Competition Stories July – September 2022

Welcome to the Competition Stories – an exploration of recent courts and competition law agencies’ decisions. Authored by Makis Komninos, a renowned expert in the field. This column aims to go through the latest and most important developments in competition law in recent months. We call them “stories” because Makis has promised to include some anecdotes from time to time, and not just stay at the black letter. Enjoy!


With much delay, in part due to my workload, I am back. This time my “stories” cover three months: July, August, and September 2022. In no other period have the EU Courts been as active in competition law, notwithstanding the summer recess. We had a number of very important AG Opinions, including in Lithuanian Railways, Unilever Italia, Facebook, and of course the General Court’s Illumina/Grail and Google Android judgments!

I. Court of Justice Judgements

The second half of 2022 started in a somewhat undistinguished manner when it comes to Court of Justice judgments. F. Hoffmann-La Roche was an unconventional ruling and in reality, not a competition ruling as such, but rather one relating to national remedies. The case relates to the previous (synonymous) preliminary reference ruling of 2018, which dealt with the Italian Competition Authority’s (ICA) decision in 2014, which found that Hoffmann-La Roche and Novartis had concluded an anti-competitive agreement designed to achieve an artificial differentiation between Avastin and Lucentis, two pharmaceutical products. The two pharmaceutical companies disagreed with the referring court’s judgment (see here) on a factual point and on the way it applied the 2018 ruling and urged it to revisit its judgment, although this was not allowed under the national procedural rules. Italy’s Supreme Administrative Court (the Consiglio di Stato) – somewhat overzealously – decided to turn to the Court of Justice once again and asked whether EU law provided an exceptional legal basis that required a revision of its final judgment and under what conditions. The Court of Justice decided to proceed to a ruling without an AG’s Opinion and without an oral hearing, which already gives you an idea… For the Court, “EU law does not have the effect of requiring Member States to establish remedies other than those established by national law, unless it is apparent from the overall scheme of the national legal system in question that no legal remedy exists that would make it possible to ensure, even indirectly, respect for the rights that individuals derive from EU law, or that the sole means of obtaining access to a court is effectively for individuals to break the law” (para. 47). In the case at hand, there was nothing to suggest that Italian law had the effect of making it impossible or excessively difficult to exercise rights conferred on individuals by EU law. Besides, the Court held, it was open to a Member State to “restrict the possibility of seeking revision of a judgment of its supreme administrative court to exceptional and strictly limited situations, which do not include a situation where, according to the litigant who has been unsuccessful before that supreme administrative court, the latter has infringed the interpretation of EU law provided by the Court in response to its request for a preliminary ruling” (para. 50). The Court then relied on Article 267 TFEU to clarify that the preliminary reference procedure meant that the referring court has to respect the interpretation of EU law, as given by the Court of Justice. However, this did not mean that the Court of Justice should be overwhelmed (my words) with references about whether the referring court did or did not respect the previous preliminary ruling. And this also meant that EU law should not require a national supreme court to allow a revision proceeding merely in order to send a new reference to Luxembourg. One can clearly sense a degree of irritation here…

Landkreis Northeim was another unconventional and rather unimportant ruling. It had to do with the interpretation of the scope of the Commission’s trucks cartel decision. The Court held that the decision also covered specialised vehicles, including garbage trucks. This means that individuals (in the case at hand: a local authority) that bought such trucks could bring a follow-on claim for damages. The Court held that the defendants’ reliance on secondary EU legislation that referred to “trucks” and “special purpose vehicles” was “irrelevant” (para. 39). What counted was the operative part and the reasoning of the decision. The interesting part of the story is that the Commission, prior to adopting the decision, had sent a request for information (RFI) to the companies concerned in order to get their turnover figures with a view to imposing fines. In that RFI, the Commission excluded military and firefighting trucks which led the defendants to believe that specialised trucks were excluded from the scope of the decision. The Court, however, considered the RFI again as irrelevant. It held that “the purpose of such a request for information is not to define or specify the products covered by the anti-competitive conduct” (para. 56). What happened is that the Commission decided to depart from its Fining Guidelines and calculate the basic amount of the fine “without having recourse to the full value of the undertakings’ sales of trucks in the EEA” (decision, para. 112). It excluded the sales of some specialised vehicles, which, however, according to the Court, did not mean that garbage trucks were not within the scope of the decision. So quite a curious case.

II. Advocates General Opinions

Let me start with another ground-breaking Opinion by AG Rantos, the “hero” of Servizio Elettrico Nazionale (see my stories of November-December 2021). This is his Lithuanian Railways Opinion. It relates to the appeal from the General Court’s judgment upholding the Commission’s decision in the Baltic Rail case (for the rather extraordinary facts, see my stories of November-December 2020). The AG saw this as an opportunity to clarify the limits of the refusal to supply case law and to provide for certain guiding principles. His Opinion, as we will see in one of my upcoming “stories”, proved very influential for the Court’s judgment (be a bit patient). AG Rantos says that the General Court did not err in treating the removal of the track as an independent and distinct form of abuse and not as a refusal of access. This was so for a number of reasons, but perhaps the most important one is that the conduct in question is not so much about the refusal to grant access as it is really about the destruction of own infrastructure for exclusionary aims and follows the “logic of predation” (paras 78-79). In other words, the AG relies on a sort of “no economic sense” test. Therefore, the case was different from the refusal to supply cases. I think the Opinion is a masterly exposition of the refusal to supply case law and the rationale of the so-called Bronner criteria and will stay as one of the foundational Opinions in this area and in Article 102 TFEU generally.

A few days later, the busy AG delivered another leading Opinion: his Unilever Italia Opinion in another preliminary reference coming from the Consiglio di Stato on appeal from an ICA infringement decision. In reality, there were two main questions at stake. One is about the concept of a single economic unit, which the ICA used in a rather curious way in its infringement decision, considering that Unilever and its distributors were a single economic unit. The second question was about the legal test in exclusive dealing cases. In short, does Intel apply only to exclusivity rebates or does it also apply to exclusive dealing? For me, this was obvious from the very text of Intel, but anyway… Interestingly, it was never doubted that the ICA had to look at all circumstances and at the parameters mentioned in para. 139 of Intel. This was undisputed, and actually, the ICA argued that it had run such an analysis. What was disputed is whether the ICA ought to take into account certain economic studies prepared by Unilever, which were inspired by the AEC test. In particular, Unilever had relied on an “efficient breach of contract” analysis done by their economists, which sought to attribute a value to each contract and hence check whether the competitors could match that value. For the AG, the approach has to be similar and consistent across the board: economic arguments should be seriously taken into account in all exclusionary abuses, including both exclusivity rebates and exclusive dealing. He believed that, not only as a matter of substantive legal test (Intel) but also as a matter of due process, a competition authority is required to take such arguments into account and address them. Contrary to his Opinion in Servizio Elettrico Nazionale, AG Rantos’s Unilever Italia Opinion is much less “academic” and does not read like an article. Maybe the AG thought that too much academic analysis might confuse the Court, which might have happened to a degree in Servizio Elettrico Nazionale (see my stories of May-June 2022). Perhaps, this explains why this time the Court followed the AG entirely on the substantive question, as we will see soon in one of my next stories (again, be patient). Clearly, another very thoughtful Opinion by AG Rantos. When it comes to the other major question, the application of the concept of single economic entity to this case, the AG proposed to confirm the ICA’s theory that a dominant company and the distributors it tightly controls can form a “single economic unit” and, therefore, Article 102 liability can be attributed to the dominant company just on the basis of its distributors’ conduct. This will be the case “where, having regard to the economic, organisational and legal links between the producer and its distributors, the producer exercises decisive influence on the distributors, so that they consider themselves compelled to reproduce the conduct conceived and implemented by the producer, as they are unable to act independently on the market. That would be the case, in particular, where, by virtue of that contractual coordination, the distributors, first, bear none of the financial risks linked with the sales of the producer’s product or, secondly, enter into exclusive contracts with the producer” (para. 55). As we will see, this is a point where the Court did not exactly follow the AG… In a way, the AG himself had hinted that this might indeed be a false question in para. 48 of his Opinion. To be continued…

The July 2022 trinity of AG Rantos Opinions was completed with his American Airlines Opinion. This relates to an old clearance decision of 2013, where the Commission had cleared the merger between US Airways and American Airlines subject to the release of a daily slot pair at London Heathrow and to a few other remedies. The question was whether the General Court erred in upholding a 2018 Commission decision approving Delta as the buyer of grandfathering rights of slots. AG Rantos found that the General Court made no legal error and duly examined the objectives of the Merger Regulation and of the remedies in merger cases. I will stop here because this is quite a fact-specific case and there were specific arguments about the text that the Commission had used in its 2018 decision. So maybe your time would be better spent focusing on other developments.

Next, we have AG Pitruzzella’s French Supermarkets Opinion (in Case C-682/20 P, Les Mousquetaires and ITM Entreprises SAS (Intermarché) v Commission) – full disclosure: my firm is involved in this case. This relates to dawn raids conducted by the Commission in the context of an investigation against the Intermarché and Casino supermarkets. Procedurally, it is quite complicated. There are three cases (Cases C-682/20 P, C-690/20 P, and C-693/20 P) and two Opinions (in Cases C-682/20 P and C-690/20 P). There were three appeals against the Commission’s decisions ordering the dawn raids (by the two supermarkets and their joint purchasing alliance). The General Court, in Cases T-255/17, T-249/17, and T-254/17 (respectively corresponding to the numerical order of the above appeals), held that there was not enough evidence for the Commission to order a dawn raid with regard to one of the two alleged infringements and partially annulled the decisions. On further appeal to the Court of Justice, the investigated parties contested the General Court’s findings to the extent that it confirmed part of the decisions. Interestingly, the three appeals were not formally joined, although they were heard at the same time and jointly. In reality, the main Opinion accompanies all three cases, notwithstanding that it is formally attached only to Case C-682/20 P. The shorter Opinion of the same date by AG Pitruzzella, which is attached to Case C-690/20 P, relates only to a narrower legal question and I am not going to bother about it. The readers should concentrate on the main Opinion. Interestingly, the AG proposed to set the General Court’s judgment aside (in part) and to annul the Commission’s decisions to conduct the dawn raids. The issue at stake was that the Commission’s decisions were vitiated by a procedural irregularity because the Commission, to launch the dawn raids, had relied on interviews with supermarket suppliers that were not correctly recorded. Indeed, it appeared that some minutes looked exactly the same and some were drafted after the raids. The Commission’s failure to record the interviews properly means that the information obtained in these interviews could not be used to adopt the decisions ordering the dawn raids, AG Pitruzzella concluded. As we will see in a few weeks, when I catch up with my 2023 stories, the Court of Justice followed the AG. This shows how far the case law of the Court has gone in this area!

In July 2022, we had two Opinions in Servier, delivered by AG Kokott (more specifically, one in Case C-176/19 P, which is the Commission’s appeal, and another in Case C-201/19 P, which is Servier’s appeal) – again, full disclosure: my firm is involved in this case. This is one of the long-standing cases that were opened after the 2009 pharma sector inquiry. The Commission adopted an infringement decision in 2014, in which it imposed a hefty fine on Servier and a number of generic manufacturers for concluding a series of agreements (mostly patent settlements) aimed at protecting Servier’s bestselling blood pressure medicine, perindopril, from price competition by generics. The Commission found violations of both Articles 101 and 102. As a result, there were a number of appeals to the General Court. In its 2018 judgment, the General Court rejected Servier’s appeal and confirmed the Commission’s decision, but only partially. It also annulled the decision for the parts (i) that Servier’s patent settlement agreements with generic manufacturer Krka were anti-competitive and (ii) that Servier had also abused its dominant position. For the General Court, the Commission had wrongly defined the relevant market, and, therefore, the Article 102 finding collapsed. AG Kokott disagrees with the latter findings of the General Court and thinks that the Commission’s decision should stand in its entirety. First, the AG thought that the General Court erred in law when it took the view that the Servier-Krka agreements, comprising, inter alia, a patent settlement agreement and a technology licensing agreement, were based on the recognition of the validity of the patent in dispute by Krka and not on a reverse payment scheme. The AG sided with the Commission in taking the view that those agreements constituted a restriction of competition by object because the licence constituted a transfer of significant value by Servier in favour of Krka. For the AG, there was no consideration from Krka but rather a commitment not to compete with Servier on EU markets not covered by the licensing agreement. Furthermore, the AG thought that the General Court had erred in finding that those agreements had no anti-competitive effect insofar as the Commission had established that they did have the effect of eliminating Krka as a potential competitor of Servier. Finally, the AG also criticised the General Court for its annulment of Article 102-related findings and disagreed with it on the question of the definition of the relevant market. All in all, the General Court is wrong in all its findings against the Commission and the Commission did everything by the book. Those of us who are familiar with AG Kokott’s Opinions are not surprised, since she has established herself over the last twenty years as the deacon of a stricter approach to EU competition law.

In September 2022, we also had an interesting Opinion in Repsol, by AG Pitruzzella. This was a private enforcement case (again), in a preliminary reference from Spain (again), dealing with the inter-temporal application of the Damages Directive(again)… The question was about whether Article 9(1) of the Directive, which provides for the probative value of infringement decisions of NCAs, should be applied to the case in question. The facts of the case are rather undistinguished (as it is usually the case with private enforcement preliminary references). The Spanish competition authority had taken decisions in 2001 and 2009 finding that Repsol had committed RPM violations in its contracts with resellers. The Spanish courts confirmed the decisions. Then, there were actions for damages. The question was whether the claimants could use the local NCA infringement decisions as irrefutable proof that an infringement had taken place, as Article 9(1) of the Directive commands. AG Pitruzzella thought that Article 9(1) was not temporally applicable to the case. This was because it is a “substantive provision” and therefore cannot be applied retroactively, as Article 22 of the Directive makes clear. Following the Volvo/DAF line of case law (see my stories of May-June 2022), the AG checked whether the situation at issue in the main proceedings arose before the expiry of the time limit for transposition of the Directive or whether it continued to produce effects after the expiry of that time limit. In this case, the action for damages concerned an infringement resulting from restrictions on competition contained in contracts the effects of which ceased before the expiry of that date. In addition, the claim referred to alleged harm that was caused during a period that elapsed before that date. So it was for the national court to determine the evidential value of the local NCA infringement decisions pursuant to the well-known principle of national procedural autonomy, provided that the principles of equivalence and effectiveness were observed. The AG relied then on the principles of effectiveness and on legal certainty to opine that the NCA’s decision ought to be accorded “at least the value of prima facie evidence of the existence of the infringement for the purposes of the civil action” on one important condition: there must be a coincidence between the infringement of Article 101, as established in the local infringement decision, and the infringement alleged in the civil action as regards the nature of the infringement and its material, personal, temporal and territorial scope. A well-written and sensible Opinion.

In September 2022, we also had AG Rantos’s Opinion on the celebrated Facebook case in a preliminary reference from Germany. Although this goes back to the 2019 decision of the Bundeskartellamt, which had lots of adventures in Germany (see e.g. for an analysis here and here), the case before the Court of Justice is only minimally related to competition law, to the extent it raises some questions of competence. For the biggest part, the case is about the GDPR, and I am not going there. The competition law question as put by the German referring court was whether an NCA, such as the Bundeskartellamt, can rely on breaches of the GDPR to enforce competition law, even if it is not the competent supervisory authority within the meaning of Article 51 et seq. of the GDPR. The national court was also asking whether Article 4(3) TEU (duty of co-operation) poses any limitation if, at the same time, the competent supervisory authority is investigating an undertaking’s contractual terms relating to data processing. I am sure all readers would have preferred more questions on the substance of the competition analysis in this case, but I guess the national court did not have much leeway, since the German authority had deliberately refused to apply EU competition law, so no reference questions could be made on Article 102… The AG, first, took the view that the GDPR and the competition rules have different objectives and this means that (i) conduct relating to data processing may breach the competition rules even if it complies with the GDPR and (ii) conversely, unlawful conduct under the GDPR does not automatically mean that it breaches competition law. So, for the AG, while a competition authority does not have jurisdiction to rule on an infringement of the GDPR as such, it may nevertheless, in the exercise of its own powers and without prejudice to the powers of the competent supervisory authority under the GDPR, incidentally take account of the compatibility of a commercial practice with the GDPR. In that respect, the compliance or non-compliance of that conduct with the provisions of the GDPR may, in the light of all the circumstances of the case, be an important indication of whether that conduct amounts to a breach of competition rules. However, the competition authority must take account of any decision or investigation by the competent supervisory authority, inform the latter of any relevant details, and, where appropriate, consult it. Interestingly, the AG relied on Article 4(3) TEU to “discover” the latter obligation, although that provision applies between EU and Member State organs and not horizontally between different Member States. In so doing, the AG relied on a rather expansive and teleological reading of Article 4(3) TEU, as well as on “the principle of sound administration as a general principle of EU law, which includes, inter alia, an extensive duty of diligence and care on the part of national authorities” (para. 28). We will see soon how the Court of Justice will deal with this question methodologically.

Finally, we had AG Kokott’s Tráficos Manuel Ferrer Opinion in September 2022. Again, a private enforcement preliminary reference from Spain! I am not sure if all these Spanish cases show a strength or a weakness of private antitrust enforcement in that country… The issue at stake was whether the national rules on the award of legal costs were compatible with the principle of effectiveness. In particular, whether the principle of effectiveness requires the setting aside of a national rule, according to which the legal costs are to be divided equally between the claimant and the defendant, when the former has been successful in part. The AG was sympathetic to the position of the claimants. She took the view that although Article 101, in conjunction with the principle of effectiveness, does not in principle preclude such a national rule, the national court must interpret that rule in conformity with EU law. This may mean that when the claimant has been partially unsuccessful because it was excessively difficult or practically impossible to quantify the harm, the defendant is to bear all the costs or, depending on the circumstances of the case, may be required to bear at least a reasonable portion of the claimant’s costs. In addition, the referring court asked some questions relating to the quantification of harm and the estimation of the amount of damages. In that respect, the AG followed Volvo/DAF, which had already held that Article 17(1) of the Damages Directive, concerning the possibility of estimating the amount of harm, is procedural in nature. Therefore, the national provisions transposing this rule were applicable in the main proceedings in accordance with Article 22(2) of the Directive. In the case at hand, the defendant had actually granted the claimants access in a data room at its premises to all data taken into consideration in its expert report. However, this did not make any difference in terms of the quantification of harm. Again, the AG was sympathetic to the claimants. She took the view that the above did not rule out the existence of information asymmetry between the parties or that it was practically impossible or excessively difficult to quantify the harm. So the national level may, indeed, have to be more proactive in undertaking an estimation of the harm pursuant to Article 17(1) of the Damages Directive.

III. General Court

  1. Illumina/Grail

On 13 July 2022, the General Court delivered its Illumina/Grail judgment, in what promises to be a major saga in EU competition law (see under IV below). The way I have always understood the Merger Regulation and its thresholds is that it created an irrebuttable presumption that mergers below EU and national thresholds can be seen as pro-competitive, except for the very rare case that the merger amounts to an abuse of a dominant position (see the Towercast case – to be analysed in one of my next stories). That particular exceptional scenario is the ultimate safety valve that ensures that there is no enforcement gap. However, ever since the Commission decided to change its “approach” with the publication of its new Guidance on Article 22 of the Merger Regulation (see my stories of March-April 2021), this certainty has been lost. The Illumina/Grail transaction was the first “victim”.

The Commission received a complaint against the transaction in December 2020, and in February 2021, it sent a letter to Member States “inviting” them to refer the case under Article 22 of the Merger Regulation. On 19 April 2021, the Commission “accepted” pursuant to Article 22(3) of the Merger Regulation a referral under Article 22(1), made by the French Competition Authority and joined by five more NCAs (see further below under IV), and asserted its jurisdiction to examine the concentration. It is this step of the Commission that Illumina challenged before the General Court in the present case. Illumina argued that the Commission lacks the competence to review its acquisition of Grail and that there was a legal error in the Commission’s interpretation of Article 22. The General Court adjudicated the case under an expedited procedure.

The General Court delivered an astounding victory for the Commission. Although the act challenged was the above step of the Commission in the specific case and not the new Guidance, in essence, the General Court confirmed the Commission’s rather bold move to change its “approach” in accepting referrals from Member States in cases where a merger falls below national notification thresholds. The General Court stressed that Article 22 of the Merger Regulation allows that and relied on a literal, historical, systematic, and teleological interpretation of that provision to reach that conclusion. First, according to the General Court, the text of Article 22 refers to “any concentration” and is therefore quite open-ended (para. 89 et seq.). Second, while Article 22 originally, indeed, aimed at Member States without their own merger control system (the famous “Dutch clause”), that does not mean that that provision was to be applied to that situation alone (paras 97-117). The General Court also relied on subsequent legislative amendments to make the same point. Third, the general structure of the Merger Regulation makes clear that its scope and therefore the extent of the Commission’s powers depend primarily on the exceeding of the turnover thresholds that define the Community dimension but also, in the alternative, on the referral mechanisms laid down, inter alia, in Article 22 (paras 123-124). Fourth, the objective of the Merger Regulation is to permit effective control of all concentrations with significant effects on the structure of competition in the European Union, and the referral mechanism at issue is a “corrective mechanism” forming part of that objective (paras 165, 177).

The General Court also dismissed Illumina’s pleas that the referral request was made out of time. According to Article 22, Member States must refer a case to the Commission at most within 15 working days of the date on which the concentration was notified, or if no notification is required, when it was otherwise “made known”. According to Illumina, the merger was announced in September 2020, i.e. many months before its referral. However, according to the judgment, the words “made known” should be understood as the active transmission of information to the Member State concerned, which is appropriate for it to be able to assess, on a preliminary basis, whether the necessary conditions for the purposes of a referral have been satisfied (para. 200 et seq.). Therefore, the Commission’s invitation letter was an appropriate means for the “making known” referred to in Article 22. That being so, in the context of the examination of the subsidiary complaints alleging infringement of the principles of legal certainty and good administration, the General Court stressed that the Commission was, nevertheless, required to comply with reasonable time limits. A period of 47 days between the complaint and the invitation letter was unreasonable. Nevertheless, since it was not established that this procedural irregularity affected Illumina’s capacity to defend itself effectively, it could not justify the annulment of the contested decisions of the Commission to accept jurisdiction.

Finally, the Court also rejected Illumina’s plea based on the violation of its legitimate expectation (meaning that the Commission’s approach prior to March 2021 was not to accept such referrals). The Court’s reasoning shows how difficult it really is for applicants to prevail on the grounds of legitimate expectation: “the right to rely on the principle of the protection of legitimate expectations presupposes that precise, unconditional and consistent assurances originating from authorised, reliable sources have been given to the person concerned by the competent authorities of the European Union” (para. 254). Good luck with that!

In the end, it will now be for the Court of Justice to give definitive answers to these important questions. I would not take anything for granted…

  1. Google Android

Full disclosure: my name is on the list of lawyers representing Google, so my views are biased… Anyway, this was a huge victory for the Commission and Vice-President Vestager personally. The highest antitrust fine ever imposed by the Commission survived almost intact. The judgment offers an interesting juxtaposition in terms of how the EU Courts treat pricing and non-pricing abuses (see my views on this question here). Google was successful with regard to the former findings of the decision but failed with regard to the latter.

Let me start with market definition and dominance, which is a point that Google heavily contested because of the overall competition with Apple. As a reminder, the decision had found Google dominant in three markets: (i) the national markets for general search services (not just mobile search) – this was exactly the same as in the Shopping decision; (ii) the worldwide market (excl. China) for the licensing of smart mobile OS; and (iii) the worldwide market (excl. China) for Android app stores. I don’t need to stress that the latter two markets were defined in such a way that Google was essentially found a monopolist, with Android in the first and Play in the second of these two markets 🙂… The General Court followed the Commission’s approach entirely. It did not agree that Apple’s iOS and the App Store constrained Android and Play. With regard to Android, the General Court did accept that there could be an “indirect” constraint, but that did not make any difference. In addition, the General Court rejected the proposition that Android’s open-source nature represented a constraint. With regard to Play, the General Court did agree that Google competes against Apple (systems competition), but this was not enough. Finally, the General Court rejected Google’s argumentation that there was a mismatch and self-inconsistency between the finding of dominance in the general search markets, “where the service is offered to end users, and the tying abuse, which related to contracts with OEMs” (paras 259-261). It is also noteworthy that the General Court confirmed the Commission’s use of the SSNDQ test (“small but significant and non-transitory degradation of quality”) to find Google dominant in the market for licensable OSs (para. 180).

Let me now go to the abuse parts. The Commission’s decision found three types of abuses:

  1. it had considered illegal tying the fact that Google required OEMs, as part of its Mobile Application Distribution Agreements (MADA), to pre-install Google Search and Chrome; in reality, there were two forms of tying: tying the Google Search app with the Play Store (abuse of dominant position on the market for Android app stores) and tying Chrome with the Google Search app and the Play Store (abuse of dominant positions on the market for Android app stores and on the national markets for general search);
  2. it had also considered illegal tying the fact that Google signed Anti-Fragmentation Agreements (AFA) that prevented manufacturers from selling smart mobile devices with non-compatible “Android Forks”;
  3. it had taken issue with Google’s Revenue Sharing Agreements (RSA), which gave financial incentives to certain OEMs and MNOs for exclusively pre-installing Google Search on a portfolio of their devices (per device RSAs were not challenged); this was seen as a typical exclusivity rebates violation.

With regard to the first abuse, the MADA-related tying, the General Court fully confirmed the Commission’s findings. Before doing so, the General Court discussed at some length the legal test in tying cases (paras 283-299). This part is quite interesting because the case law on tying is old and certainly predates the post-Guidance Paper new generation of cases, which confirmed that a violation of Article 102 needs to be based on actual or likely effects leading to anti-competitive foreclosure. This part of the judgment admittedly could have been written better; it could have been clearer as to whether tying is an abuse “by effect”. I think that is what the General Court is saying, in particular in para. 295, but the second part of the second sentence of that paragraph is utterly off-putting and disappointing (“an examination [of effects] serves […] to clarify the gravity of the conduct in question, which will facilitate determination of the appropriate level of any penalty”)… Hopefully, the Court of Justice will clarify that fundamental point.

Going to the substance, Google argued that pre-installation amounted to neither exclusivity nor default status. In reality, Google said, the MADA conditions protected it from the risk of exclusive pre-installation by rivals. The OEMs could and did install rivals on a non-exclusive basis and downloads were an effective means to counter pre-installation. So there was no foreclosure. The General Court did not agree. It confirmed the Commission’s findings that the MADA conditions foreclosed rivals and that pre-installation was a “significant competitive advantage” because of the “status quo bias” it created. For the General Court, pre-installation and default were more or less the same things (paras 326-335), and the competitive advantage of “status quo bias” could not be offset by competitors. The judgment has quite harsh words on that point: “A distinction must be made in this respect between theoretical competition assumptions and the practical reality, where the competitive alternatives to which Google refers appear to have little credibility or real impact due to the ‘status quo bias’ arising from the MADA pre-installation conditions and the combined effects of those conditions with Google’s other contractual arrangements, including RSAs” (para. 428). Disappointingly, the General Court did not really respond to Google’s argument that it was an error to conflate “significant competitive advantage” with anti-competitive foreclosure. Another point of contention is the General Court’s brushing aside Google’s counterfactual arguments, which were made in a particularly forceful way at the oral hearing. The argument is that Android has created numerous opportunities for rivals (search engines and browsers). Would these opportunities have existed absent the MADA preinstallation conditions? The General Court held, however, that “even taking into account the pro-competitive effects generated by the Android platform”, the MADA pre-installation conditions amounted to an abuse of a dominant position (para. 592). In reality, the judgment never examined the “but for” world. In addition, it concluded that Google’s conduct was not objectively justified and that Google could not argue that the MADA preinstallation conditions were necessary to recoup the costs related to keeping Android free (paras 613-617). Finally, a point that will certainly figure prominently on appeal was the General Court’s reliance on the combined effects of the MADA preinstallation conditions and of other agreements that were not abusive. In so doing, the judgment relied on per-device RSAs (which the decision did not challenge) and portfolio RSAs (which were found not to be abusive). According to para. 452, “it must be noted that the taking into account as a factual element of the combined effects of MADAs and RSAs does not in any way depend on whether or not the RSAs are abusive, irrespective of whether they are portfolio-based RSAs constituting an abuse according to the Commission’s analysis, which is challenged by Google in the context of the third plea, or device-based RSAs which are not considered abusive in the contested decision”.

With regard to the second abuse, again the General Court fully upheld the decision. The licensing of Play and of the Search app only if OEMs entered into an AFA meant that OEMs could not distribute devices based on incompatible “forks”. This limited market access for non-compatible Android “forks”. As simple as that. The General Court was not persuaded by Google’s arguments that the AFA addressed the risk of fragmentation and thus fostered competition in the Android ecosystem. It took the view that incompatible “forks” were not part of that ecosystem. While “the obligations that are intended to ensure the compatibility of Android-compatible forks and interoperability within the ‘Android ecosystem’ [had] pro-competitive effects […] which are not disputed” (para. 890), the AFA abuse extended to “restrictions of competition that arise outside that ecosystem”, and this meant that Google’s objective justification could not stand (para. 891). The General Court again rejected Google’s counterfactual arguments and held that the Commission “was not also required” to conduct such an analysis (para. 893).

The Commission’s findings related to the third abuse did not survive. These findings should now be considered definitively annulled, since this part of the General Court judgment was not appealed to the Court of Justice. The annulment was essentially based on two grounds: first, the coverage of the relevant market was too low to show foreclosure, and second, the AEC test had not been conducted with adequate rigour. This all meant that the abuse could not be considered sufficiently established (para. 800). In particular, starting with the analysis of the coverage, in para. 679, the General Court restated the Intel obligation to analyse “coverage” and noted that, in previous cases, coverage was between 39% and 85% and that it was “plausible” (para. 685) that the coverage here was between 0% and 5% (the real figure is confidential)… The General Court also deplored the fact that the Commission never really calculated the coverage in its decision (para. 686). The Court went even further and said that even if some of the Commission’s arguments were correct (that the RSAs covered important segments), again coverage of 10-20% is not enough (para. 692). These references are very useful because it is the first time the EU Courts refer to insufficient coverage numbers. To be clear, the General Court did accept in paras 696-697 that sometimes qualitative elements may be taken into account while establishing “coverage”, e.g. the fact that the segment covered by the practices in question may be of “strategic importance”. The problem, however, was that the Commission never really went there and did not run its case like that. As to the AEC test, the General Court started from the premise that “where, as in this instance, the AEC test is applied, it must be conducted rigorously” (para. 644). Well, now you see why the Commission recently nixed the AEC test 🙂… The problems here were that the Commission had committed errors in (i) taking into account Google’s operational costs instead of requesting it to provide figures on its incremental costs, (ii) taking the 12% that all search rivals achieved on PCs as the maximum share that an as efficient competitor could achieve on mobile devices, when Seznam achieved 26%, and (iii) conflating actual competitors and as efficient competitors, when analysing what level of pre-installation an as efficient competitor could achieve.

Then, the General Court rejected certain procedural arguments relating to the taking of minutes by the Commission for meetings with third parties. In the post-Intel world, these types of arguments are heard quite often in Luxembourg, and sometimes they are even successful… According to the General Court, the minutes may have been somewhat inadequate, but this procedural error did not breach Google’s rights of defence, as Google had not demonstrated that absent the error it would have been better able to defend itself. However, Google prevailed in its second procedural argument relating to the Commission’s failure to hold a hearing on the AEC test. The Commission had made some adjustments in its AEC test but instead of sending a supplementary statement of objections, which would have entitled Google to an oral hearing, it only sent letters of facts, which do not trigger the obligation to hold an oral hearing. Finally, although Google had succeeded in the annulment of one out of the three abuses in the decision, the reduction of the fine was minimal. The fine was reduced only by 5% to EUR 4.125 billion, based on the General Court’s unlimited jurisdiction. In reality, if one scratches below the surface, the General Court increased the fine in one of the intermediate stages for its calculations. It suffices to read para. 1081: when referring to the fixed gravity coefficient of 11%, the General Court said that this “did not sufficiently reflect the reality of the implementation of the infringement and in particular its intensity during the period concerned, particularly as regards […] Google’s anticompetitive conduct in the years 2012 to 2014”. As regards the implications of annulling the findings on the RSAs, the General Court held that the finding of a single and continuous infringement was not affected. The remaining abuses (MADA and AFA) formed part of an overall strategy that was not vitiated by any illegality, and anyway, the RSA remained relevant as “elements of factual context for the purpose of assessing the exclusionary effects of the first and second aspects of the single and continuous infringement [meaning the MADA and AFA abuses]” (para. 1018).

IV. European Commission

As far as Commission developments are concerned, in July 2022, the Commission accepted commitments in Network sharing – Czech Republic, a case that had stalled for a number of years. The commitments were probably as best as what the Commission could have got here. They were offered by T-Mobile CZ, CETIN, and O2 CZ, as well as their parent companies Deutsche Telekom and PPF Group. The companies must ensure that their network-sharing agreements do not reduce infrastructure competition which enables competition and innovation in the wholesale and retail telecommunications markets in Czechia. In particular, the companies concerned committed to (i) modernising the mobile network equipment to enable more flexibility and independence for the two sharing parties in certain radio frequencies; (ii) reviewing and changing the financial conditions for unilateral network deployments in order to remove financial disincentives for such unilateral deployments; (iii) improving the contractual provisions limiting information exchange to the minimum necessary for the operation of the shared network; (iv) implementing measures to ensure that CETIN effectively prevents information spill-over between T-Mobile CZ and O2 CZ; and (v) not extending the geographical scope of the existing network sharing to Prague and Brno for a period between 7 and 10 years, in order to ensure that each player continues to deploy its own 2G, 3G and 4G networks in full independence in these two large cities.

In the merger control area, I would mention Bouygues/Equans, where the Commission conditionally approved Bouygues’s acquisition of Equans subject to Bouygues divesting Colas Rail Belgium in its entirety, including all assets, personnel, and ongoing and future contracts of both its railway contact lines and track installation businesses. As a result, Colas Rail Belgium remains an independent competitor to Bouygues and Equans in the relevant market in Belgium. Another interesting case is D’Ieteren/PHE, where the Commission conditionally approved D’Ieteren’s acquisition of PHE subject to D’Ieteren divesting Mondial Pare-Brise and the Glass Auto Service label in their entirety. Both were Phase I clearances.

Finally, in early September, the Commission prohibited the ill-fated Illumina/Grail deal in one of the most dramatic competition events of 2022. This was a vertical merger! Illumina has a strong position in the supply of Next Generation Sequencing (NGS) systems for genetic and genomic analysis, while Grail, a customer of Illumina, uses NGS systems to develop cancer detection tests. The Commission found that Grail and its rivals are currently engaged in an innovation race to develop and commercialise early cancer detection tests. While there is still uncertainty about the exact results of this innovation race and the future shape of the market for early cancer detection tests, protecting the current innovation competition was crucial to ensure that early cancer detection tests with different features and price points will come to the market. According to the Commission, Illumina would have had the ability and the incentive to engage in foreclosure strategies against Grail’s rivals, who depend on Illumina’s technology, from access to an essential input they need to develop and market their own tests. It could for instance refuse to supply its NGS systems to Grail’s rivals, increase the prices, or degrade the quality and delay supplies. The Commission considered that those strategies would have resulted in a significant detrimental effect on competition in developing and marketing NGS-based cancer detection tests. Illumina offered mostly behavioural remedies, in particular a licence open to NGS suppliers to some of Illumina’s NGS patents, a commitment to stop patent lawsuits in the US and Europe against an NGS supplier, and a commitment to conclude agreements with Grail’s rivals under the conditions set out in a standard contract, but the Commission was not impressed. Interestingly, the Commission also announced that it would adopt a divestiture decision, as well as take (new) interim measures to deal with the short-term issues before the divestiture decision is taken (in addition to the interim measures decision of October 2021). Indeed, the new interim measures decision was taken on 28 October 2022. In other words, we are heading for a world record of Commission decisions and appeals in a single case. We have 11 Commission decisions and 6 possible appeals (and more than 6 judgments if Grail continues filing its own appeals)! Here is the list – not in chronological order – but better to understand legally, bearing in mind the appeals:

  1. Six Commission decisions accepting the referral requests of France, Belgium, Greece, Iceland, the Netherlands, and Norway (19 April 2021) – appeal to and judgment of the General Court (see above) – appeals by Illumina and Grail to the Court of Justice (respectively, pending Cases C-611/22 P and C-625/22 P);
  2. Commission decision prohibiting the acquisition (6 September 2022) – appeal to the General Court (pending Case T-709/22);
  3. Commission decision in the main case for the early closing against the standstill obligation & imposing a fine (decision still to be adopted – Statement of Objections sent on 19 July 2022) – no doubt there will be appeals to the General Court;
  4. Commission interim measures decision for the early closing against the standstill obligation (29 October 2021) – appeals to the General Court by Illumina and Grail (respectively, pending Cases T-755/21 and T-23/22);
  5. Commission decision ordering divestiture (decision still to be adopted – Statement of Objections sent on 5 December 2022) – no doubt there will be an appeal to the General Court here too;
  6. Commission decision ordering interim measures post the prohibition decision and prior to the divestiture (28 October 2022) – appeal to the General Court (pending Case T-5/23).

Executive Vice-President Vestager had actually said that the unwinding of the deal would be a first for the Commission, but the famous Schneider/Legrand case is another such case. And it actually shows what could well happen here. If decision no. 1 is annulled, then decision no. 2 collapses too, and in that case, the further decisions will collapse. That’s because there will be no legal basis for the second, the third, and so on. This supervening event will mean that the second, third etc. application for annulment will instantly be successful and the General Court will issue a summary judgment in Illumina’s (and Grail’s) favour. This is what happened in Schneider/Legrand, where the General Court annulled the Commission’s divestiture decision (Case T-77/02), since the prohibition decision was also annulled (Case T-310/01). The domino effect! Anyway, prepare for a long-standing saga.

Finally, at the end of September, the Commission adopted its Guidelines on the application of EU competition law to collective agreements regarding the working conditions of solo self-employed persons. The Guidelines clarify when certain self-employed persons can get together to negotiate better working conditions collectively without breaching EU competition rules. The difficulty is that self-employed persons are considered “undertakings” and thus risk infringing competition rules when negotiating collectively on their fees or other trading conditions. The Commission clarifies that competition law does not apply to solo self-employed persons that are in a situation comparable to workers. These include solo self-employed people who: (i) provide services exclusively or predominantly to one undertaking; (ii) work side-by-side with workers; and (iii) provide services to or through a digital labour platform. The Commission states that it will not enforce EU competition law against collective agreements made by solo self-employed people who are in a weak negotiating position and face an imbalance in bargaining power due to negotiations with economically stronger companies or when they bargain collectively pursuant to national or EU legislation. My personal view is that the risks have always been low if not inexistent because I can’t see a competition authority ever opening or prioritising such a case, but we are living in interesting times, hence the Guidelines

Makis Komninos


Citation: Makis Komninos, Competition Stories: July – September 2022, Network Law Review, Spring 2023.

Read all the competition stories over here: link

Related Posts