In Case T‑286/09

(1)General overview of the part of the contested decision devoted to Lenovo

412The Commission conducted the AEC test for Lenovo in recitals1457 to 1508 of the contested decision. The Commission analysed, first of all, the extent and nature of the rebates, on the basis of the 2007 MoU.

413The Commission then calculated the ASP, the costs and the required number of x86 CPU units.

414Finally, the Commission assessed the contestable number of x86 CPU units. In its principal calculation, it limited its assessment to the notebook segment (see recitals1473 to 1478 of the contested decision), whereas, in its alternative calculations, it responded to Intel’s assertions that the contestable number of x86 CPU units should also include the desktop segment (see recitals1479 to 1508 of the contested decision). Those alternative calculations are subdivided, first, into a response from the Commission to Intel’s allegations regarding the overall contestable number of x86 CPU units and, second, into a confirmatory calculation, made on the basis of a comparison with data from a document entitled ‘Statement of Work April 2006’, adopted pursuant to an agreement between AMD and Lenovo.

(2)The conditional portion of the rebates

415In recital1461 of the contested decision, the Commission stated that the amount of the rebates in question was set out in the 2007 MoU, which provided for a financial support of USD180 million for 2007, in the form of quarterly payments.

416In recital1462 of the contested decision, it was stated that the payments under the 2007 MoU were incremental to payments which Intel continued to make under other previously agreed financial support programmes, separately from the 2007 MoU. The Commission therefore took the view that they had to be attributed entirely to the outcome of the agreement on the 2007 MoU. All payments and favourable commercial terms provided for in the 2007 MoU were conditional on Lenovo cancelling its plans for notebook computers with AMD-based x86 CPUs.

417In recital1463 of the contested decision, it was stated that, in the submission of 5February 2009, Intel had put forward the argument that the relevant value for the size of the rebates was only USD138 million. That is explained by the fact that, out of the financial support for Lenovo of USD180 million provided for in the 2007 MoU, only USD135 million were awarded in the form of cash. The remainder of the financial support was awarded 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 hub facility in China. The Commission stated that Intel had submitted that, whilst the value of those two non-cash contributions for Lenovo was USD20 million and USD24 million respectively, their cost for Intel was much lower, namely USD1.7 million and USD1.3 million respectively. Intel had submitted that, for the purposes of the as-efficient-competitor analysis, those elements had to be assessed not by reference to their value for Lenovo, but by reference to their economic cost to Intel. Intel reached the amount of USD138 million by adding the USD1.7 million and USD1.3 million costs to the USD135 million cash funding.

418In recital1464 of the contested decision, the Commission stated that, before discussing the validity of Intel’s argument on the appropriate measure to be used in the as-efficient-competitor analysis, it had noted the disparity between the alleged economic cost of the contributions for Intel and their value to Lenovo. The ratio between the value to Lenovo and the alleged economic costs for Intel was 1176% (20:1.7) for the warranty extension and 1846% (24:1.3) for the hub facility. The Commission stated that Intel had provided certain calculations carried out for the purpose of the 5February 2009 submission related to the Supplementary Statement of Objections of 2008 in support of its assertion regarding the economic cost of the contributions, but that it had failed to explain the reason for the stark discrepancy between those costs and their value to Lenovo.

419In recital1465 of the contested decision, the Commission noted that, without prejudice to the observation mentioned above, Intel’s argument that the appropriate measure to be used in the as-efficient-competitor analysis was not the value of such items to Lenovo but their economic cost for Intel was based on a misunderstanding of the principles of that analysis.

420In that regard, in recital1466 of the contested decision, the Commission stated that ‘the as-efficient-competitor analysis assesses the price at which a competitor which is as efficient as the dominant company– but which is not dominant– would have to offer its products in order to compensate the customer for the loss of the conditional benefits granted by the dominant company and which would result from that customer’s switching the contestable share of its supply needs away from the dominant company to the hypothetical as-efficient competitor’.

421Lastly, in recital1467 of the contested decision, the Commission stated that it was clear from the foregoing that the relevant measure was the loss for the customer, as that was the loss which the as-efficient competitor would have to compensate, and not the economic costs to the dominant company, in the event that the two figures diverged. That difference was illustrated, according to the Commission, by the case of the supply hub. As a dominant undertaking, Intel already had a supply hub in China, in respect of which it claimed that it needed to make only marginal improvements, with an economic value of USD1.3 million, in order to be in a position to offer Lenovo a full benefit value of USD24 million. However, the Commission stated that a competitor as efficient as the dominant company, but which was not dominant, would normally not have put in place such a facility at that stage. In order to compensate Lenovo for the loss of the benefit linked to improved use of Intel’s supply hub, the as-efficient competitor would therefore have to have awarded Lenovo a monetary payment equivalent to the economic value of the improved supply hub facility for Lenovo.

422The applicant makes a general assertion that the 2007 MoU did not support the conclusion that its rebates would have resulted in it foreclosing an as-efficient competitor. The Commission’s analysis, first, overstates the allegedly conditional portion of the discount, second, it understates the contestable share and, third, it overstates Intel’s costs. As regards, more specifically, the conditional portion, the applicant states that, so far as the rebates under the 2007 MoU are concerned, the contested decision concludes in recitals1461 and 1474 to 1477 that conditional rebates of USD180 million were granted for a contestable share of only 0.9 to 1.1 million notebook computers. However, according to the applicant, the amount of conditional rebates was only USD138 million.

423According to the applicant, the Commission’s methodology for taking account of those non-cash advantages is incorrect, on the ground that, for the purposes of the AEC analysis, the conditional rebate ought to take account of the costs incurred by the applicant in providing such rebates, and not of the value which they represent to Lenovo. The supplemental Shapiro-Hayes report of 28January 2009 (‘the Supplemental Shapiro-Hayes report’) calculated the cost to Intel of the two non-cash advantages at approximately USD3 million. Of the USD3 million, USD1680073 (rounded to USD1.7 million) correspond to the warranty extension and another USD1256948 (rounded to USD1.3 million) to Intel’s cost of offering Lenovo a supply hub.

424The Commission disputes all of the applicant’s arguments. The Commission submits that, in order to assess the conditional portion of the rebates, the contested decision assessed the incremental funding awarded by Intel to Lenovo in 2007 under the 2007 MoU. That approach is not disputed. According to the Commission, the decision established that that incremental funding amounted to USD180 million, based on the examination of slides prepared by Intel during the negotiation of the 2007 MoU. The Commission’s view is that it was correct to rely on the value of the non-cash advantages granted to Lenovo, instead of taking as a basis the cost for Intel of those advantages. According to the Commission, the as-efficient-competitor analysis involves an assessment, in essence, of the compensation which a hypothetical as-efficient competitor would have to offer to Lenovo for the loss of Intel’s rebates. In order to be incentivised to choose the as-efficient competitor, Lenovo would expect to be compensated for its own losses, and not for Intel’s losses.

425Moreover, the Commission submits, relying on Annex B.31 to the defence, that Intel has adduced no sound evidence that there was a divergence of views between Lenovo and it on the value of the non-cash advantages granted to Lenovo, nor, a fortiori, that Lenovo calculated a different value for those non-cash advantages. It also states that documents contemporaneous with the facts set out in the file prove that Lenovo considered those non-cash advantages to be a significant help for it and that Lenovo had requested them from Intel as from the beginning of the negotiations.

426According to the Commission, it is incorrect to assert, as the applicant does, that the value of the non-cash advantage for the purposes of the as-efficient-competitor analysis is the cost to the dominant undertaking of those advantages. The reply attempts, according to the Commission, to circumvent the error which vitiates the reasoning in the application, claiming that ‘by definition, an as-efficient competitor could provide the same non-cash benefit to Lenovo at the same cost as Intel’. That ignores the fact that the as-efficient competitor is smaller than Intel. The Commission refers to recital1467 of the contested decision, which explained that the as-efficient competitor would normally not have a supply hub in China at that stage. It would therefore, according to the Commission, have to provide cash compensation for the loss of the advantages granted by Intel to Lenovo.

427On that point, the Commission contends that the reply merely responds, first, by stating that an as-efficient competitor would necessarily have a supply hub in China and, second, by alleging that AMD had one. In that regard, the Commission considers that the first response provided in the reply is a mere assertion. There is, according to the Commission, no reason why a competitor, even one who is as efficient, would necessarily have a supply hub in China. As regards the second response provided in the reply, the as-efficient-competitor analysis is concerned, according to the Commission, with a hypothetical competitor, not AMD. In any event, the document to which Intel refers merely states that AMD had ‘facilities’ in China, which does not demonstrate the existence of a supply hub, let alone one which is equivalent to that of Intel.

428According to the Commission, paragraphs22 to 37 of Annex D.39 to the rejoinder show that, even if it were accepted that the as-efficient competitor had a supply hub in China, its cost in making that hub available to Lenovo would be significantly higher than the cost for Intel of providing that advantage. The same is true of the warranty extension. If, as Intel claims, the cost of the two non-cash advantages amounts to USD3 million for Intel, then the provision of the same benefits to Lenovo would cost an as-efficient competitor at least USD38 million. That amount is calculated using two of Intel’s assumptions which the Commission disputes, namely, first, that the as-efficient competitor would have a supply hub in China and, second, that Intel’s cost in awarding the non-cash advantages was USD3 million.

429According to the Commission, in any event, the key argument in the application and in the reply that the cost of the two non-cash advantages for Intel is USD3 million misrepresents Intel’s own evidence. Paragraphs38 to 44 of Annex D.39 to the rejoinder demonstrate, according to the Commission, that contemporaneous internal Intel documents show that the applicant calculated that, in reality, the cost of the two non-cash advantages would be equal to, or even higher than, their value to Lenovo. Their cumulated cost for Intel was found to be USD47 million, and not USD3 million as Intel alleges.

430Before considering the parties’ arguments regarding the two non-cash advantages, it should be noted that the applicant does not deny having mentioned the figures of USD20 million for the warranty extension and USD24 million for the supply hub, in a presentation which it prepared for Lenovo. The applicant claims, however, that those figures ought to be replaced, for the purposes of the AEC test, by USD1.7 million and USD1.3 million for each of them, in order to reflect its costs and not the benefit to Lenovo. The Commission attributed USD44 million of the USD180 million of conditional discounts to the non-cash advantages on the basis of the value which those services represented for Lenovo. From a reading of recital1465 of the contested decision, it is apparent that it cannot be accepted that the Commission took into consideration, in that decision, Intel’s calculations estimating its costs of offering the USD3 million of non-cash advantages or that it analysed that figure.

431The Commission’s approach consists, in essence, of taking the view that, even if it is accepted that an as-efficient competitor may, in principle, offer non-cash advantages, the fact remains that making available a supply hub or a warranty extension is more expensive for the competitor than the dominant undertaking, in particular where the value of the non-cash advantages is compared with the contestable share. The Commission also maintains that Intel adduced no sound evidence that there was a divergence of views between Lenovo and Intel on the value of the non-cash advantages granted.

432The applicant criticises that analysis by the Commission. According to the applicant, the Supplemental Shapiro-Hayes report and the Salop-Hayes report demonstrate that that methodology is incorrect and that an appropriate analysis of the as-efficient competitor takes into account the cost to Intel of providing those non-cash advantages. The applicant refers to the Salop-Hayes report, which states as follows:

‘For the as-efficient-competitor test, the conditional discount would include the cost to Intel of providing those benefits, not the value to Lenovo. The focus of the as-efficient-competitor test is to determine whether Intel’s incremental revenues on the contestable share exceed its incremental costs of providing that volume, taking into account the reduction in Intel’s profits from the conditional discounts. The reduction in Intel’s profits involves the cost to Intel of the non-cash benefits.’

433In that regard, the foundation of the AEC test applied by the Commission in the present case is set out, in particular, in recitals1003 and 1004 of the contested decision.

434In recital1003 of the contested decision, the Commission explains the logic inherent in the AEC test, stating that ‘in essence, this examines whether Intel itself, in view of its own costs and the effect of the rebate, would be able to enter the market at a more limited scale without incurring losses’.

435In recital1004 of the contested decision, the Commission states that the as-efficient-competitor analysis is a purely hypothetical exercise in the sense that it attempts to analyse 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 entering. That analysis is, in principle, independent of whether or not AMD was actually able to enter.

436It follows from the foregoing that the hypothetical competitor whose ability to enter the market is assessed notwithstanding Intel’s pricing practices is an as-efficient competitor, namely an operator capable of supplying x86 CPUs under the same conditions as Intel. As is apparent from recital1003 of the contested decision, the AEC test amounts, in essence, to examining whether Intel itself could have entered the market despite the system of rebates at issue. It is apparent from recital1004 of that decision that, in principle, the only difference between the situation of the hypothetical competitor and the actual situation of Intel on the market is that that hypothetical competitor does not have an equivalent sales base. In the light of the clarifications provided in recital1005 of the contested decision, that reference to the lack of equivalent sales base must be interpreted as meaning that, due to Intel’s status as an unavoidable trading partner, the hypothetical as-efficient competitor is likely to win from Intel only the contestable share of customer requirements for x86 CPUs.

437As the applicant correctly observes, when the Commission assessed, in the contested decision, the value of the non-cash advantages offered by the applicant in the context of examining the breadth of the rebates granted to Lenovo, its reasoning was not conducted on the basis of the hypothetical competitor being capable of selling x86 CPUs to Lenovo while also offering non-cash advantages to Lenovo on the same terms as Intel.

438In recital1466 of the contested decision, the Commission expressed the view that it was a question of assessing the price that an as-efficient competitor, which is not the dominant company, would itself have had to pay to Lenovo to compensate for the loss of the non-cash advantages offered by Intel to Lenovo, such as an extension to the hub or a warranty extension. In recital1467 of that decision, in order to justify that approach, the Commission relied on the example of the supply hub. The Commission found that, unlike Intel, which had a supply hub in China, to which certain adaptations could be made in order to offer a non-cash advantage to Lenovo, a competitor as efficient as the dominant undertaking, but which was not dominant and which was therefore of smaller scale, would not normally have had such a facility at that stage.

439Thus, the Commission proceeded on the basis of an assumption which was contrary to the foundation of the AEC test set out in recitals1003 and 1004 of the contested decision, which is based on the principle that the hypothetical competitor is as efficient as Intel, in particular from the perspective of the costs of extending a hub or a warranty. The Commission in fact conducted its reasoning by reference to a less efficient competitor, which is not, however, the relevant economic operator for assessing whether the rebate at issue is capable of having a foreclosure effect.

440None of the arguments put forward by the Commission is such as to invalidate that finding.

441The Commission– which, admittedly, refers to the fact that Intel had evaluated the amount of benefit for Lenovo at a high value (USD20 million and USD24 million respectively)– provides no response, in the contested decision, to the question of what the cost would have been for an as-efficient competitor 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. The same logic applies to costs associated with a warranty extension.

442In that regard, the parties stated, in response to the questions put by the Court at the 2020 hearing, that economies of scale ought not to be taken into consideration as a differentiating factor, but that the costs of an as-efficient competitor ought to be regarded as being the same as for Intel. However, such explanations on the part of the Commission contradict the approach adopted in recitals1466 and 1467 of the contested decision, which take account of the size of the as-efficient competitor to emphasise, inter alia, that a hub comparable to that of Intel would not be in place at that stage.

443In addition, in so far as the Commission referred, before the Court, to the actual size of an as-efficient competitor’s hub (see paragraph426 above, in fine), it must be stated, as Intel submitted at the 2020 hearing, that that factor was not analysed in the contested decision. The same applies to the quantified assessments, submitted by the Commission for the first time in Annex D.39 to the rejoinder, aiming to evaluate the actual costs to Intel in relation to the non-cash advantages (see paragraphs429 and 430 above).

444The Court, however, cannot take account of those additional analyses, which were submitted during the procedure before it to substantiate the AEC test in the contested decision without substituting its own reasoning for that of the Commission set out in that decision. The case-law cited in paragraph150 above prohibits the Court from making such a substitution.

445As regards the Commission’s assertions in recital1464 of the contested decision, concerning the alleged stark discrepancy between the economic costs put forward by Intel for the grant of non-cash advantages and their value to Lenovo, it must be noted that, irrespective of the fact that the value to Lenovo is not decisive for the AEC test analysis, as is apparent from the minutes of a statement of 2June 2009 by L10, [confidential], Lenovo did not accept that the negotiations with Intel concerned an exact value for the non-cash advantages. L10 considered, in essence, that the approach calculated in USD in relation to those advantages could be radically different from the sum presented by Intel. In his view, in essence, that undertaking attempted to give itself credit for factors in respect of which it did not calculate a monetary value, such as distribution via a hub. Intel attempted to persuade it that those factors had an economic value, even though they were, rather, an operational benefit. L10 stated that he had not given any credit from a monetary perspective to those non-cash advantages. Lastly, as regards the Commission’s reference to the L10 email of 12January 2006, which acknowledged the importance of the non-cash advantages, it should be noted that they are not quantified in dollars in that email.

446It is also apparent from a series of emails dated 26November 2006 to 28November 2008, entitled ‘RE: Intel Meet Comp Response November 27 06.ppt’ that Intel’s negotiating tactics made various references to exaggerated advantages, in particular by positioning elements which it intended to provide to the business partner in any event as an advantage. In those circumstances, the Commission cannot draw the inference, even if only implicitly, as is the case in recital1464 of the contested decision, from those data alone concerning the negotiations on non-cash advantages, that the actual costs, as stated by Intel, were minimised. Similarly, it is necessary to reject as ineffective the Commission’s claim, referred to in paragraph614 of the defence and referring to Annex B.31 to the defence, that Intel had not demonstrated that there were divergences of views between Lenovo and itself regarding the value of the non-cash advantages granted. The question is what costs were necessary in order to offer them, and not Lenovo’s perception of their value.

447In addition, it is not sufficient to rely, as the Commission did in recital1464 of the contested decision and, subsequently, in paragraph614 of the defence, referring to paragraph416 of Annex B.31 to the defence, on the argument that Intel failed to explain the stark discrepancy between its alleged costs of USD3 million and the amount of USD44 million to Lenovo. It was for the Commission to assess, directly in the contested decision and not in calculations submitted for the first time before the Court, what the costs of an as-efficient competitor would have been if it had had to offer to an OEM such as Lenovo non-cash advantages equivalent to those proposed by Intel (see also paragraph444 above).

448Furthermore, in so far as the Commission carried out, for the first time before the Court, in paragraph326 of the rejoinder, referring, by way of illustration, to AnnexD.39 to the rejoinder, calculations of the costs for the situation in which account was to be taken of the as-efficient competitor having a supply hub in China, it must be noted that the result reached by the Commission regarding the costs differs, in any event, from the result set out in the contested decision, irrespective of the fact that those calculations were made out of time and do not form part of the reasons for the contested decision, which applied a different test. First, as is apparent from paragraph36 of Annex D.39 to the rejoinder, the cost to an as-efficient competitor is USD20690000, and not USD24 million as stated in recital1463 of the contested decision in relation to the supply hub. Second, as regards the warranty extension, the cost of which for an as-efficient competitor is also calculated for the first time at USD17473664 in paragraph30 of Annex D.39 to the rejoinder, that differs from the USD20 million stated in the contested decision.

449Lastly, the Commission’s assertion in paragraph327 of the rejoinder, referring to paragraphs38 to 44 of Annex D.39 to the rejoinder, that the applicant’s key argument that the cost of the two non-cash advantages, which amounted to USD3 million for Intel, contradicts Intel’s own evidence, cannot succeed.

450As regards Intel’s internal documents with references D.41 and D.42 annexed to the rejoinder, from which it is apparent, according to the Commission, that Intel estimated the costs of non-cash advantages at USD47 million rather than at USD3 million, these were not mentioned in the contested decision and do not therefore form part of the reasons for that decision. On reading recital1465 of the contested decision, it appears inconceivable that the Commission took them into consideration in its principal analysis as set out in that decision, in so far as it states in the decision that ‘Intel’s argument that the appropriate measure to be used in the as-efficient-competitor analysis is not the value of such items to Lenovo but their economic cost to Intel is based on a misunderstanding of the principles of the analysis’.

451In any event, even if the Commission’s reference to the documents mentioned in paragraph450 above had been admissible, it could not have been inferred from those documents that Intel had incorrectly minimised its costs by stating that the two non-cash advantages corresponded to USD1.7 million and USD1.3 million respectively. The documents to which the Commission refers relate to a situation in which negotiations with Lenovo were underway and in which Intel sought to demonstrate the significance of its commercial proposals, by presenting them in a favourable light to Lenovo (see also paragraphs445 and 446 above). A specific analysis of those documents, carried out subject to what has already been established, purely as a precautionary measure, demonstrates that they lack clarity and accordingly do not make it possible to uphold the Commission’s position.

452Accordingly, first, in the document entitled ‘Intel chart entitled “2006 vs. 2007 Trend”’, the advantage relating to the enlargement of the supply hub falls within the item ‘Incremental 07 Spending’ and a reference including the term ‘billing impact’ is made in that chart. However, the figure of USD24 million, concerning the hub, is included in the column entitled ‘Contra’ and not in the column entitled ‘Expense’. That indicates that what was involved was Intel’s estimate of the equivalent value of the use of the hub, as explained in paragraph71 of the Supplemental Shapiro-Hayes report and illustrated in Annex10 to that report, and not Intel’s costs of such a hub or of modifying it. Similarly, the cost of the warranty extension is calculated in paragraph70 of the Supplemental Shapiro-Hayes report and in Annex9 thereto at USD1.7 million. In those circumstances, it is not necessary to give a ruling on DrHayes’ allegations made at the 2020 hearing that, because of the limited number of x86 CPU failures, increasing the warranty from one to three years did not involve significant incremental costs.

453Second, while the tables set out in Annex D.42 to the rejoinder enable a link to be established between the costs for Intel and the advantages for Lenovo, those tables do not show the overall cost of modifying a supply hub, quantified in the contested decision at USD24 million. In any event, it cannot be ruled out that the aim of that document may have been to present the proposal in a favourable light during the negotiations with Lenovo.

454Accordingly, in the light of the errors of assessment made by the Commission, it is not necessary to assess certain additional Intel arguments regarding whether AMD actually had a hub in China, since AMD’s situation was not, in any event, decisive for the AEC test.

455Consequently, it must be concluded that the Commission erred in making a quantified assessment of the non-cash advantages offered by Intel to Lenovo, by using the amounts of USD20 million and USD24 million respectively, from which it estimated the amount of the rebates at USD180 million. That amount of USD180 million is therefore itself vitiated by an error.

456In the light of the foregoing, it must be observed that, in recital1507 of the contested decision, the Commission stated that the conclusions which it had reached as regards the capability of the rebates granted to Lenovo to have foreclosure effects were based on the comparison between the required number of units and the contestable number of units established in recital1478 of that decision as well as on the considerations in recitals1479 to 1506, which set out an alternative test for the required share on the combined segments of desktop computers and notebook computers. However, as is apparent from recitals1472, 1478 and 1503 to 1506 of the contested decision, in the context of both the comparison mentioned above and the alternative test, the Commission took into account a conditional portion of USD180 million for its analyses regarding the definition of the required share, for the purposes of comparing the required share with the contestable share of the x86 CPU units. The error in the quantified assessment of the non-cash advantages offered by Intel to Lenovo therefore affected all the component parts of the examination of the rebates granted to that OEM.

457Consequently, there being no need to assess the substance of Intel’s arguments regarding the contestable number of units to be taken into consideration, it must be found that the Commission has not established to the requisite legal standard the validity of the conclusion set out in recital1507 of the contested decision that in 2007 Intel’s rebates were capable of having or likely to have anticompetitive foreclosure effects, since even an as-efficient competitor would have been prevented from supplying Lenovo’s notebook x86 CPU requirements.

(e)The alleged errors in the AEC test applied to MSH

458The applicant submits that the AEC analysis concerning MSH in the contested decision, apart from the fact that it overstated Intel’s AAC, contains two errors concerning, first, the ‘double conditional rebate’ method (‘the double rebate method’) and, second, the conditional portion of the payments. In the applicant’s view, correcting one of those errors would demonstrate that MSH passed the AEC test.

459The Court considers it appropriate to examine first the substance of the line of argument aimed at demonstrating that the Commission erred in applying the double rebate method.

460The applicant disputes, in essence, the relevance of the figures used to apply that method and the inferences drawn from it by the Commission.

461For its part, the Commission considers that all the arguments must be rejected since no error vitiates the application of the double rebate method.

462The Commission asserts, first, that, in order to be able to sell computers of a particular brand to MSH, an as-efficient competitor would have to ensure not only that MSH was ready to purchase computers based on its CPUs, but also, and most importantly, that OEMs were ready to manufacture those computers. Therefore, in the Commission’s view, Intel’s practices at different levels of the supply chain could have a cumulative effect.

463Second, the Commission maintains that, in order to demonstrate that Intel’s payments to MSH were capable of having an anticompetitive foreclosure effect when taken together with Intel’s practice towards an OEM, it suffices to show that capability of effect by reference to a representative example of a conditional payment granted by Intel to one OEM without having to repeat the same exercise for each OEM.

464Third, the contested decision analyses the cumulation of Intel’s payments to MSH with Intel’s naked restrictions, in particular with respect to AMD-based x86 CPU Lenovo notebooks for the period from June to December 2006.

465Furthermore, the Commission maintains that Annex B.31 to the defence analyses Intel’s other arguments in detail. The Commission argues, in essence, that that annex established that (i) the contested decision provides proper justification that the rebates granted to NEC for the relevant quarter are representative of the entire relevant period; (ii) NECCI could not plausibly have provided the entire contestable share of MSH; and (iii) the contested decision does not rely on the assumption that 100% of Intel’s rebates to NEC were conditional.

466In that regard, it must be noted, as the applicant also does, that the Commission initially found, in recital1565 of the contested decision, that Table58 in recital1564 of that decision showed that, using the normal calculation method, Intel did not pass the AEC test for 1997, 1998 and 2000. As the applicant, in essence, submits, the Commission therefore acknowledged, at least implicitly, that, if the normal calculation method were used, the effective price resulting from Intel’s conditional payments to MSH was far above the AAC during the entire period of the alleged infringement, namely from 2002 to 2007.

467Subsequently, as is apparent from recitals1561 and 1566 of the contested decision, the Commission nevertheless adapted the AEC test by taking the view that, when Intel provided a conditional rebate to an OEM, an as-efficient competitor would have had to provide two payments: one to ensure that it captured the OEM’s contestable share and the other to ensure that it captured MSH’s contestable share. In taking account of that double rebate, the Commission came to the conclusion, in recital1568 of the contested decision, that Intel did not pass the AEC test during the entire period in question, except for 2004.

468It therefore follows from the recitals of the contested decision mentioned above that it is appropriate to proceed on the assumption that Intel passed the AEC test using the normal calculation method and that it was only by taking into consideration the existence of a double rebate that the Commission, using its own figures, was able to prove that Intel’s payments to MSH were capable of causing anticompetitive foreclosure during the entire period in question, except for 2004.

469As regards the assessment of those facts, it should be noted, as a preliminary point, that the applicant does not contest the double rebate method as such. The applicant acknowledges, in essence, that, in order to be able to sell computers of a particular brand to MSH, an as-efficient competitor had to ensure not only that MSH was ready to purchase computers based on its CPUs, but also, and most importantly, that OEMs were ready to manufacture those computers. Therefore, Intel’s practices at different levels of the supply chain could have had a cumulative effect.

470By contrast, the applicant disputes the figures used by the Commission to carry out its calculations. As the applicant states, the contested decision calculates the amount of the double rebate by assuming that each OEM which supplied MSH benefited from a conditional rebate equivalent to the total rebates offered to NEC in the fourth quarter of 2002 and that such OEMS would have lost the entirety of that rebate if MSH began selling computers with AMD-based x86 CPUs. On the assumption that 100% of the rebates provided to MSH were conditional, the Commission concluded that, for the whole of the period in question, except 2004, Intel’s rebates would have foreclosed an as-efficient competitor.

471The Court considers that that analysis contains two flaws, each of which is capable of invalidating the results of the AEC test for MSH based on the discounts granted by Intel to NEC in the fourth quarter of 2002.

472First, as the applicant submits, the Commission assumes, in recitals1566 and 1567 of the contested decision, that the rebates granted to NEC are an appropriate proxy for conditional rebates on all Intel-based computers sourced by MSH from any OEM. That assumption is not substantiated in any way.

473Intel states, without being contradicted by the Commission, that MSH purchased only 4% of its computer requirements from NEC during the period from 2002 to 2007 and that, apart from NEC, the main OEM suppliers of computers to MSH between 2002 and 2007 were Fujitsu, Acer, HP, Compaq, Toshiba and Medion. At the very least, the Commission’s position is necessarily based on the assumption that MSH purchased computers from OEMs other than NEC.

474However, the Commission does not claim, nor does 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, on conditions comparable to the rebates relating to computers purchased from NEC.

475It is therefore apparent that the contested decision based its analysis of the dual rebate method on the rebates granted by Intel to NEC over the course of a single quarter which represented only part of MSH’s purchases. Therefore, as the applicant submits, it must be held that the Commission’s assumption that all of MSH’s suppliers were subject to substantial conditional rebates identical to those enjoyed by NEC is unfounded and, in any event, entirely unsubstantiated.

476That conclusion is, moreover, borne out by the wording of recital1566 of the contested decision, in which the Commission simply states, in order to illustrate the double rebate method, that the ‘[Supplementary Statement of Objections of 2008] took the example of NEC as [an OEM representative of that situation]’, and by the wording of recital1567 of that decision, in which the Commission states that ‘section 4.2.3.4 assessed the Intel conditional rebates to NEC in the fourth quarter of 2002 (as that is the only quarter where sufficient data are available for the Commission to perform an analysis of the capability of the rebates to foreclose an as-efficient competitor)’. It therefore follows from those recitals of the contested decision that the Commission appears to have relied on the example of NEC and on one single quarter not simply because of its relevance, but because it was the only quarter for which it had been able to obtain information in order to carry out the AEC analysis for MSH.

477The Commission claims, in that regard, that it is sufficient to refer to a single representative example, since the AEC test is used solely to demonstrate the anticompetitive capability– and not the actual effects– of a commercial practice. However, the Court considers that, when the Commission chooses a quantitative approach to demonstrate that capability, it must ensure that the data used are reliable and must at least explain how such data may be extrapolated. The Commission, however, has in no way proven that NEC’s figures were ‘representative’ for all OEMs.

478Second, and in any event, as the applicant submits, the Commission’s analysis assumes that NEC and all the other OEM suppliers of MSH benefited, between 1997 and 2007, from conditional rebates identical to the rebate received by NEC for a single quarter. That implies, therefore, that, even if they were representative for all OEMs, the rebates granted to NEC for the fourth quarter of 2002 were stable over a 10-year period. However, first, the Commission has in no way proven that that was the case. The only justification upon which the Commission appears to rely is that set out in recital1567 of the contested decision, according to which the only data available to it were data concerning NEC’s rebates in the fourth quarter of 2002. However, as the applicant points out, the fact of not being able to obtain additional evidence does not permit the Commission to found its conclusions on assumed facts. Second, it must be recalled that, as is apparent from paragraphs404 to 411 above, it has been demonstrated that, as regards NEC, the Commission made an error of assessment in extrapolating the results which it had reached for the fourth quarter of 2002 for the whole of the alleged infringement period.

479Therefore, without there being any need to give a ruling on the other arguments put forward by the parties as referred to in paragraphs458 to 465 above, it must be held that the Commission erred in finding that Intel’s conditional rebates to NEC in the fourth quarter of 2002 constituted sufficient data to carry out the AEC test for MSH over the entire infringement period.

480Since the Commission has not demonstrated that the conditions for extrapolation were satisfied, it must therefore be held, without there being any need to give a ruling on the second argument concerning the conditional portion of the payments (see paragraph458 above), that the applicant is justified in maintaining that the application of the AEC test concerning MSH is vitiated by an error of assessment which applies to the entire period examined.

481In the light of the foregoing, the Commission has not established to the requisite legal standard the validity of the conclusion, set out in recital1573 of the contested decision, that, on the basis of the considerations set out in recitals1559 to 1572 of that decision, during the period from the last quarter of 1997 to 12February 2008, the Intel payments to MSH were capable of having or likely to have anticompetitive foreclosure effects, either in themselves or as a reinforcing factor of Intel’s conduct vis-à-vis other actors on the market, since even an as-efficient competitor would have been prevented from entering the relevant part of the market.

(f)Conclusions on the AEC test

482In the light of all the considerations set out in paragraphs179 to 480 above, there being no need at all to assess the applicant’s various claims concerning the cost analysis, the Court must accept the applicant’s argument that the AEC analysis carried out by the Commission in the contested decision is vitiated by errors.

C.The argument that the contested decision did not properly analyse and take account of the criteria referred to in paragraph139 of the judgment on the appeal

483According to the applicant and ACT, the Commission’s findings in the contested decision regarding the foreclosure capability of Intel’s rebates do not take proper account of all the criteria set by the Court of Justice in paragraph139 of the judgment on the appeal. The fact of failing to take account of even one of those criteria should, they argue, result in the General Court annulling the contested decision.

484The applicant and ACT maintain that, of those five criteria, at least three were not examined appropriately. They submit that, although the contested decision contains an analysis of the first and third criteria referred to in paragraph139 of the judgment on the appeal, namely the extent of Intel’s dominant position on the relevant market and the conditions and arrangements for granting Intel’s rebates, it fails, in any event, to analyse the criteria relating to the share of the market covered, the duration and amount of the rebates or the existence of a strategy designed to exclude from the market competitors that are at least as efficient as the dominant undertaking.

1.The coverage rate

485In recital1577 of the contested decision, in Section 4.2.4 which relates to the strategic importance of the OEMs which benefited from Intel’s rebates, the Commission stated, in essence, that, because of their market share, their strong presence in the more profitable segment of the market and their ability to legitimise a new processor in the market, some OEMs, in the present case Dell and HP, were strategically more important than others in providing x86 CPU manufacturers with access to the market. The Commission also took the view, in recital1597 of the contested decision, that the OEMs targeted by Intel’s conduct held a significant part of the market and that, moreover, they 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 coverage of the abusive practices had to be regarded as significant.

486The applicant and ACT submit, in essence, that, by simply stating in recital1597 of the contested decision that the OEMs targeted by Intel’s conduct held a significant part of the market and that they were strategically the most important, which had a more significant impact on the overall market than would have corresponded to their aggregate market share alone, the contested decision did not properly take account of the criterion of the coverage rate for the purpose of analysing whether Intel’s rebates and payments were capable of leading to foreclosure.

487Furthermore, the applicant states that that finding in the contested decision was set out following the conclusion, in recital1001 of that decision, that Intel’s rebates and payments fulfilled the criteria to be regarded as abusive, whereas the judgment on the appeal requires that the Commission analyse the market coverage before making a finding of abuse. In addition, the applicant and ACT argue that the evidence relied on by the Commission was not sufficient for the view to be taken that the market share covered by Intel’s conduct was significant.

488The Commission disputes the substance of the arguments put forward by the applicant.

489First, the Commission observes that market coverage was considered in Section4.2.4, of the contested decision, in the context of the strategic importance of the OEMs which received rebates from Intel. The Commission dwells in particular on the fact that, although paragraph139 of the judgment on the appeal simply identifies market coverage as a factor, this must be applied in the context of each case and, in the present case, the strategic importance of the part of the market covered is to be taken into consideration in assessing that factor as demonstrating the capability of Intel’s loyalty rebates to foreclose competition. Similarly, it is necessary to take account of the fact that Intel was an unavoidable trading partner for the OEMs, which gave Intel significant leverage over its customers, since it would have been unrealistic for them to switch to an all AMD or majority AMD product line-up.

490Second, the Commission maintains that, as regards the share of the market covered, the applicant no longer relies on the assertion, set out in paragraph115 of the application, that the market coverage of Intel’s practices did not exceed 2% in any year, but appears to accept the fact that the Court found, in paragraph194 of the initial judgment, that market coverage was approximately 14% on average during the period in which the infringement was committed and submits that it could be inferred from certain evidence that the market shares of the OEMs concerned by the rebates at issue exceeded 25%.

491Third, as regards the complaint raised by Intel in its observations that the finding in recital1597 of the contested decision was set out after the conclusion had been reached, in recital1001 of that decision, that those rebates and payments fulfilled the criteria to be regarded as abusive (see paragraph487 above), the Commission takes the view that the applicant mischaracterises the contested decision. Recital1001 of that decision relies on the judgment of 13February 1979, Hoffmann-La Roche v Commission (85/76, EU:C:1979:36), according to which loyalty rebates infringe Article102 TFEU. However, as set out in recital1597 of the contested decision, the Commission states that the subsequent analysis shows that targeting such strategically important OEMs has a more significant impact on the overall market than would correspond to their aggregate market share alone. Consequently, the coverage of the abusive practices has to be regarded as ‘significant’ and recital1616 of the contested decision reaches the overall conclusion that the loyalty induced by the rebates had complementary effects which significantly diminished the ability of other competitors to compete and sell their products on the merits of their x86 CPUs.

492It should be recalled that it is apparent from paragraph139 of the judgment on the appeal that the share of the market covered by the contested practice is one of the criteria which the Commission must take into account for the purposes of assessing the foreclosure capability of rebates and conditional payments (see paragraphs119 and 125 above).

493In the first place, in the circumstances of the present case, it cannot be ruled out that Section 4.2.4 of the contested decision, concerning the strategic importance of the OEMs which benefited from Intel’s rebates, might be relevant in examining the coverage rate. That section deals with certain factors which are a priori relevant to examining the foreclosure capability of a system of rebates, such as targeting certain pricing practices at the most profitable segments of the market or using, to the detriment of a competitor, the largest market operators’ power to legitimise a product.

494The fact remains that, contrary to the Commission’s assertions, and regardless of whether the finding in recital1597 of the contested decision was set out after the conclusion had been reached in recital1001 of that decision that Intel’s rebates and payments fulfilled the criteria to be regarded as abusive, the content of Section4.2.4 of the contested decision, concerning the strategic importance of OEMs which benefited from Intel’s rebates, and in particular recital1597 of that decision upon which the Commission relies in order to find that the share of the market covered had been examined, cannot be interpreted as constituting in itself sufficient examination, in the circumstances of the present case, of the share of the market covered by the contested practice, within the meaning of paragraph139 of the judgment on the appeal.

495Irrespective of the fact that the Commission relied on the market shares of certain OEMs and on the assumption that the Commission could legitimately confine itself to relying on the market shares of certain OEMs rather than examining the share of the market covered by the contested practice, as mentioned in paragraph139 of the judgment on the appeal, recitals1578 to 1580 of the contested decision take account solely of Dell’s and HP’s market shares, and exclude the other OEMs concerned by the contested practice, as pointed out by the applicant and ACT. It should be added that the market shares taken into account cover only the period from the first quarter of 2003 to the final quarter of 2005. Therefore, not only does that decision cover only one part of the entire period covered by that decision, that is to say, from October 2002 to December 2007, but it also disregards the period from 2006 to 2007, during which Lenovo and MSH were concerned. Lastly, as the applicant and ACT point out, it is apparent from recitals1578 to 1580 of the contested decision that 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 practice as regards HP relates to corporate desktops, as stated in Article1(b) of the contested decision.

496In the second place, in its main observations, the Commission relies on the Court’s finding, in paragraph194 of the initial judgment, that the market coverage was approximately 14% on average during the period of the infringement and submits that it could be inferred from certain evidence that the market shares of the OEMs concerned by the rebates at issue exceeded 25%. The Commission also states that ‘Intel’s complaint… that the Commission relied on HP’s market share across all segments is unfounded; the [contested] decision does not rely on any specific figure as regards HP and the [initial] judgment’s reliance on the average coverage of 14%, contrary to Intel’s assertion… does not take account of the segment-specific loyalty rebate with HP’.

497However, it is necessary to reject the Commission’s argument based on the fact that the Court found, in paragraph194 of the initial judgment, that market coverage was approximately 14% on average during the period in which the infringement was committed, which the applicant did not dispute, or that the market shares of the OEMs concerned by the rebates exceeded 25%.

498It must be stated that the rates of 14% or 25% do not appear anywhere in the contested decision subsequent to an examination of the coverage rate. Consequently, for the purposes of reviewing the legality of the contested decision, as regards the share of the market covered by the contested practice, the Court is not in a position to place reliance on those rates, even if they are derived from evidence in the file, since they were not included in the contested decision and, by definition, the Commission could not have acted on the basis of that evidence.

499Consequently, there being no need to give a ruling on the Commission’s arguments regarding HP’s market share, it must be held that the Commission failed to determine the share of the market covered by the practice at issue, contrary to the requirement placed on it pursuant to paragraph139 of the judgment on the appeal. It should be added that that is, moreover, contrary to the Commission’s own guidelines on the analysis of cases falling within the scope of Article102 TFEU, and in particular contrary to paragraph20 of the Guidance on the Commission’s enforcement priorities in applying Article [102 TFEU] to abusive exclusionary conduct by dominant undertakings (OJ 2009 C45, p.7).

500In the light of the foregoing, it must therefore be concluded that the applicant and ACT are correct in maintaining that the contested decision is vitiated by errors in that it did not consider properly the criterion relating to the share of the market covered by the contested practice.

2.Duration and amount of the rebates

501In the application and in its main observations, Intel criticises the failure to analyse, in the contested decision, the importance, first, of the duration of the conditional rebates and payments offered and, second, of the amounts of those rebates and payments. Intel maintains, in particular, that it is not possible, in order to assess the foreclosure effects of the rebates at issue on an as-efficient competitor, to accumulate short-term agreements concluded with the OEMs and MSH. In its view, in order to do so, it is necessary to take into consideration the duration of each of those agreements.

502The Commission submits that the conditions for granting the rebates and various payments granted by Intel were analysed, for each OEM, in SectionVII.4.2.2 of the contested decision. Those analyses concerned the nature and modus operandi of the exclusivity or quasi-exclusivity conditions to which the payments and rebates were subject, the amounts of the rebates and, lastly, the decisive nature of the conditionality of the payments and rebates for each of the OEMs, and for MSH, when they assessed the possibility of obtaining part of their supplies of x86 CPUs from AMD. In particular, the Commission states in the defence that the brevity of the notice of termination period for certain agreements, including those with HP, did not change the harmful effects on competition. Accordingly, if Intel had terminated the HPA agreements following contravention by HP of its quasi-exclusivity obligation, HP would have lost the rebates over the entire remaining duration of the agreement and, at least potentially, over the duration of the extension of that agreement.

503In its main observations, the Commission submits, in essence, that the applicant has not challenged the passage in the initial judgment which is relevant in the light of the judgment on the appeal, namely the findings in paragraph195 of the initial judgment, in which the Court considered the implications of the duration of the rebate agreements for their foreclosure capability. The Commission therefore submits that it is necessary to regard as definitive the findings in paragraph195 of the initial judgment, to the effect that the duration of Intel’s agreements did not impair their ability to foreclose competition.

504The Commission, in its main observations, also submits, with regard to paragraph195 of the initial judgment, that, even if Intel were permitted to renew its challenge to the contested decision by reference to the assessment of the significance of the duration of its agreements, there is no reason to depart from the conclusion reached in the initial judgment. First, according to the Commission, if, as found in the contested decision, Intel fails the AEC test, Intel’s insistence on the OEMs being able to withdraw from the loyalty rebate agreements is illogical. An as-efficient competitor simply could not compete. Second, even if Intel were to pass the AEC test, it is inherent to an as-efficient competitor’s bid for the OEMs’ business in such circumstances that it would have to accept a much lower level of profitability on those sales than Intel. Third, the Commission reiterates the allegation that the overall duration of an Intel loyalty rebate scheme would be a factor in how long an as-efficient competitor would have to accept reduced profitability in ‘capturing an OEM’s custom from Intel’ on those sales. Accordingly, for HP, any competitor wishing to displace Intel would have to be prepared to offer terms which would offset the loss of Intel’s rebates over at least the full term of the HPA1 agreement. Moreover, the Commission maintains that each set of agreements with the OEMs ran for a period sufficient for Intel’s actions to be capable of foreclosing competition, since those agreements focused on the most profitable periods for sales of x86 CPU processors, early in the life cycle of a new design. The Commission also asserts that the duration of Intel’s practices cannot be isolated from their timing, since they sought to overcome Intel’s inability to produce a timely technical response to AMD’s 64-bit x86 CPUs.

505During the 2020 hearing, the Commission submitted a document to the Court concerning the recitals of the contested decision which, in its view, evaluated the various criteria as set out in paragraph139 of the judgment on the appeal, including duration.

506In the first place, it is necessary to reject the argument raised by the Commission in its main observations that the complaints regarding the duration and amount of the conditional rebates and payments are inadmissible. It is sufficient to note that the applicant’s arguments put forward in its main and supplementary observations in that regard relate clearly to those put forward in paragraphs102 and 111 to 114 of the application. Consequently, in accordance with the case-law cited in paragraph106 above, those complaints are admissible.

507In the second place, it is apparent from paragraph139 of the judgment on the appeal that the analysis of the duration and amount of the conditional rebates and payments, which are the object of the contested practice, is one of the criteria which must be taken into account in assessing the foreclosure capability of those practices.

508First, it is true that, in the contested decision, the Commission examined, on several occasions, factors relating to the duration of the rebates.

509First of all, recitals1013 to 1035 of the contested decision concern the time horizon for the AEC test. The Commission found, in particular, in recitals1015 and 1017 of the contested decision, that, in certain circumstances, there could be quarterly adjustments to rebate practices and stated, in recitals1017 to 1028 of the contested decision, that, because of the fact that the relevant market was very fast-moving, innovation in the relevant sector made it difficult or even impossible to make long-term predictions. Similarly, recitals1025 to 1027 of the contested decision contain a reference to the duration of the contracts and to the fact that it was necessary regularly to ‘refresh’ product cycles.

510Next, recitals201 and 202 of the contested decision indicate that the Commission considered that some of the relevant negotiations between Intel and the OEMs were conducted on a quarterly basis. Those negotiations also concerned a relatively short period, which could allow an as-efficient competitor to offer its own x86 CPUs to those OEMs more easily. Similarly, in recitals965 to 968 of the contested decision, the Commission examined Intel’s argument that the 30-day termination notice provision in the HPA agreements gave HP greater freedom to compare its offers with those of AMD, and indicated that Intel’s status as an unavoidable trading partner and the effects of its rebates resulted in that argument being rejected. At the 2020 hearing, the Commission stated that, in some cases, as regards HP, there had been several renewals of the agreements with Intel on a monthly basis. As regards Dell, the Commission stated, in recital1227 of the contested decision, that, due to the absence of any written contract with Intel, as regards the rebates granted under the Meet Competition Programme, those rebates were the subject of ‘constant’ oral renegotiation, such that Intel had great flexibility to change discounts.

511The fact remains that, first, the purpose of recitals1013 to 1035 of the contested decision was solely to define the time horizon 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 rebates for each of the OEMs concerned. The Commission concluded that, in relation to the AEC test, it would proceed on the assumption that the relevant time horizon was one year.

512Consequently, the time horizon was used in that regard 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 rebates at issue could have a foreclosure effect. Such an examination does not therefore constitute an analysis of the duration of the rebates as a factor which is capable, in itself, of demonstrating their ability to have a foreclosure effect.

513Second, it is apparent from recitals201, 202, 965 to 968 and 1227 of the contested decision that the Commission examined the duration and form of the OEMs’ commitments to Intel giving 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.

514However, although those aspects of the time horizon appeared to it to be relevant, the Commission examined them only in a haphazard and limited manner, in recitals201, 202, 965 to 968 and 1227 of the contested decision. It did not carry out a thorough and exhaustive examination for all OEMs of those aspects in so far as they were capable of determining or strengthening the capability of Intel’s pricing practices at issue to have a foreclosure effect.

515It is apparent from the foregoing that the Commission did not examine the duration of the rebates as a factor intrinsically relevant to determining the capability of Intel’s pricing practices at issue to have a foreclosure effect.

516Second, the Commission asserts, in essence, that, even if the AEC test did not demonstrate the capability of the rebates at issue to have a foreclosure effect, it is the total period during which the applicant applied the rebates and exclusivity payments to OEMs which should 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 must 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. In the Commission’s view, those considerations are apparent from paragraphs93 and 195 of the initial judgment and are therefore definitive.

517In that regard, first, it is apparent from paragraph81 above that the operative part of the judgment on the appeal sets aside the entirety of the initial judgment. Consequently, the Court must, following the referral back, carry out a fresh examination of the parties’ arguments concerning the duration of the rebates, without being bound by paragraphs93 and 195 of the initial judgment, which it does not accept for its own account.

518Second, it is apparent from paragraphs138 and 139 of the judgment on the appeal that, in the case where the undertaking concerned submits, during the administrative procedure, on the basis of supporting evidence, that its conduct was not capable of restricting competition and, in particular, was not capable of producing the alleged foreclosure effects, the Commission is required to assess all of the criteria referred to in paragraph139 of that judgment, and not solely the criterion relating to the duration of the rebates set out in that paragraph. Accordingly, the sole reference to the period during which the rebates were granted to the OEMs and MSH is not sufficient, in itself, notwithstanding the conclusions which may be drawn from the AEC test, to justify definitive conclusions as to the foreclosure effects thereby produced.

519Third, the Commission cannot succeed in its assertion that the duration of Intel’s practices cannot be isolated from their timing, since they sought to overcome Intel’s inability to produce a timely technical response to AMD’s 64-bit x86 CPUs. For the same reasons as those set out in paragraph518 above, that argument, even if it appears as such in the contested decision, is not in itself sufficient to justify definitive findings as to the foreclosure effects thereby produced.

520There being no need to rule on the applicant’s arguments concerning the amounts of the rebates, it is apparent from the foregoing that the Commission erred in failing to examine, in the contested decision, the duration of the rebates as evidence making it possible to determine the capacity of Intel’s pricing practices at issue to have a foreclosure effect.

3.Conclusions on taking account of the criteria mentioned in paragraph139 of the judgment on the appeal

521In the light of all the considerations set out in paragraphs485 to 520 above, there being no need to analyse the applicant’s complaints regarding the criteria concerning the amount of the rebates and the strategy aiming to exclude competitors from the market, it must be concluded that the applicant is justified in maintaining that the Commission’s analysis in the contested decision of the criteria mentioned in paragraph139 of the judgment on the appeal is vitiated by a number of errors. The Commission did not consider properly the criterion relating to the share of the market covered by the contested practice and did not analyse correctly the duration of the rebates.

D.Conclusion on the application for annulment of the contested decision

522It is apparent from paragraphs124 to 126 above that, although a system of rebates set up by an undertaking in a dominant position on the market may be characterised as a restriction of competition, since, given its nature, it may be presumed to have restrictive effects on competition, the fact remains that what is involved is, in that regard, a mere presumption and not a per se infringement of Article102 TFEU, which would relieve the Commission in all cases of the obligation to examine whether there were anticompetitive effects. Where an undertaking in a dominant position submits, during the administrative procedure, on the basis of supporting evidence, that its conduct was not capable of restricting competition and, in particular, was not capable of producing the alleged foreclosure effects, the Commission must analyse the foreclosure capability of the system of rebates by applying the five criteria listed in paragraph139 of the judgment on the appeal. In addition, where the Commission has carried out an AEC test, that test is one of the factors which the Commission must take into account in assessing whether the system of rebates is capable of restricting competition.

523In the present case, the applicant submitted, during the administrative procedure, on the basis of supporting evidence, that the rebates at issue were not capable of producing the alleged foreclosure effects. In recitals1002 to 1573 of the contested decision, the Commission carried out an AEC test and, in the light of the results of that test, concluded, in recitals1574 and 1575 of that decision, that Intel’s rebates and payments at issue were capable of having or likely to have anticompetitive foreclosure effects, since even an as-efficient competitor 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.

524However, it follows from all of the foregoing that, first, the AEC test carried out in the contested decision is vitiated by errors and, second, as regards the criteria mentioned in paragraph139 of the judgment on the appeal, the Commission did not consider properly the criterion relating to the share of the market covered by the contested practice and did not analyse correctly the duration of the rebates.

525It should be noted, as regards the rebates granted to HP, that it has been held, in paragraph334 above, that the Commission did not establish to the requisite legal standard its finding that, during the period from November 2002 to May 2005, the rebate which Intel granted to HP was capable of having or was likely to have an anticompetitive foreclosure effect, since it has not demonstrated that those effects were present for the period between 1November 2002 and 31September 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, that could not demonstrate to the requisite legal standard the foreclosure effect of the rebates granted to HP, since the Commission did not consider properly the criterion relating to the share of the market covered by the contested practice and did not analyse correctly the duration of the rebates.

526Therefore, the Commission is not in a position to determine that the applicant’s rebates and payments at issue were capable of having or likely to have anticompetitive effects and that they therefore constituted an infringement of Article102 TFEU.

527Consequently, the Court considers that the grounds of the contested decision are not capable of serving as a basis for Article1(a) to (e) of that decision.

528Furthermore, in reply to a question from the Court of 2April 2012, seeking to ascertain, as regards a possible adjustment to the amount of the fine should the contested decision be the subject of annulment in part, the relative value of the infringements consisting of the exclusivity payments compared with the infringements consisting of the naked restrictions, the Commission, in a reply lodged on 8May 2012, replied solely in relation to the gravity of the infringements, arguing that it had assessed the conduct in question as a whole and had come to the view that those infringements complemented and mutually reinforced one another.

529Since the Court is not in a position to identify the amount of the fine relating solely to the naked restrictions, Article2 of the contested decision must therefore also be annulled.

530Article3 of the contested decision must be annulled in so far as it concerns the exclusivity rebates.

531The action is dismissed as to the remainder, having regard, in particular, to the findings in the initial judgment which the Court accepts, as set out in paragraphs96 to 98 above.

The head of claim seeking the annulment of the fine or a reduction in the amount of the fine

532In the light of the foregoing, it is not necessary to give a ruling on the second head of claim, put forward in the alternative.