each quarter concerned.

186 The final column of that table, to which paragraph 175 of the judgment under appeal also refers, shows that, for the period from the fourth quarter of the 2003 tax year to the fourth quarter of the 2006 tax year, the required share was between 4.9%, which is the value calculated for the first quarter of the 2004 tax year, and 12.1%, which is the value calculated for the fourth quarter of the 2006 tax year. As the General Court states in that paragraph of the judgment under appeal, the Commission took the view that for the last nine quarters of the period considered, the required share was higher than the contestable share.

187 After rejecting, in paragraphs 189 to 201 of the judgment under appeal, Intel’s arguments complaining that the Commission had infringed the principle of legal certainty by taking into account the 2004 spreadsheet, the General Court examined, in paragraphs 202 to 256 of that judgment, Intel’s arguments concerning the assessment of the contestable share as 7.1%.

188 In that regard, it is apparent from paragraphs 203 and 240 of the judgment under appeal that Intel disputed the contestable share of 7.1% used by the Commission, relying, in the first place, on evidence other than the 2004 spreadsheet and, in the second place, on data relating to the actual switch of Dell’s requirements to AMD in 2006 and in 2007, which, in its submission, demonstrated that Dell’s contestable share could be greater than 7.1%.

189 In the first place, the evidence on which Intel relied comprises, as is apparent from paragraphs 204 to 206, 215 and 220 of the judgment under appeal, the D 1 email, the D 5 email, the statements made by Dell executives in private litigation between Intel and AMD in the State of Delaware and the I1 declaration, referred to in paragraphs 147 to 151 of the present judgment.

190 The General Court analysed that evidence in paragraphs 213 to 232 of the judgment under appeal. In paragraph 233 of that judgment, the General Court held that it is apparent from the D 1 email and the D 5 email, the statements made by Dell executives in private litigation between Intel and AMD in the State of Delaware and the I1 declaration, which, taken together, supported one other, that, during 2005, the switch of Dell’s requirements to AMD could have involved up to 25% of the volume of x86 CPUs, and not 7% as was set out in the 2004 spreadsheet.

191 In those circumstances, the General Court held, in paragraph 234 of the judgment under appeal, that the evidence relied on by Intel cast doubt on the argument that Dell’s contestable share had to be assessed solely on the basis of the 2004 spreadsheet which mentioned a 7% volume switch of Dell’s requirements to AMD for 2005, from which the Commission had inferred a contestable share of 7.1%.

192 In the second place, the General Court observed, in paragraph 243 of the judgment under appeal, that the Commission had expressly acknowledged, in recital 1245 of the decision at issue, that it was possible to calculate a contestable share of more than 7.1%, being between 8.2% and 10.1%, on the basis of observations drawn from the actual switch of part of Dell’s requirements to AMD in 2006 and in 2007.

193 In that context, the General Court noted, first, in paragraph 244 of the judgment under appeal, that the very existence of those estimates was sufficient to demonstrate that the assumption of a 7.1% contestable share was not the only conceivable assumption and cast doubt on the substance of the assessment made by the Commission in the decision at issue.

194 Second, in paragraphs 246 to 251 of the judgment under appeal, the General Court rejected the Commission’s arguments that the switch of Dell supplies to AMD in 2006 and in 2007 was of only limited bearing when assessing the situation during the infringement period, that it was necessary at the very least to adapt certain calculation parameters, in particular the level of the contested rebates during 2006, and that the Commission had carried out, as an alternative, an AEC test in the decision at issue which took account of the situation in 2006 and in 2007, which supported its findings.

195 In that context, the General Court held, in paragraph 254 of the judgment under appeal, that it was apparent from the decision at issue that it was possible to determine a contestable share for Dell of between 8.2% and 10.1% on the basis of evidence other than the 2004 spreadsheet. Thus, the General Court reiterated the assessment that the very existence of estimates to that effect in the decision at issue itself demonstrated that the assumption of a 7.1% contestable share in the case of Dell was not the only conceivable assumption, which led it to doubt the substance of that assumption, relied on by the Commission in the decision at issue.

196 In paragraph 255 of the judgment under appeal, the General Court held that that finding and the finding made in paragraph 234 of that judgment, which concerns the assessment of the contestable share on the basis of evidence, relied on by Intel, other than the 2004 spreadsheet, when taken together, supported the existence of doubt as to the assessment of that contestable share in the decision at issue.

197 In the light of those findings, the General Court held, in paragraph 256 of the judgment under appeal, that the evidence put forward by Intel was capable of giving rise to doubt in the mind of the General Court as to whether the contestable share for Dell had to be set at 7.1%. Consequently, according to the General Court, the Commission had not demonstrated to the requisite legal standard that the assessment of that contestable share was well founded.

198 Moreover, in paragraphs 257 to 271 of the judgment under appeal, the General Court examined, for the sake of completeness, the substance of the Commission’s analysis of Dell’s contestable share as regards the initial part of the relevant period, between December 2002 and October 2003. In that regard, the General Court pointed out, in paragraph 260 of the judgment under appeal, that Table 22 illustrated clearly that the contestable share was greater than the required share, for the first four quarters set out in that table, even if the calculations of the required share and the contestable share made by the Commission were accepted.

199 After analysing the arguments put forward by the Commission in that regard, the General Court held, in paragraphs 270 and 271 of the judgment under appeal, that the Commission had not been able to explain or validate, a posteriori, the difference between the positive results for Intel, set out in Table 22 for the first four quarters of the infringement period, and its conclusion, adopted for the whole of that period, that Intel had not passed the AEC test.

200 The General Court’s final considerations on the AEC test in respect of the contested rebates granted to Dell are set out in paragraphs 283 to 287 of the judgment under appeal, the content of which is reproduced in paragraphs 48 to 51 of the present judgment.

201 In that regard, it should be noted that, as the Commission also submits, as carried out in the present case, the AEC test is an econometric model fed with input values making it possible, first, to calculate the required share and, second, to establish the contestable share. In calculating the required share, account is taken of the conditional rebates granted by the undertaking in a dominant position, the average avoidable cost and the average sales price for that undertaking. In calculating the contestable share, account is taken of estimates that may have been made by the undertaking that receives the rebates, but also by the other actors involved, regarding the share of the requirements that that undertaking may cover by obtaining supplies from a competitor of the undertaking in a dominant position.

202 Thus, the result of the AEC test is liable to indicate whether a pricing practice, such as loyalty rebates, adopted by an undertaking in a dominant position, with sufficiently pronounced characteristics in terms of the share of the market covered, the conditions and arrangements for granting those rebates, their duration and their amount, is capable of foreclosing a competitor as efficient as that undertaking and thus of being detrimental to competition as protected by Article 102 TFEU.

203 With regard to Dell, the positive or negative result of the AEC test depends, as is apparent from paragraphs 183 to 186 of the present judgment, on a comparison of two rates, that of the contestable share and that of the required share, expressed to the first decimal place. In order to calculate those rates, the Commission relies on assumptions that require a set of several figures to be taken into account.

204 It follows that, as the Commission again submits, it is not sufficient for the undertaking in a dominant position to call into question the accuracy of one of the calculations made in the AEC test in order to invalidate the Commission’s conclusion, based on such a test, relating to the capability of a loyalty rebate scheme to foreclose a competitor as efficient as that undertaking. The deficiency or error identified must also be capable of altering the result of the test, changing it from negative to positive, in such a way as to give rise to reasonable doubt as to the validity of the result adopted by the Commission and, therefore, as to the capability of the rebates concerned to foreclose a competitor as efficient as the undertaking in a dominant position. That type of doubt may arise due to calculation errors or due to selective or incomplete consideration of evidence.

205 In that regard, it is apparent from paragraphs 213 to 256 and 283 of the judgment under appeal that the General Court did not base its considerations relating to the AEC test with regard to the contested rebates granted to Dell on a doubt affecting a non-decisive part of the Commission’s assessment or on the mere suggestion that another result was conceivable for that test. Rather, the General Court relied on the fact that all the relevant data to be taken into consideration in assessing Dell’s contestable share – the assessment of which falls within the sole jurisdiction of the General Court unless distortion is alleged, as examined in the third part of the present ground of appeal, and of which only the 2004 spreadsheet had been taken into account by the Commission – indicated that that contestable share could in all likelihood be greater not only than the rate of 7.1% applied by the Commission, but also than the required share, so that the result of the AEC test could have been altered, changing from negative to positive, if all those data had been taken into account. It is on that basis that the General Court held, in its unappealable assessment of the evidence, that the Commission’s assumption that that contestable share was 7.1% had not been established to the requisite legal standard, so that the conclusions that the Commission had drawn from the AEC test were invalid.

206 The Commission cannot draw a valid argument to the contrary merely from the wording of paragraphs 244 and 254 of the judgment under appeal, according to which the existence of estimates in the decision at issue relating to the actual switch of part of Dell’s requirements to AMD in 2006 and in 2007 ‘is sufficient to demonstrate that the assumption of a 7.1% contestable share was not the only conceivable assumption and casts doubt on the substance of the assessment made by the Commission in the [decision at issue]’.

207 It is apparent from the wording of those paragraphs of the judgment under appeal, read in the light of paragraphs 242 and 243 of that judgment, that the General Court intended to emphasise that a contestable share greater not only than 7.1% but also than the required share was conceivable even on the basis of the Commission’s estimates in the decision at issue. Thus, that contestable share was ‘conceivable’ for the Commission itself, which nevertheless chose to rely solely on the 2004 spreadsheet. Furthermore, the General Court reached the finding in paragraph 254 of the judgment under appeal after rejecting, in paragraphs 245 to 251 of that judgment, the Commission’s arguments seeking to call into question the relevance of the data relating to the actual switch of Dell’s requirements to AMD in 2006 and in 2007.

208 It follows that the General Court did not disregard the nature of the AEC test as an econometric model for assessing the capability of the contested rebates to foreclose a competitor as efficient as the dominant undertaking, so that the first part of the third ground of appeal must be rejected.

The second part, alleging misapplication of the standard of proof laid down

–  Arguments of the parties

209 In the alternative, the Commission submits that the General Court misapplied the standard of proof which it itself laid down in paragraph 166 of the judgment under appeal. In particular, as the General Court held in paragraph 217 of that judgment, the 2004 spreadsheet, on which the Commission relied in order to calculate Dell’s contestable share, had greater evidential value than the documents or statements of Dell’s senior executives relied on by Intel. In those circumstances, the fact that the evidence relied on by Intel is not lacking in evidential value is not sufficient to call into question the evidence relied on by the Commission.

210 Intel disputes the merits of this part of the third ground of appeal.

–  Findings of the Court

211 In paragraph 166 of the judgment under appeal, the General Court observed, in essence, that when an undertaking in a dominant position raises the possibility that a circumstance arose which might affect the probative value of the evidence relied on by the Commission, it is for that undertaking to prove to the requisite legal standard the existence and relevance of the circumstance on which it relies and the influence of that circumstance on the probative value of the evidence relied on against it.

212 In that regard, it should be noted that the findings in paragraphs 217 and 218 of the judgment under appeal are a response to Intel’s argument, set out in paragraph 215 of that judgment, that (i) the documents on which Intel relied were drafted by senior Dell executives; (ii) D 1 had confirmed, under oath, in the course of private litigation between Intel and AMD in the State of Delaware, the content of his email of 10 November 2005; and (iii) in the course of that litigation, D 3, another Dell executive, had stated that he had no reason to question the accuracy of D 1’s remarks.

213 In paragraph 216 of the judgment under appeal, the General Court responded, to that particular argument, that answers given on behalf of an undertaking, such as the answer given by Dell during the administrative procedure, in which it had produced the 2004 spreadsheet, carry more weight than the answer given by an employee or one of its directors. It was therefore from the perspective of that general rule that the General Court referred, in paragraphs 217 and 218 of the judgment under appeal, to the greater evidential value and to the detailed and precise nature of the 2004 spreadsheet.

214 However, first, that general rule does not affect the Commission’s obligation to take into account all the relevant evidence in assessing whether an infringement of Article 102 TFEU may have been demonstrated to the requisite legal standard. That assessment is all the more necessary given that, as the General Court states in paragraph 172 of the judgment under appeal, the 2004 spreadsheet was also an internal Dell document speculating how the relationship between Dell and AMD could evolve, with AMD’s penetration increasing in Dell’s various business segments.

215 Second, it is apparent from the detailed examination of the evidence carried out by the General Court in paragraphs 220 to 256 of the judgment under appeal that it did not attribute to the evidence on which Intel relied greater evidential value than that of the 2004 spreadsheet, but that it considered, in its unappealable assessment of the evidence, that that evidence called into question to the requisite legal standard the validity of the contestable share established at 7.1% rather than at specific higher levels which, if they had been used, would have altered the outcome of the AEC test in Intel’s favour. In addition, in paragraphs 257 to 271 of the judgment under appeal, the General Court also considered, as noted in paragraphs 198 and 199 of the present judgment, that the Commission had not demonstrated that the rebates granted to Dell were capable of restricting competition during the initial part of the relevant period.

216 It follows that the Commission’s line of argument alleging that the General Court disregarded the standard of proof that it laid down is based on a partial reading of the judgment under appeal.

217 The second part of the third ground of appeal must therefore be rejected.

The third part, alleging distortion of the evidence in the examination of the AEC test in relation to Dell and infringement of the Commission’s rights of defence

–  Arguments of the parties

218 In the further alternative, the Commission submits that the General Court distorted the evidence, based its decision on contradictory reasoning and infringed the Commission’s rights of defence.

219 In particular, according to the Commission, the finding in paragraph 239 of the judgment under appeal, according to which there remained a doubt as to the definitive percentage of the contestable share for Dell and, more particularly, as to that percentage having to be set at 7.1%, is vitiated by a distortion of the evidence. It is apparent from paragraphs 204 to 206 of the judgment under appeal that Intel principally based its arguments on three documents, namely the D 1 email, the D 5 email and the I1 declaration, and on economic analyses based on those documents.

220 As regards the first two documents, the Commission submits that, by inferring, in paragraph 233 of the judgment under appeal, a contestable share of up to 25% of Dell’s requirements, the General Court conflated the requirements that Dell could manage to cover with AMD at the end of a period of one year with the average of those requirements over the course of that period. It is self-evident that that average was less than 25%. As the General Court acknowledged in paragraph 218 of the judgment under appeal, the Commission correctly relied on the precise and detailed nature of the information contained in the 2004 spreadsheet, which justified establishing Dell’s contestable share at 7.1%. Consequently, according to the Commission, contrary to what is stated by the General Court in paragraph 239 of the judgment under appeal, the fact that the D 1 email justifies an assessment of Dell’s contestable share between 5.6% and 10.4% does not cast doubt on the fact that Dell’s contestable share had to be set at 7.1%. In addition, the establishment of a contestable share on the basis of the documents referred to in paragraph 219 of the present judgment would require assumptions to be made as to the timing and rate of the ramp-up of the switch from Intel to AMD as Dell’s supplier, whereas the 2004 spreadsheet contained all the necessary information for that purpose. The Commission submits that the evidence relied on by the General Court enabled it to calculate the contestable share only in the form of ranges, whereas the precise and detailed information in the 2004 spreadsheet made it possible to calculate a specific figure, as acknowledged in paragraph 218 of the judgment under appeal.

221 As regards the third document, namely the I1 declaration, prepared for the purposes of the investigation and intended to exonerate Intel from any liability, the General Court characterised, in paragraph 227 of the judgment under appeal, its evidential value as poor, but then sought, in paragraph 230 of that judgment, to restore its credibility by means of contradictory grounds indicating that that declaration was distorted. Evidence that has poor evidential value cannot prevail over precise and detailed evidence. Without finding that the 2004 spreadsheet was not capable of substantiating the conclusions drawn from it in the decision at issue, the General Court applied an incorrect test consisting in determining whether the Commission’s well-substantiated assessment of the contestable share at 7.1% was the only conceivable one.

222 Moreover, in finding, in paragraph 247 of the judgment under appeal, that the Commission had used, for the purpose of assessing the contestable share, observations deduced from the partial switch of Dell’s requirements to AMD in 2006 and in 2007, the General Court distorted the decision at issue. The Commission did not take account of post-infringement data, so that it did not admit that the contestable share could be as high as 10.1% during the infringement period. The fact that the rising competitive pressure from AMD, noted in recital 1243 of the decision at issue, ultimately allowed AMD to compete with a contestable share of up to 10.1% between October 2006 and June 2007 necessarily implies lesser competitive pressure during the infringement period. The General Court’s findings in paragraphs 249 to 251 of the judgment under appeal are also based on a distortion of recitals 1245 and 1258 of the decision at issue, in so far as they assume that the Commission took into account data relating to the post-infringement period in calculating the contestable share and they ignore the information in Table 22. In paragraph 250 of the judgment under appeal, the General Court confused calendar years and Dell’s tax years, which led it to conclude incorrectly that the post-infringement period had been taken into account in the AEC test.

223 Moreover, contrary to the General Court’s finding in paragraph 263 of the judgment under appeal, no contradiction can be found between what is apparent, on the one hand, from recital 1256 of the decision at issue, according to which, in at least four quarters of the infringement period, Intel passed the AEC test and, on the other hand, from the Commission’s findings in recitals 1281 and 1282 of that decision, from which it follows that the contested rebates granted to Dell were capable of having a foreclosure effect throughout that period. Whereas recital 1263 of that decision is limited to a comparison of the required share and the contestable share, recital 1281 of that decision contains an overall assessment of the capability of the contested rebates to foreclose competition, also taking into account reinforcing factors, which the Commission is not required to quantify and the existence of which has not been disputed by Intel. Those reinforcing factors, which form part of the context of the AEC test and are relevant for the purpose of assessing the capability of the contested rebates to foreclose a competitor as efficient as Intel, were also present in the first four quarters of the infringement.

224 Similarly, the finding in paragraph 268 of the judgment under appeal, to the effect that the Commission did not take account of contestable shares of 17.3%, 22.5% and 24.2%, which are also apparent from the 2004 spreadsheet, is based on a distortion of recital 1212 of the decision at issue. The percentages referred to by the General Court relate to periods subsequent to the period taken into account for the purpose of calculating the contestable share, during which, after having gained market acceptance for its computers with AMD-based x86 CPUs, Dell would be able to increase its proportion of purchases of those CPUs. The fact that the General Court describes those higher figures as the ‘contestable share’ shows that it erred in finding that those figures called into question the Commission’s finding of a contestable share of 7.1%.

225 According to the Commission, that distortion of the decision at issue vitiates the General Court’s finding, in paragraph 270 of the judgment under appeal, that the Commission’s arguments did not make it possible to explain or validate, a posteriori, the difference between the results indicated by the Commission for the first four quarters of the infringement period in Table 22 and its conclusion, adopted for the whole of that period, that Intel did not pass the AEC test.

226 Last, the Commission submits that the General Court should have taken into account the economic analyses submitted by the Commission during the proceedings, which sought to show that, even if it were necessary to calculate the contestable share on the basis of the documents referred to in paragraph 219 of the present judgment, a contestable share of between 12.5% and 17.5% could not be inferred from them. By refusing to do so, the General Court, in paragraphs 235, 236, 252 and 253 of the judgment under appeal, infringed the Commission’s rights of defence.

227 Intel disputes the merits of this part of the third ground of appeal.

–  Findings of the Court

228 Given the exceptional nature of a complaint of distortion of the evidence, the appellant must, pursuant to Article 256 TFEU, the first paragraph of Article 58 of the Statute of the Court of Justice of the European Union and Article 168(1)(d) of the Rules of Procedure of the Court of Justice, indicate precisely the evidence alleged to have been distorted by the General Court and show the errors of appraisal which, in its view, led to such distortion. Such distortion must be obvious from the documents in the file, without there being any need to carry out a new assessment of the facts and the evidence (see, to that effect, judgment of 25 July 2018, QuaMa Quality Management v EUIPO, C‑139/17 P, EU:C:2018:608, paragraph 34).

229 While a distortion of the evidence may therefore consist in an interpretation of a document contrary to its content, it is not sufficient, in order to establish such a distortion, to show that that document could have been interpreted differently from the interpretation adopted by the General Court. To that end, it is necessary to establish that the General Court manifestly exceeded the limits of a reasonable assessment of that document, in particular by reading it in a manner contrary to its wording (judgments of 25 February 2021, Dalli v Commission, C‑615/19 P, EU:C:2021:133, paragraph 139, and of 23 March 2023, PV v Commission, C‑640/20 P, EU:C:2023:232, paragraph 134).

230 It should be recalled that the passages of the judgment under appeal relating to the AEC test applied to Dell consist of a section concerning the assessment of the contestable share (paragraphs 171 to 271 of the judgment under appeal), a section concerning an alternative calculation applied by the Commission in order to confirm that the contested rebates granted to Dell were capable of foreclosing a competitor as efficient as Intel (paragraphs 272 to 276 of that judgment) and a section concerning the taking into account of certain factors which, if they were included in the AEC test, reinforced, according to the decision at issue, the estimated foreclosure capability of those rebates. Those reinforcing factors are, first, that Dell clearly considered that any loss by it of the contested rebates would be accompanied by an increase in the contested rebates granted by Intel to Dell’s OEM competitors and, second, that the estimate of the contestable share did not take account of the fact that Dell also purchased from Intel products other than x86 CPUs, in particular chipsets (paragraphs 177 and 277 to 282 of that judgment).

231 The section relating to the assessment of Dell’s contestable share is divided into three subsections. The first subsection, which includes paragraphs 189 to 201 of the judgment under appeal, concerns Intel’s arguments based on the principle of legal certainty, which the General Court rejected. The second subsection, which includes paragraphs 202 to 256 of that judgment, concerns Intel’s arguments challenging the contestable share of 7.1% found by the Commission, which the General Court upheld. The third subsection, which includes paragraphs 257 to 271 of that judgment, concerns Intel’s claims regarding the initial part of the infringement period, from December 2002 to October 2003, which the General Court examined for the sake of completeness and also upheld.

232 As regards the section concerning the alternative calculation applied by the Commission, the General Court concluded, in paragraph 276 of the judgment under appeal, that it did not demonstrate that Intel’s rebate practices were capable of having a foreclosure effect throughout the infringement period.

233 As regards the section concerning the reinforcing factors, the General Court held, in paragraph 282 of the judgment under appeal, that those factors had been included in the decision at issue as elements which were capable of reinforcing the principal assessment regarding the existence of a foreclosure effect caused by the contested rebates and that those factors had not been analysed sufficiently by the Commission as regards the effect that they would have had on the assessment of the foreclosure capability of those rebates. Thus, according to that paragraph of the judgment under appeal, the General Court could not usefully review how those factors were taken into account, nor could consideration of those factors replace the Commission’s principal analysis of the capability of the contested rebates granted to Dell to have a foreclosure effect.

234 The arguments put forward by the Commission in support of the complaint alleging distortion of the evidence relate to the last two subsections of the section of the judgment under appeal concerning the assessment of Dell’s contestable share and to the reinforcing factors (see paragraphs 231 and 233 of the present judgment).

235 In that regard, as is apparent from paragraphs 203 and 240 of the judgment under appeal, the General Court examined in two stages Intel’s line of argument challenging the contestable share of 7.1% found by the Commission in relation to Dell. In the first place, in paragraphs 203 to 239 of that judgment, the General Court examined the evidence on which Intel had relied to dispute that contestable share and, in the second place, in paragraphs 240 to 255 of that judgment, it examined Intel’s arguments concerning the actual switch of Dell’s requirements to AMD in 2006 and in 2007.

236 In that context, first, the Commission’s argument that the General Court distorted the D 1 email or the D 5 email by conflating a volume switch of Dell’s requirements to AMD at the end of a period of one year with Dell’s contestable share is based on a misreading of the judgment under appeal. It is apparent, in particular, from paragraphs 204, 233 and 234 of the judgment under appeal that the General Court distinguishes between the concept of the switch of volume of purchases of x86 CPUs from Intel to AMD, on the one hand, and that of the contestable share, on the other. In particular, in those paragraphs, the General Court states that the projection of a switch of 25% of such volume could, according to the estimates of the parties to the dispute, translate into a contestable share between 12.5% and 17.5%, and it compares that volume switch to the 7% taken into account by the Commission in order to establish a contestable share of 7.1%. In those circumstances, it must be held that, under the guise of a claim of distortion, the Commission in fact seeks, by relying on the high evidential value of the 2004 spreadsheet as compared with that of the D 1 email, to have the Court of Justice reassess the evidence. That line of argument must therefore be rejected as inadmissible.

237 Second, as regards the Commission’s line of argument relating to the I1 declaration, it is apparent from paragraph 233 of the judgment under appeal that that declaration was only one of the four items of evidence relied on by the General Court to find, in the same paragraph, that the switch of Dell’s requirements to AMD could have involved up to 25% of the volume of x86 CPUs, and not 7% as was set out in the 2004 spreadsheet. In addition, the Commission does not indicate precisely the evidence alleged to have been distorted by the General Court. In those circumstances, the argument in question does not in fact allege distortion of the I1 declaration, within the meaning set out in paragraphs 228 and 229 of the present judgment, but rather seeks to have the Court of Justice reassess that evidence and its importance among all the items of evidence examined by the General Court, so that it must be rejected as inadmissible.

238 As regards, third, the arguments by which the Commission claims that the General Court’s findings set out in paragraphs 247 to 251 of the judgment under appeal, concerning the taking into account of the data from the actual switch of part of Dell’s requirements to AMD in 2006 and in 2007, that is to say, after the infringement period, which ended in December 2005, are vitiated by a distortion of the decision at issue, the following should be noted.

239 In paragraphs 247 and 248 of the judgment under appeal, the General Court observed that the Commission could not rely on the relevance of the information deduced from the partial switch of Dell’s requirements to AMD in 2006 and in 2007 to contradict Intel’s argument relating to the starting point for the one-year time horizon in calculating the contestable share, while at the same time denying the relevance of that information for the purpose of assessing the precise rate of the contestable share. Contrary to the Commission’s claim, in adopting that position of principle, which criticises the Commission’s selective use of the evidence, the General Court did not find that the Commission had used the observations deduced from the partial switch of Dell’s requirements to AMD in 2006 and in 2007 for the purpose of assessing the contestable share.

240 In addition, in paragraph 249 of the judgment under appeal, the General Court rejected the Commission’s argument, set out in its supplementary observations on the referral of the case back to the General Court, according to which the relevance of the actual switch of part of Dell’s requirements to AMD in 2006 and in 2007 had to be adapted because of the increase in the contested rebates granted by Intel during that period. In that regard, the General Court observed that the Commission had not adjusted the contestable share used in recital 1245 of the decision at issue, assessed at between 8.2% and 10.1%, to take account of that factor. It is not apparent from that paragraph of the judgment under appeal that the General Court considered that the Commission had applied contestable share rates set at between 8.2% and 10.1% during the infringement period. The General Court merely observes therein that, if the Commission took the view that a certain factor affected the reliability of the rates referred to in recital 1245 of that decision, it ought to have adjusted them to take account of that factor.

241 In paragraph 250 of the judgment under appeal, the General Court refers to another argument that the Commission put forward in its supplementary observations on the referral of the case back to the General Court, according to which the AEC test had ‘incorporated’ AMD’s actual market shares in Dell’s purchases in 2006 and in 2007 and that the results of those calculations supported the findings of the decision at issue ‘for the period ending in 2005’. In that regard, in paragraph 251 of the judgment under appeal, the General Court observed that the Commission had not adapted the contestable share used for 2005 based on the calculations made in recital 1245 of that decision, which relate to the post-infringement period. Without such adaptation, it cannot be claimed that the results of the calculations made in that decision supported the findings in that decision relating to the infringement period, especially since they show a high contestable share, which rather calls those findings into question.

242 It follows from those considerations that the General Court correctly distinguished between the data in the AEC test relating to the infringement period and the data relating to the post-infringement period, which did not form part of that test. Thus, the General Court examined, in response to Intel’s arguments, whether the Commission’s assessments concerning the relevance of the data relating to the post-infringement period were well founded, without distorting the content of the relevant recitals of the decision at issue.

243 Furthermore, the Commission’s arguments that the General Court infringed its rights of defence by refusing, in paragraphs 236 and 253 of the judgment under appeal, to take into account the additional analyses, referred to in paragraphs 235 and 252 of that judgment, produced by the Commission for the first time during the proceedings in order to contradict evidence allegedly produced by Intel for the first time before the General Court, cannot succeed. As is apparent from paragraphs 156 to 161 of the present judgment, the General Court did not take account of evidence produced by Intel for the first time before it, so that the arguments put forward by the Commission in support of that ground of appeal are, in any event, based on an incorrect premiss.

244 In the light of the assessments set out in paragraphs 235 to 243 of the present judgment, the Commission’s arguments relating to the assessment carried out by the General Court, for the sake of completeness, in paragraphs 257 to 271 of the judgment under appeal, concerning the analysis of Dell’s contestable share regarding the initial part of the infringement period, from December 2002 to October 2003, must be rejected as ineffective. As stated in paragraph 257 of the judgment under appeal, the General Court carried out that assessment even though the conclusion that it had reached in paragraph 256 of the judgment under appeal, according to which the Commission had not demonstrated to the requisite legal standard that the assessment of Dell’s contestable share made in the decision at issue was well founded, already invalidated, in itself, the Commission’s assessment of that contestable share.

245 Last, the Commission’s argument that the General Court erred in law by requiring the reinforcing factors relied on by the Commission in order to demonstrate the capability of the contested rebates to foreclose a competitor as efficient as Intel to be quantified is based on a misreading of paragraphs 278 to 282 of the judgment under appeal. It follows from those paragraphs that the General Court rejected the analysis carried out in the decision at issue concerning the transfer of the contested rebates granted to Dell to Dell’s competitors and the loss of those rebates in connection with the purchase of chipsets on the ground that, without an assessment in figures in that regard, the General Court could not review effectively whether those two reinforcing factors were able to counteract the deficiencies stemming from the errors that vitiated the AEC test relating to Dell’s contestable share or to demonstrate the foreclosure capability of those rebates. Consequently, the General Court did not consider that the reinforcing factors had to be precisely set out in figures in all circumstances, but that, if they were to replace the Commission’s principal analysis of the capability of the contested rebates granted to Dell to foreclose a competitor as efficient as Intel, the examination of those factors had to be able to demonstrate that capability.

246 It follows that the third part must be rejected, as must the third ground of appeal in its entirety.

The fourth ground of appeal, alleging several errors of law and infringement of the Commission’s rights of defence in the examination of the AEC test in relation to HP

247 The fourth ground of appeal comprises two parts.

The first part, alleging errors of law in the examination of the AEC test in relation to HP

–  Arguments of the parties

248 According to the Commission, the General Court made errors of law that vitiate the judgment under appeal by holding, in paragraphs 319 and 335 of the judgment under appeal, that the calculation of HP’s required share did not permit the inference that the contested rebates granted to HP were capable of foreclosing competition for the period from 1 November 2002 to 30 September 2003.

249 In that regard, first, the Commission claims that the General Court applied an incorrect standard of proof which, as regards the AEC test relating to HP, rendered paragraphs 292 to 320 of the judgment under appeal unlawful. Thus, the General Court erred in law in finding, in paragraph 335 of the judgment under appeal, that the Commission had not demonstrated that, during the period from November 2002 to May 2005, the contested rebates granted to HP were capable of having an anticompetitive foreclosure effect.

250 In particular, the General Court did not dispute that the decision at issue also provided the required share for the period from November 2002 to September 2003. However, in paragraphs 304 and 305 of the judgment under appeal, the General Court took issue with the fact that, in order to determine that share, the Commission relied on figures derived from the arithmetic sum or average of the figures regarding only three quarters, namely the fourth quarter of the 2003 tax year and the first and second quarters of the 2004 tax year. According to the Commission, those figures are sufficiently representative for the entire period of the first agreement concluded between Intel and HP, in force between November 2002 and May 2004 (‘the HPA 1 agreement’), which is why the Commission relied on them.

251 The General Court therefore took issue with the Commission’s choice of an economic parameter, in the present case the reference period chosen to calculate the required share. That economic parameter is not an issue of fact, but an aspect of a complex economic assessment in respect of which the Commission enjoys broad discretion, and the General Court’s review must therefore be limited to verifying whether there has been any manifest error of assessment, such as the Commission’s use of an unlawful method. Intel did not produce an alternative calculation for the quarters concerned and the Commission’s approach was, in any event, favourable to Intel. In those circumstances, the General Court substituted its own assessment for that of the Commission.

252 Moreover, the Commission’s choice of methodology to determine that reference period is based on implicit reasoning that stems from the fact that Table 35 of the decision at issue presents the required share as covering the entire duration of the HPA 1 agreement and from the fact that, as stated in recital 346 of the decision at issue, the most decisive factor, namely the amount of the contested rebates granted to HP, remained stable throughout the period covered by that agreement, which Intel did not dispute.

253 Second, the Commission submits that the General Court erred in law by refusing to take into account the fact that Intel had not challenged, during the administrative procedure, the periods used by the Commission for its calculations in the statement of objections and, moreover, that Intel had even used those periods for its own calculations, which it submitted to the Commission during that procedure.

254 Third, according to the Commission, even if the General Court’s finding that the Commission had not demonstrated any anticompetitive foreclosure effect during the period between November 2002 and September 2003 were well founded, that would not justify finding, as the General Court did in paragraph 335 of the judgment under appeal, that that error vitiated the entire analysis that the contested rebates granted to HP were capable of foreclosing competition during the period from November 2002 to May 2005, which includes the period from June 2004 to May 2005 covered by the second agreement concluded between Intel and HP (‘the HPA 2 agreement’). On the contrary, as the General Court acknowledged in paragraph 525 of the judgment under appeal, the AEC test could be regarded as conclusive for the period from October 2003 to May 2005.

255 Fourth, the Commission submits that, in concluding in paragraph 334 of the judgment under appeal that the decision at issue failed to provide reasons, to the requisite legal standard, as regards the reinforcing factors, the General Court necessarily referred to both reinforcing factors taken into account in that decision, cited in paragraph 321 of the judgment under appeal, namely the Commission’s use of figures that were favourable to Intel and the risk of the contested rebates initially intended for HP being transferred to another competitor of HP. However, the General Court examined the reasoning of the decision at issue only with regard to the second of those factors, so that the judgment under appeal did not state reasons, even implicitly, so far as concerns the first of those factors. In any event, the General Court also erred in law by holding, in paragraphs 328, 331 and 334 of the judgment under appeal, that the decision at issue failed to provide reasons, to the requisite legal standard, as regards the second reinforcing factor. In that regard, the Commission observes, in particular, that the General Court in fact required it to go beyond an assessment of the capability of the contested rebates to foreclose a competitor as efficient as Intel and to quantify the foreclosure effects, by reference to a factor that is hypothetical and unrelated to the parameters of the AEC test. Consequently, even if the General Court were correct in holding, in the judgment under appeal, that the decision at issue did not demonstrate the existence of an anticompetitive foreclosure effect for the period from November 2002 to September 2003, the general conclusion drawn in paragraph 335 of that judgment is incorrect.

256 Intel disputes the merits of the Commission’s arguments.

–  Findings of the Court

257 The General Court set out the factual background relating to the contested rebates granted to HP in paragraphs 288 and 289 of the judgment under appeal. It is apparent from those paragraphs that Intel had concluded with HP, for the period from November 2002 to May 2005, two successive agreements on equipping corporate desktop computers with x86 CPUs. The first of those agreements, the HPA 1 agreement, covered a period from November 2002 to May 2004 and the second, the HPA 2 agreement, covered a period from June 2004 to May 2005. The grant of the rebates provided for in those agreements was subject to the unwritten condition that HP would source from Intel at least 95% of its x86 CPU requirements in order to equip its corporate desktop computers. According to paragraph 288 of the judgment under appeal, which refers, in that regard, to recital 1406 of the decision at issue, the Commission concluded on the basis of the AEC test that those rebates were capable of having anticompetitive foreclosure effects.

258 According to paragraphs 298, 299, 303 and 321 of the judgment under appeal, that conclusion is based, on the one hand, on a comparison of the contestable share and the required share and, on the other, on two reinforcing factors. Those factors consist, first, of the Commission’s use of figures that were most favourable to Intel and, second, of the fact that, if HP were to switch its x86 CPU purchases to AMD, Intel could in turn transfer the contested rebates initially intended for HP to another competitor of HP, such as Dell.

259 It follows from paragraph 299 of the judgment under appeal that, according to the decision at issue, the Commission had established HP’s contestable share at 7%.

260 The requisite share, for its part, is shown in Table 34, which is found in recital 1334 of the decision at issue (‘Table 34’), to which paragraph 303 of the judgment under appeal refers. The final column of that table shows that, for the period from the third quarter of the 2003 tax year to the third quarter of the 2006 tax year, the required share was between 9.7%, which is the value calculated for the fourth quarter of the 2003 tax year, and 18.9%, which is the value calculated for the first quarter of the 2005 tax year. The Commission thus concluded, in recital 1387 of the decision at issue, to which paragraph 298 of the judgment under appeal refers, that the required share was above the contestable share for the entire infringement period.

261 In addition, Table 35, which is found in recital 1337 of the decision at issue, to which paragraph 292 of the judgment under appeal refers, shows, according to that recital, the robustness of the Commission’s conclusion and is presented in such a way as to suggest that it encompasses the entire period covered by the HPA 1 agreement. That table shows a required share of 10.6% for the period covered by the HPA 1 agreement.

262 It is apparent from paragraph 291 of the judgment under appeal that Intel claimed that the Commission made, inter alia, four errors concerning (i) the contestable share; (ii) the amount of the conditional portion of the contested rebates; (iii) the infringement period examined; and (iv) the reinforcing factors taken into consideration. The General Court focused its analysis on the errors alleged in the third place and in the fourth place.

263 As regards the infringement period examined, the General Court observed, in paragraph 303 of the judgment under appeal, that Table 34 covered the period from the fourth quarter of the 2003 tax year to the third quarter of the 2005 tax year, with the result that it did not incorporate any data relating to the months of November and December 2002 or the first three quarters of the 2003 tax year.

264 Similarly, in paragraph 304 of the judgment under appeal, the General Court noted that the figures for the HPA 1 agreement found in the first line of Table 35 are derived from the arithmetic sum or average of the figures in the first three lines of Table 34. The General Court set out the details underlying that assessment in paragraph 305 of that judgment, adding, in paragraph 306 thereof, that the Commission had not maintained that that mathematical relationship was the product of a coincidence or that the various values set out were identical for the missing three quarters and for the following three quarters.

265 In those circumstances, the General Court held, in paragraph 307 of the judgment under appeal, that the months of November and December 2002 and the first three quarters of the 2003 tax year had not actually been taken into account by the Commission in the calculations which gave rise to the figures set out in Table 35. It added, in the same paragraph, that the calculation of the required share throughout the duration of the HPA 1 agreement which produced the results set out in Tables 34 and 35 therefore did not cover the whole of the period between November 2002 and May 2005, in relation to which the Commission had taken the view that it could demonstrate that there was an anticompetitive foreclosure effect caused by the contested rebates granted to HP.

266 In paragraphs 308 to 320 of the judgment under appeal, the General Court rejected the arguments that the Commission had based, first, on the fact that the result of a quarterly calculation does not differ fundamentally from the result of the overall calculation which is alleged to have been made and, second, on references to an annex to the defence and an annex to the rejoinder.

267 As regards the Commission’s arguments alleging that the AEC test involves complex economic assessments in respect of which the Commission enjoys discretion that the General Court should not have infringed, it should be borne in mind that the Commission carried out an AEC test in order to demonstrate that the contested rebates were capable of foreclosing a competitor as efficient as Intel.

268 As the Advocate General states, in essence, in points 83 to 85 of her Opinion, the calculation of the required share involves methodological choices that the Commission must make in such a way as to ensure that its conclusions are based on all the factually accurate, reliable and consistent information necessary for that purpose. In that context, the Commission was entitled to decide to analyse the capability of the contested rebates granted to HP to foreclose a competitor as efficient as Intel on a quarterly basis for a period from November 2002 to May 2005. That being so, in the absence of any other means used in the decision at issue to demonstrate that capability, the AEC test must be able usefully to support the conclusion that those rebates have such a capability for the whole of that period.

269 First, the Commission does not claim that the findings of fact made by the General Court, which fall within its unappealable assessment, to the effect that Table 34, which concerns the calculation of the required share, does not cover the period up to the beginning of the fourth quarter of the 2004 tax year and, moreover, that Table 35 is equally incomplete despite a reference to the HPA 1 agreement, are the result of any distortion of evidence or distortion of the decision at issue itself.

270 Second, as the Advocate General states, in essence, in points 91 and 92 of her Opinion, even though the use of the mechanism of extrapolation is not a choice that is prohibited in principle, in so far as that mechanism is intended to infer the unknown from data that is known, it must be based on a concrete logical pattern, defined and made explicit by the party bearing the burden of proof, in the present case the Commission.

271 The Commission does not claim, nor is it apparent from the judgment under appeal, that it relied before the General Court on the representativeness of the data relating to the period after the third quarter of the 2003 tax year in order to justify extrapolating those data to the part of the infringement period up to the end of that quarter, still less that such representativeness is apparent from the decision at issue. On the contrary, footnote 1658, mentioned in recital 1337 of the decision at issue, to which paragraphs 292 and 297 of the judgment under appeal refer, states that the period examined does not coincide fully with the duration of the agreements at issue because, as regards the HPA 1 agreement, the figures provided by HP did not cover the entire duration of that agreement. Furthermore, that reference rules out the possibility of considering that the decision at issue contains an implicit statement of reasons relating to the representativeness of the data relating to the period after the third quarter of the 2003 tax year, which could have justified extrapolation. Moreover, since the Commission does not challenge the statement set out in paragraph 306 of the judgment under appeal, described in paragraph 264 of the present judgment, it must be held that the General Court did not err in law in holding, in paragraph 307 of the judgment under appeal, that the months of November and December 2002 and the first three quarters of the 2003 tax year had not actually been taken into account by the Commission.

272 Third, in paragraph 318 of the judgment under appeal, the General Court examined, for the sake of completeness, the calculations produced by the Commission as an annex to the rejoinder to demonstrate that the results relating to HP’s required share would have been the same for the missing two months and three quarters as the results relating to that share for the quarters that were actually taken into account by the Commission. In that regard, the General Court considered that there was nothing to demonstrate the accuracy of the premiss that, because the contested rebates were stable during the period covered by the HPA 1 agreement, the results of the required share for that period would also be applicable to the period that was not taken into account, especially since there was no indication that the volume of HP’s purchases and the average sales price had remained the same throughout the infringement period. The Commission merely states that the amount of the contested rebates lost by HP is the most important factor in the calculation of the required share, without, however, claiming that the General Court’s assessment is vitiated by a distortion of the evidence.

273 As regards the Commission’s argument that Intel had not disputed, during the administrative procedure, the periods used by it for its calculations, it must be borne in mind that the Commission is required to assess, inter alia, the possible existence of a strategy aiming to exclude, through the grant of loyalty rebates, competitors that are at least as efficient as the dominant undertaking from the market where that undertaking submits, during the administrative procedure, that its conduct was not capable of restricting competition and, in particular, of producing the alleged foreclosure effects (see paragraphs 138 and 139 of the judgment on the appeal).

274 It follows that the Commission is entitled to base its assessment on the same premisses and evidence as those submitted by the undertaking concerned, just as it may carry out its own analyses on the basis of other evidence obtained during its investigation.

275 However, the fact that Intel submitted, during the administrative procedure, calculations which, even if based on the reference periods used by the Commission, show, according to Intel, that the contested rebates were not capable of foreclosing a competitor as efficient as it is clearly distinguishable from a situation in which Intel produced, during that procedure, evidence relating to one part of the infringement period that it itself extrapolated to another part of the infringement period, which the Commission in turn relied on in the statement of objections and, in the absence of any challenge, in the decision at issue. That distinction applies irrespective of whether, as it claims, Intel had identified, during the administrative procedure, the gap relating to the period from November 2002 to July 2003.

276 In that context, it should be borne in mind that, as is already apparent from paragraph 163 of the present judgment, and as the General Court correctly pointed out in paragraph 300 of the judgment under appeal, there is no requirement under the law of the European Union that the addressee of the statement of objections must challenge its various matters of fact or law during the administrative procedure, if it is not to be barred from doing so later at the stage of judicial proceedings (judgment of 1 July 2010, Knauf Gips v Commission, C‑407/08 P, EU:C:2010:389, paragraph 89).

277 Thus, since Intel submitted, during the administrative procedure, on the basis of supporting evidence, that its conduct had not been capable of restricting competition with regard to HP, thereby triggering the Commission’s obligation to assess the possible existence of a strategy aiming to exclude competitors that were at least as efficient as Intel from the market, there was nothing to prevent it from challenging before the General Court the elements of that assessment relating to the period at issue that formed part of the statement of reasons for the decision at issue. It follows that the General Court did not err in law in adopting the same approach in paragraphs 300 to 302 of the judgment under appeal.

278 The Commission’s argument that the General Court erred in law in finding that the gap in the AEC test relating to the part of the infringement period that lasted until the end of the third quarter of the 2003 tax year vitiated the analysis relating to the entire infringement period, which ran from November 2002 to May 2005, is based on a misreading of paragraph 335 of the judgment under appeal. In that paragraph, the General Court stated that the finding, set out in recital 1406 of the decision at issue, that the contested rebates granted to HP were capable of having an anticompetitive foreclosure effect during the period from November 2002 to May 2005 had not been established to the requisite legal standard only ‘since [the Commission had] not demonstrated that those effects were present for the period between 1 November 2002 and [30] September 2003’. Thus, in paragraph 525 of the judgment under appeal, the General Court accepted that that fact was not capable of leading to the annulment of the decision at issue for the entire infringement period taken into account in relation to HP. In those circumstances, as is apparent from the same paragraph of that judgment, the General Court annulled the decision for the entirety of that period only after finding, in paragraphs 485 to 520 of that judgment, that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and that it had not analysed correctly the duration of those rebates.

279 It follows that, contrary to what the Commission claims, the General Court did not base the annulment of the decision at issue with regard to the contested rebates granted to HP on the gap affecting only part of the infringement period.

280 The Commission’s argument that the General Court examined only the reinforcing factor of the risk of the contested rebates initially intended for HP being transferred to another competitor of HP, excluding the factor of the Commission’s use of figures that were favourable to Intel, must also be rejected. It is sufficient to note in that regard that, in view of the nature of the deficiency found by the General Court consisting in the complete lack of evidence relating to the first part of the infringement period, a ‘reinforcing factor’ based on the Commission’s use of evidence that was favourable to Intel could not, by definition, concern that part. The General Court was therefore not required to devote any particular analysis to that factor.

281 As regards the Commission’s argument that the General Court in fact required it to go beyond an assessment of the capability of the contested rebates to foreclose a competitor as efficient as Intel and to quantify the foreclosure effects produced by a reinforcing factor, it is apparent from paragraphs 328 and 329 of the judgment under appeal that the failure to state reasons found by the General Court in paragraph 331 of that judgment concerns, in essence, the possibility that the reinforcing factor consisting of a transfer of the contested rebates initially granted to HP to one of its competitors would make it possible to reject all of Intel’s arguments concerning the AEC test and to counteract all the errors contained in the decision at issue in that regard.

282 In that regard, in paragraph 330 of the judgment under appeal, the General Court noted that the argument put forward by the Commission at the hearing, according to which HP refused AMD’s offer to provide it with one million x86 CPUs free of charge solely because of the consequences that acceptance would have on its relationship with Intel, was a simple unsubstantiated assumption, which could not make up for the lack of reasoning in the decision at issue as to the influence of the reinforcing factor consisting of a transfer of the contested rebates initially granted to HP to one of its competitors on the Commission’s conclusions stemming from the AEC test.

283 In that context, it is apparent from a combined reading of paragraphs 321, 334 and 335 of the judgment under appeal that the General Court examined whether recitals 1390 to 1395 of the decision at issue, which deal with the reinforcing factors, contained grounds that made it possible to make up for the lack of evidence underlying the calculation of HP’s required share for the two months and the three quarters for which the AEC test had proved deficient.

284 In that regard, it must be borne in mind that the capability of the contested rebates to foreclose a competitor as efficient as Intel was assessed by means of a comparison, carried out as part of the AEC test, of HP’s required share and its contestable share. That comparison is based, in turn, on a calculation taking into account the amount of the contested rebates, which are lost if the underlying exclusivity conditions are not met, the volume of HP’s purchases and Intel’s average sales price and average avoidable cost. As is apparent from recitals 1393 and 1394 of the decision at issue, to which reference is made in paragraph 321 of the judgment under appeal, the possible transfer of those rebates to one of HP’s competitors is a quantifiable parameter, discussed in the contacts between HP and AMD, that is capable of influencing, inter alia, the effect created by the loss of those rebates taken into account for the purpose of calculating the required share. Similarly, in recital 1285 of the decision at issue, relied on by the Commission in support of that part of the fourth ground of appeal, the Commission states that the calculation of the required share concerning HP, carried out in the decision at issue, is ‘based on the assumption that the potentially lost rebate is not distributed to [HP’s] competitors’.

285 Consequently, as the General Court states in paragraphs 328 and 329 of the judgment under appeal, if the Commission intended to rely on the possible transfer of the contested rebates granted to HP to one of its competitors in order to establish the capability of those rebates to foreclose a competitor as efficient as Intel irrespective of the gaps in the AEC test, the Commission was required to set out the reasoning illustrating the influence of that transfer on one or other of the factors underlying the calculation of the required share.

286 The General Court’s analysis would be no less well founded if, as the Commission submits in support of this part of the fourth ground of appeal, recital 1285 of the decision at issue were to be understood as meaning that the parameters for calculating the required share with regard to HP preclude taking into account the possible transfer of the contested rebates granted to HP to one or more of that undertaking’s competitors. If such a transfer is not included in the parameters of that calculation, it can only exacerbate the impact of anticompetitive foreclosure, as the Commission noted in recital 1392 of the decision at issue. As is apparent from paragraphs 267 to 280 of the present judgment, the General Court did not err in law in concluding that the Commission had not demonstrated the anticompetitive foreclosure capability of the contested rebates for two months and three quarters of the infringement period, so that an exacerbation of the impact of a possible anticompetitive foreclosure cannot be taken into account for that period, in the absence of a specific statement of reasons to that effect.

287 It follows that, contrary to what is claimed by the Commission, the General Court did not implicitly require it to go beyond an assessment of the capability of the contested rebates to foreclose a competitor as efficient as Intel or to quantify the foreclosure effects. The General Court merely held that, in order to find that a reinforcing factor such as that relied on by the Commission is capable of allowing Intel’s arguments concerning the calculations made for the purposes of the AEC test to be rejected and the errors vitiating that test to be counteracted, the factors that make it possible to recognise that capability must, like the main factors of those calculations, be apparent from the statement of reasons for the decision finding the infringement.

288 In the light of the foregoing, the first part of the fourth ground of appeal must be rejected.

The second part, alleging infringement of the Commission’s rights of defence in the examination of the AEC test in relation to HP

–  Arguments of the parties

289 The Commission states that, in paragraphs 316 and 317 of the judgment under appeal, the General Court refused to take into consideration certain evidence produced in the course of the proceedings. That evidence shows that, even if Intel’s new argument concerning the reference period for the AEC test were accepted, that would not vitiate the outcome of that test. For the same reasons as those set out in support of the second ground of appeal, the General Court’s refusal to examine the rebuttal of Intel’s claims infringed the Commission’s rights of defence.

290 Intel disputes the merits of this part of the fourth ground of appeal.

–  Findings of the Court

291 It should be noted that, since, in the light of the arguments and evidence produced by Intel during the administrative procedure, the Commission was required to examine whether that undertaking had implemented a strategy aiming to exclude competitors that were at least as efficient as it from the market, it was for the Commission to demonstrate, in the decision at issue, that the strategy in question had such a capability.

292 In the present case, it is apparent from paragraphs 316 and 317 of the judgment under appeal that the Commission relied before the General Court on an additional calculation covering the second and third quarters of the 2003 tax year for the first time in a document produced as an annex to the rejoinder.

293 As stated in paragraph 163 of the present judgment, the General Court did not err in refusing to take account of those additional calculations produced by the Commission for the first time as an annex to the rejoinder, all the more so since the Commission does not claim to have justified their late submission at first instance, as required by Article 85(2) of the Rules of Procedure of the General Court.

294 It follows that the second part of the fourth ground of appeal and, with it, that ground of appeal in its entirety must be rejected as unfounded.

The fifth ground of appeal, alleging misinterpretation of the AEC test and of Article 102 TFEU, distortion of the evidence and infringement of the Commission’s rights of defence in the examination of the AEC test in relation to Lenovo

Arguments of the parties

295 The Commission recalls that the General Court, in its examination of the AEC test in so far as it relates to Lenovo, focused on the value of two non-cash advantages that Lenovo enjoyed as rebates in return for exclusive supply from Intel, for the loss of which a competitor as efficient as Intel would have had to offer it compensation. Those were, first, the extension of Intel’s standard one-year warranty and, second, the making available of an Intel supply hub in Shenzhen (China). In that regard, the Commission submits that, in paragraphs 436 to 439 of the judgment under appeal, the General Court disregarded the nature of the AEC test, which led to an incorrect definition of the competitor ‘as efficient’ as Intel and, ultimately, an infringement of Article 102 TFEU.

296 Moreover, first, in paragraph 454 of the judgment under appeal, the General Court did not classify the errors of assessment found in the preceding paragraphs of that judgment as ‘manifest’. According to the Commission, in view of the methodological choices required by an AEC test, only ‘manifest’ errors of assessment could lead to the annulment of a decision based on such a test.

297 Second, by relying on all the costs incurred by Intel, including those incurred in making a supply hub available to Lenovo, and not only on those relating to the production of CPUs, the General Court overlooked the fact that the ultimate issue to be resolved under Article 102 TFEU is whether there is anticompetitive foreclosure capability resulting from the existence of the contested rebates, which are also applied to the non-contestable share, which is much larger than the contestable share. Consequently, the focus should have been on whether the competitor as efficient as Intel would be able to secure the contestable share of an OEM’s requirements, notwithstanding that it operates on a more modest scale than Intel.

298 Thus, it is not the cost of the contested rebates for the dominant undertaking that should be taken into account, but the benefit that those rebates represent for the undertaking that receives them, namely the customer of that dominant undertaking. That approach, which would ensure that the dominant undertaking’s position would not operate to its advantage in the AEC test, should be adopted irrespective of whether the rebates take the form of payments in cash or in kind. In addition, it is apparent from the evidence produced by Intel that it did not consider that the costs of making a supply hub available to Lenovo were part of the average avoidable cost of a competitor as efficient as Intel.

299 In any event, according to the Commission, the General Court incorrectly rejected, in paragraph 438 of the judgment under appeal, the approach adopted by the Commission in the decision at issue, according to which a competitor of Intel would not have available a supply hub such as that made available to Lenovo and that, as a result, it would have had to offer Lenovo compensation for the loss of that non-cash advantage.

300 As regards the extension of the standard warranty, the Commission submits that the General Court disregarded the fact that Intel would continue to sell to Lenovo a far larger non-contestable share than the contestable share sold by the competitor. Assuming that the failure rate of x86 CPUs during their warranty period and the repair or replacement costs would be the same, the total cost of meeting the warranty claims on Intel’s x86 CPUs in Lenovo computers would be much higher than that of meeting the warranty claims on Lenovo computers containing x86 CPUs from the competitor as efficient as Intel. Consequently, the similar level of efficiency of the two producers does not alter the fact that, to persuade Lenovo to forego the extended warranty offered by Intel in return for exclusivity, the competitor as efficient as Intel would have to offer its x86 CPUs at a price that would offset the cost to Lenovo of warranty claims that it could not pass on to Intel. The Commission was therefore correct, in its AEC test, in treating the value to Lenovo of the extension of Intel’s standard warranty in the same way as a cash rebate.

301 In the alternative, the Commission submits that the General Court infringed the Commission’s rights of defence by refusing to examine the evidence to rebut Intel’s claims that the Commission submitted in the course of the proceedings as Annex D.39 to the rejoinder. To the extent that the General Court did examine the evidence produced by the Commission as Annex D.39 to the rejoinder, from which it follows, according to the Commission, that Intel estimated the costs of the non-cash advantages at USD 47 million, it did not give reasons for its refusal, in paragraph 452 of the judgment under appeal, to take account of a document entitled ‘Intel chart entitled 2006 vs. 2007 Trend’, from which it is apparent that the cost of extending the standard warranty amounted to USD 23 million for Intel. If the General Court did not take that document into account on the ground that the aim of that document may have been to present Intel’s proposal in a favourable light during the negotiations with Lenovo, that ground constitutes a distortion of the document in question since that document is marked ‘Intel confidential’.

302 Intel disputes the merits of the fifth ground of appeal.

Findings of the Court

303 In paragraph 98 of the judgment under appeal, the General Court noted that Intel and Lenovo had concluded a declaration of intent, the 2007 Memorandum of Understanding (‘the 2007 MoU’), which was subject to an unwritten exclusivity condition. In paragraph 415 of that judgment, the General Court referred to recital 1461 of the decision at issue, in which the Commission had stated that the amount of the contested rebates granted to Lenovo was set out in the 2007 MoU, which provided for a financial support of USD 180 million for 2007, in the form of quarterly payments.

304 In paragraph 417 of the judgment under appeal, the General Court referred to recital 1463 of the decision at issue, which sets out Intel’s argument that only a value of USD 138 million was relevant for assessing the amount of those rebates. Thus, it is explained, in paragraph 417 of that judgment, that, out of the financial support for Lenovo of USD 180 million provided for in the 2007 MoU, only USD 135 million were awarded in the form of cash. The remainder of that support was granted in the form of non-cash advantages, namely the extension of Intel’s standard one-year warranty and the offer of improved use of an Intel supply hub in China. The Commission stated that Intel had submitted that, whilst the value of those two non-cash contributions for Lenovo was USD 20 million and USD 24 million respectively, their cost for Intel was much lower, namely USD 1.7 million and USD 1.3 million respectively. Intel had also submitted that, in order to determine whether a competitor was as efficient as Intel, those elements had to be assessed not by reference to their value to Lenovo, but by reference to their economic cost to Intel. Intel reached the amount of USD 138 million by adding those costs of USD 1.7 million and USD 1.3 million to the financial support paid in cash, namely USD 135 million.

305 In paragraphs 420 and 421 of the judgment under appeal, the General Court states that, in recitals 1466 and 1467 of the decision at issue, the Commission, first, rejected the latter argument made by Intel on the ground that the AEC test required an examination of the price at which a competitor which is as efficient as the dominant company – but which is not dominant – would have had to offer its products in order to compensate the customer for the loss of the conditional benefits granted by the dominant undertaking and which would result from that customer’s switching the contestable share of its supply needs away from the dominant undertaking to the hypothetical as-efficient competitor. Second, the Commission had inferred therefrom that it was necessary to assess the loss for the customer, since that was the loss which the as-efficient competitor would have to compensate, and not the economic costs to the dominant undertaking, in the event that the two figures diverged.

306 In paragraphs 433 to 439 of the judgment under appeal, the General Court found that the method used by the Commission to assess the extent of the contested rebates granted to Lenovo was based on an assumption that was contrary to the foundation of the AEC test, as set out in recitals 1003 and 1004 of the decision at issue.

307 In that regard, in paragraph 434 of the judgment under appeal, the General Court pointed out that, according to recital 1003 of the decision at issue, the purpose of the AEC test is to examine whether Intel itself, in view of its own costs and the effect of the contested rebates, would be able to enter the market at a more limited scale without incurring losses. Thus, in paragraph 435 of that judgment, the General Court emphasised that, as set out in recital 1004 of that decision, the AEC test is a hypothetical exercise, in the sense that it attempts to analyse whether a competitor as efficient as Intel, in terms of producing and delivering x86 CPUs that provide the same value to customers as Intel, but which would not have as broad a sales base as Intel, would be foreclosed from entering.

308 It should also be borne in mind that, as the General Court stated in paragraphs 154, 157 and 158 of the judgment under appeal, reproduced in paragraph 43 of the present judgment, the AEC test, as applied in the present case, establishes the ‘effective price’, namely the price at which a competitor as efficient as Intel would have had to offer its x86 CPUs in order to compensate an OEM for the loss of any exclusivity payment granted by Intel, and the effective price must then be compared with ‘Intel’s measure of viable cost’, namely the average avoidable cost. A system of exclusivity payments is capable of foreclosing competitors as efficient as Intel if the effective price is below Intel’s average avoidable cost. As is, moreover, set out in paragraph 44 of the present judgment, the positive or negative result of the AEC test is ultimately determined by means of a comparison of the contestable share and the required share, the latter being the proportion of the requirements of a customer of Intel that a competitor as efficient as Intel must obtain in order to be able to enter the market without incurring losses.

309 In that regard, it is apparent from paragraph 412 of the judgment under appeal, which refers to recitals 1457 to 1508 of the decision at issue, that the extent and nature of the contested rebates granted to Lenovo are factors taken into account when calculating the required share. When those rebates are granted in cash, their value is objective and is the same for both Intel and the undertaking that receives them. By contrast, where they are granted in non-cash form, even in part, they must be assessed.

310 It should be noted, in that context, that it is true that a competitor as efficient as Intel must offer its products at a price that compensates for the loss of the conditional benefits granted by the dominant undertaking. However, that compensation does not necessarily have to take the form of a cash advantage equal to the value of the non-cash advantage for the customer concerned. It may consist in a non-cash advantage equivalent to that which that customer will lose following its decision to obtain the contestable share from the competitor of the undertaking in a dominant position. It is, also in that case, irrelevant that that advantage may represent, from the subjective point of view of the customer benefiting from it, a value that differs from the cost that the competitor as efficient as the undertaking in a dominant position had to incur in order to grant it to that customer.

311 Consequently, it is necessary, in accordance with the principles of the AEC test, to assess a rebate granted in the form of a non-cash advantage by taking into account a hypothetical competitor with a cost structure similar to that of the undertaking in a dominant position (see, to that effect, judgments of 14 October 2010, Deutsche Telekom v Commission, C‑280/08 P, EU:C:2010:603, paragraph 198, and of 19 January 2023, Unilever Italia Mkt. Operations, C‑680/20, EU:C:2023:33, paragraph 59).

312 Such an approach is all the more justified because, as the General Court correctly pointed out in paragraph 190 of the judgment under appeal, it is also consistent with the general principle of legal certainty. The account taken of the costs of the dominant undertaking enables that undertaking, in the light of its special responsibility under Article 102 TFEU, to assess the lawfulness of its own conduct, since it knows its own costs (see, to that effect, judgment of 14 October 2010, Deutsche Telekom v Commission, C‑280/08 P, EU:C:2010:603, paragraph 202).

313 That being so, it cannot be ruled out, as the Advocate General states in points 160 and 165 of her Opinion, that the cost of the non-cash advantage may require a certain adjustment to allow account to be taken of the fact that the as-efficient competitor does not hold a dominant position. Such an adjustment may be necessary where the costs of the competitor as efficient as the dominant undertaking are affected because it satisfies only the contestable share of the customers, which is smaller than the non-contestable share of those customers, supplied by that undertaking.

314 It follows, as the Advocate General observes in point 155 of that Opinion, that, in finding, in paragraphs 437 and 438 of the judgment under appeal, that the Commission’s reasoning was not conducted on the basis of a hypothetical competitor being capable of selling x86 CPUs to Lenovo while also offering non-cash advantages to Lenovo on the same terms as Intel and, in paragraph 439 of that judgment, that, in so doing, the Commission proceeded on the basis of an assumption that was contrary to the foundation of the AEC test set out in the decision at issue, the General Court did not substitute its assessment for a valid analysis carried out by the Commission, but that it highlighted an internal inconsistency in the AEC test as applied in the present case by the Commission.

315 As the General Court observes, moreover, in paragraph 441 of the judgment under appeal, without being challenged by the Commission in the appeal, the decision at issue provides no response to the question of what the cost would have been for a competitor as efficient as Intel if it had had to provide access to a supply hub or simply modify its own existing hub so as to extend it to an OEM, as Intel offered to Lenovo, or what the costs of a warranty extension would have been.

316 It is apparent from paragraph 428 of the judgment under appeal that the Commission presented such an analysis for the first time in Annex D.39 to the rejoinder. However, that analysis goes beyond the limits defined by the statement of reasons for the decision at issue, so that, as it stated in paragraphs 443 and 444 of the judgment under appeal, the General Court cannot take it into account.

317 In that regard, the complaint raised by the Commission in the alternative, alleging infringement of its rights of defence (see paragraph 301 of the present judgment), must be rejected for the reasons set out in paragraphs 164 to 167 of that judgment, from which it follows that the General Court did not err in law or infringe the Commission’s rights of defence in stating, in paragraph 444 of the judgment under appeal, that it could not take account of the evidence produced for the first time as an annex to the rejoinder and thus go beyond the limits defined by the statement of reasons for that decision without substituting its own reasoning for that of the Commission set out in the decision at issue.

318 Moreover, in paragraph 448 of the judgment under appeal, the General Court examined, for the sake of completeness, the analysis contained in Annex D.39 to the rejoinder and rejected it on the ground that the costs incurred, according to the Commission, by a competitor as efficient as Intel having a supply hub in China differ from the value for Lenovo of the non-cash advantages that the Commission took into account for the purposes of its AEC test. In any event, the Commission does not claim that that assessment, carried out for the sake of completeness by the General Court, is vitiated by any distortion of the evidence.

319 Similarly, the Commission’s argument that, in paragraph 452 of the judgment under appeal, in which the General Court examined the evidential value of the document entitled ‘Intel chart entitled 2006 vs. 2007 Trend’, produced by the Commission for the first time as Annex D.41 to the rejoinder, the General Court ignored the cost to Intel of extending the standard warranty as set out in that document and accepted the lower cost calculated in the supplementary Shapiro-Hayes report, must also be rejected.

320 As is apparent from the words ‘In any event’ at the beginning of paragraph 451 of the judgment under appeal, the General Court examined the evidential value of that document for the sake of completeness. The Commission does not dispute the General Court’s principal finding in paragraph 450 of that judgment that the information in that document was not mentioned in the decision at issue. In those circumstances, the Commission’s argument is ineffective.

321 It follows from the foregoing considerations that the fifth ground of appeal must be rejected.

The sixth ground of appeal, alleging an incorrect assessment of the consequences to be drawn from the errors found in the AEC test

Arguments of the parties

322 The Commission raises the sixth ground of appeal in the alternative, should the Court of Justice reject the other grounds of appeal. In that context, the Commission submits that the General Court erred in law as regards the consequences to be drawn, first, from the errors found in the AEC test and, second, from the concept of a competitor as efficient as Intel as derived from the judgment on the appeal.

323 In particular, first, the Commission submits that the finding of errors vitiating the AEC test does not mean that Intel passed that test and that the contested rebates were not capable of foreclosing competition, but only that the Commission did not prove, in the decision at issue, that Intel had failed that test. In that regard, referring to its main observations on the referral of the case back to the General Court, which are based in particular on recitals 1037 and 1038 of that decision, the Commission recalls that ‘Intel’s measure of viable cost’ adopted in that decision is that of the average avoidable cost, which is a conservative, favourable cost benchmark for Intel, since it disregards the very high fixed costs that a competitor of Intel would have to incur in order to remain a viable competitor in the long term. Thus, Intel’s failure to pass the AEC test, taking into account the average avoidable cost, is only one consideration to be taken into account in the overall assessment that is needed to answer the question whether the contested rebates were capable of foreclosing competition. The Federal Republic of Germany adds that, in some circumstances, such as those of the present case, exclusivity agreements may cause the trading partners to be more reticent about accepting equivalent offers from competitors of their supplier due to the commitment entered into and the dependency with regard to the non-contestable share of the products purchased.

324 According to the Commission, the General Court therefore distorted the argument put forward by it in its observations on the referral of the case back to the General Court by holding in paragraphs 518 and 519 of the judgment under appeal that the sole reference to the period during which Intel implemented the contested rebates and to their timing was not sufficient, in itself, notwithstanding the conclusions that could be drawn from the AEC test, to justify definitive conclusions as to the foreclosure effects thereby produced.

325 It follows that the General Court’s conclusion, set out in paragraphs 526 and 527 of the judgment under appeal, that the Commission had not established that the contested rebates and the payments at issue were capable of having or likely to have anticompetitive foreclosure effects is vitiated by an error of law. Before reaching that conclusion, it was for the General Court, as part of an overall assessment of the circumstances of the case, to examine whether the AEC test was a particularly decisive factor in the decision at issue and whether the errors identified in that test were significant enough to lead to the annulment of that decision.

326 Second, the Commission submits that that conclusion disregards the fact that the concept of an ‘as-efficient competitor’ is not based solely on considerations relating to prices and costs. In that regard, it is apparent from paragraph 139 of the judgment on the appeal that the question whether the contested rebates were capable of foreclosing a competitor as efficient as Intel also depends on parameters such as the choice, quality and innovation demonstrated by that competitor’s products. It is apparent from paragraphs 116 to 122 and 134 of the judgment under appeal that the General Court failed to take account of those criteria, even though they were relevant to the concept of an ‘efficient competitor’ such as AMD, whose products are both innovative and high-performance and, consequently, attractive to the OEMs.

327 Intel disputes the merits of that line of argument.

Findings of the Court

328 In the first place, it must be borne in mind that it is for the Commission to prove the infringements of the competition rules which it has found and to adduce evidence capable of demonstrating to the requisite legal standard the existence of the constituent elements of an infringement (see, to that effect, judgments of 6 January 2004, BAI and Commission v Bayer, C‑2/01 P and C‑3/01 P, EU:C:2004:2, paragraph 62, and of 16 February 2017, Hansen & Rosenthal and H&R Wax Company Vertrieb v Commission, C‑90/15 P, EU:C:2017:123, paragraph 26).

329 Those constituent elements must be apparent from the statement of reasons for the measure finding the infringement, since the Courts of the European Union cannot, as noted in paragraphs 138 and 163 of the present judgment, alter them by substituting, in the context of the review of legality referred to in Article 263 TFEU, their own reasoning for that of the author of the measure in question.

330 In the second place, it follows from paragraphs 175 to 181 of the present judgment that, although the grant of loyalty rebates by an undertaking in a dominant position may constitute an infringement of Article 102 TFEU, the fact that that undertaking submits, during the administrative procedure, on the basis of supporting evidence, that its conduct was not capable of restricting competition and, in particular, of producing the alleged foreclosure effects requires the Commission to carry out an analysis to determine the existence of that capability.

331 In that situation, since the demonstration of the actual or potential effect of restricting competition must be assessed in the light of all the relevant factual circumstances, the Commission is required to analyse, as the Court of Justice emphasised in paragraph 139 of the judgment on the appeal, not only the extent of the undertaking’s dominant position on the relevant market, the share of the market covered by the contested rebates, the conditions and arrangements for granting those rebates, their duration and their amount, but also the possible existence of a strategy aiming to exclude competitors that are at least as efficient as that undertaking from the market.

332 It is on the basis of that analysis that the Commission must, in that case, fulfil its obligation, described in paragraphs 328 and 329 of the present judgment, to establish, in the statement of reasons for the decision adopted following its investigation, that the conduct complained of amounts to an abuse of a dominant position within the meaning of Article 102 TFEU.

333 In the third place, it follows from paragraphs 142, 143, 147 and 149 of the judgment on the appeal and, in particular, from paragraphs 30, 31, 152 to 158, 175, 283, 285, 286, 297 to 299, 339, 412, 413, 457 and 466 to 468 of the judgment under appeal that it is by means of the AEC test, which seeks to establish the price at which a competitor as efficient as Intel would have had to offer its x86 CPUs in order to ‘compensate’ an OEM for the loss of the contested rebates, that the Commission stated reasons, in the decision at issue, for the capability of the contested rebates to foreclose such a competitor and thereby to restrict competition, with regard to the OEMs and to MSH, in a way that is prohibited by Article 102 TFEU.

334 As is apparent, moreover, from paragraphs 147 and 149 of the judgment on the appeal, it is with regard to the fact that the Commission had carried out an AEC test that the Court of Justice referred the case back to the General Court, so that the General Court could examine whether the contested rebates were capable of restricting competition in the light of Intel’s arguments seeking to expose alleged errors made by the Commission in that test. As the General Court observed in paragraph 523 of the judgment under appeal, the Commission concluded, in the light of the results of that test, that the contested rebates and the payments at issue granted by Intel were capable of having or likely to have anticompetitive foreclosure effects, since even a competitor as efficient as Intel would have been prevented from supplying Dell, HP, NEC and Lenovo for their x86 CPU requirements or from having MSH sell computers equipped with its x86 CPUs.

335 In that context, it follows from paragraph 157 of the judgment under appeal that the Commission took into account, for the purposes of the AEC test, Intel’s average avoidable cost. It is common ground that the Commission relied on that cost benchmark on the ground that, in view of the more favourable nature of that benchmark for Intel, the fact that it would be impossible for the competitor as efficient as Intel to cover such a cost indicated clearly that it would sacrifice its profits and incur losses when making sales to the customers covered by the conduct of the undertaking in a dominant position, so that competition would be foreclosed.

336 In the fourth place, it is apparent from paragraphs 485, 493 to 495, 509 and 510 of the judgment under appeal that the Commission devoted a number of recitals of the decision at issue to the analysis of the share of the market covered by the contested rebates and of their duration and that, among the recitals relating to that duration, several are in the part of the decision at issue devoted to the AEC test. As stated in paragraphs 62 to 76 of the present judgment, the General Court found, following an examination carried out in paragraphs 483 to 521 of the judgment under appeal, the specific aspects of which are not disputed in the appeal, that the Commission had not considered properly the criterion relating to the share of the market covered by the contested rebates and had not analysed correctly the duration of those rebates.

337 It is in those circumstances that, as stated in paragraphs 80 to 82 of the present judgment, the General Court found, in paragraphs 524 to 527 of the judgment under appeal, that the errors which it had identified in the AEC test, the share of the market covered by the contested rebates and the duration of their application justified the annulment of Article 1(a) to (e) of the decision at issue.

338 In the light of the foregoing considerations, it was not for the General Court to examine, before annulling the decision at issue in part, whether the contested rebates were capable of foreclosing a competitor as efficient as Intel by relying, for the purposes of that examination, on factors different from those on which the Commission had relied in order to establish that capability.

339 In particular, it was not for the General Court to examine whether that capability could be established by means of reasoning that did not contain the errors found by it, in so far as such reasoning is not set out in the decision at issue in a coherent way from which it may be inferred that it may support the operative part of that decision irrespective of the grounds vitiated by errors of law or of fact that render the Commission’s findings invalid.

340 In that regard, first, it is apparent from paragraphs 133 to 147 of the judgment under appeal that the analysis carried out in the decision at issue, intended to demonstrate that the contested rebates constitute an abuse irrespective of the conclusions drawn by the Commission from the AEC test, is vitiated by an error of law in so far as it starts from the premiss that the contested rebates were abusive irrespective of whether they were capable of foreclosing a competitor as efficient as Intel. Second, as regards the AEC test, the Commission does not maintain that that decision contains an additional examination taking into account a cost benchmark other than the average avoidable cost that would make it possible to find that capability, or that it established, in that decision, that the conduct complained of also included other components that would have been capable of giving rise, irrespective of the conclusions to be drawn from the AEC test, to an anticompetitive foreclosure effect that Intel pursued by exploiting the non-contestable share.

341 In those circumstances, the General Court did not err in law in holding, in paragraphs 518 and 519 of the judgment under appeal, that the sole reference, notwithstanding the conclusions that may be drawn from the AEC test, to the period during which Intel implemented the contested rebates and to their timing was not sufficient, in itself, to justify definitive conclusions as to the foreclosure effects thereby produced. That consideration appears all the more justified in the light of the General Court’s findings concerning, first, the share of the market covered by the contested rebates and their duration, set out in paragraphs 492 to 500 and 508 to 515 of the judgment under appeal respectively and, second, the contested rebates as implemented with regard to NEC and to MSH, the specific aspects of the findings made in that regard not being disputed in the appeal.

342 Moreover, the Commission’s argument that the General Court failed to take account of the fact that AMD is an ‘efficient’ competitor, as defined by the judgment on the appeal, because its products are high-performance, innovative and attractive, is also irrelevant.

343 As is apparent from paragraph 435 of the judgment under appeal, the AEC test is a hypothetical exercise that makes it possible to determine whether a competitor which is as efficient as Intel, in terms of producing and delivering x86 CPUs that provide the same value to customers as Intel, but which would not have as broad a sales base as Intel, would be foreclosed from the market as a result of the contested rebates. It follows that that analysis is independent of AMD’s actual ability to remain on the market. The analysis in question may therefore demonstrate that the contested rebates were, in breach of Article 102 TFEU, capable of foreclosing a competitor presumed to be as efficient as Intel in terms of costs, choice, quality and innovation, even if AMD had not itself been foreclosed, just as it may reveal the lack of such capability, despite the fact that one or other of the dominant undertaking’s competitors left the market or was marginalised.

344 It follows from the foregoing that the sixth ground of appeal must be rejected and, consequently, that the appeal must be dismissed in its entirety.

Costs

345 In accordance with Article 184(2) of the Rules of Procedure of the Court of Justice, where the appeal is unfounded, the Court is to make a decision as to the costs.

346 Under Article 138(1) of those rules of procedure, applicable to appeal proceedings by virtue of Article 184(1) thereof, the unsuccessful party is to be ordered to pay the costs if they have been applied for in the successful party’s pleadings.

347 Since Intel and ACT have applied for costs to be awarded against the Commission and the Commission has been unsuccessful, it must be ordered to pay the costs.

348 Pursuant to Article 140(1) of the Rules of Procedure of the Court of Justice, applicable to appeal proceedings by virtue of Article 184(1) thereof, the Member States and institutions which have intervened in the proceedings are to bear their own costs.

349 Since Intel and ACT have applied for costs to be awarded against the Federal Republic of Germany, the latter must bear its own costs and pay those incurred by Intel and ACT in connection with the intervention.

On those grounds, the Court (Fifth Chamber) hereby:

1. Dismisses the appeal;

2. Orders the European Commission to bear its own costs and to pay those incurred by Intel Corporation Inc. and by Association for Competitive Technology Inc.;

3. Orders the Federal Republic of Germany to bear its own costs and to pay those incurred by Intel Corporation Inc. and by Association for Competitive Technology Inc. in connection with the intervention.