General Court Partially Annuls European Commission Decision in Google Android
October 20, 2022
In September 2022, the General Court partially annulled the European Commission’s 2018 Google Android decision, which fined Google €4.3 billion for abuses of dominance relating to apps it offers for its Android mobile operating system (“OS”).
The Court also found that the Commission’s investigation suffered from procedural errors. It reduced the fine by €200 million.
The Court upheld the Commission’s finding of abuse regarding the tying of certain Google apps. It also maintained the Commission’s finding that Google had unlawfully restricted competition from alternative versions of Android via its anti-fragmentation agreements. The judgment nonetheless underscores the Commission’s obligation to examine alleged exclusionary effects rigorously. This judgment is the third annulment of allegedly abusive conduct this year due to a lack of rigour in the Commission’s assessment of alleged anticompetitive effects or infringements of parties’ rights of defence.
Cleary Gottlieb acted as co-counsel for Google before the European Commission and General Court.
Google introduced Android in 2008. Android’s launch followed the introduction of Apple’s iPhone in 2007, which defined a new category of innovation in mobile telephony.
The Android OS is made available on a free, open-source basis to original equipment manufacturers (“OEMs”) and mobile network operators (“MNOs”). Google’s business model in mobile depends principally on the distribution of its revenue-generating services, in particular Google Search, on smart mobile devices produced by third-party OEMs. By contrast, Apple, which is vertically integrated in the production of smart mobile OSs (iOS and iPadOS) and devices (iPhones and iPads), generates revenue primarily from the sale of its devices.
On July 18, 2018, following a four-year investigation, the Commission fined Google €4.3 billion for abusing its alleged dominant position in the markets for licensable smart mobile OSs (in which Android was found dominant), Android app stores (in which its app store, Play, was found dominant), and online general search services (in which Google Search was found dominant). The Commission challenged provisions in three of Google’s agreements with OEMs and MNOs in relation to Android mobile devices:
- First, the Commission challenged provisions in the Mobile Application Distribution Agreement (“MADA”) under which Google licenses its apps and services to Android OEMs as a suite. It found that the MADA contained two unlawful tying arrangements between: (a) the Play Store and the Google Search app; and (b) the Play Store, the Google Search app, and Chrome (Google’s web browser).
- Second, the Commission challenged provisions in the Anti-Fragmentation Agreement (“AFA”), now called the Android Compatibility Commitment (“ACC”), under which OEMs that wished to preinstall Google’s apps under the MADA agreed not to sell devices running versions of Android that did not comply with Android’s baseline compatibility standard (referred to as “incompatible Android forks”).
- Third, the Commission challenged Google’s portfolio-based Revenue Share Agreements (“RSAs”), under which Google compensated OEMs and MNOs for sole promotion of the Google search app across an agreed portfolio of devices. The Commission did not, though, challenge device-by-device deals, under which OEMs and MNOs were free to choose the devices on which they wished to promote Google apps.
In the Commission’s view, these agreements formed part of Google’s overall strategy to cement its dominant position in the online general search market at a time when the importance of the mobile internet was growing.
Google challenged all four abuses on appeal, as well as the Commission’s assessment of market definition and dominance and the financial penalty imposed. Google also alleged that its procedural rights had been infringed during the Commission’s investigation.
General Court Judgment
The Court largely dismissed Google’s appeal. It did, however, annul the infringement regarding Google’s portfolio-based RSAs and uphold Google’s pleas based on infringements of its rights of defence. The Court reduced the fine by €200 million.
A summary of the Court’s findings on each plea is below.
Market Definition & Dominance
The Court upheld the Commission’s finding that Android and iOS belonged in separate relevant product markets, that iOS did not sufficiently constrain Android, and, by the same logic, that Play was not sufficiently constrained by Apple’s App Store.
The Court’s starting point was that iOS could at best pose an “indirect” constraint on Android because it is not available for license by OEMs. It could not be included in the relevant market, which contained only licensable smart mobile OSs. The Court largely confirmed the Commission’s assessment of the evidence of competitive constraint. For example, it accepted that Google’s internal documents demonstrated that Apple is a “competitor” and a “competitive relationship” between Android and iOS, but held that the internal documents were inconclusive as to the importance of iOS as a competitive constraint. The Court largely dismissed evidence Google put forward on parallel innovations in iOS and Android, holding that Google innovated Android to boost distribution of Google Search, not compete with Apple.
The Court endorsed the Commission’s assessment of dominance by reference to a “Small But Significant And Non-Transitory Decrease In Quality” (“SSNDQ”) framework, and agreed that the SSNDQ did not have to be quantifiable. On the facts, the Court found that users and developers would not switch to iOS in response to an SSNDQ. This was down to users’ loyalty to their OS and costs of switching to another OS.
The Court accepted Google’s argument that the quality of the OS is important to users when buying a mobile device. The Court said, however, that there were, in addition, other “factors determining a user’s choice,” which meant that Google could degrade Android’s quality without users switching. The Court’s reasoning is unsatisfactory because the fact that there are multiple parameters of competition does not necessarily mean any one of them is irrelevant. In addition, the Court did not address Google’s position that these other factors (such as the battery life and ease of use) depend on the quality of the mobile OS, a point that the Commission itself had made in its Google/Motorola merger decision.
The Court held that the Commission committed an error in finding that Apple’s pricing policy was a barrier to switching by users of high-end Android devices, but concluded that this error was “of no consequence” as users’ switching behaviour depended on other factors. The Court did not, however, assess whether the Commission’s findings on pricing played a particularly important role in its analysis, such that the Commission’s error could have undermined its overall conclusions. The Court also rejected Google’s argument that Android is competitively constrained by the open-source license to Android, which allows Android substitutes to be developed.
On Play, the Court accepted Google’s argument that Android and Play compete as a system with iOS and Apple’s App Store. The Court disagreed, however, that the App Store sufficiently constrains Play, basing its finding largely on the same grounds that iOS was found not sufficiently to constrain Android.
Finally, the Court rejected Google’s argument that the Commission’s finding of dominance in general search did not match the alleged abuse, which concerned licensing of the Google Search app to OEMs. The Court reasoned that the tying abuses and AFA abuse relied on Play’s dominance and not only on Google’s search dominance, and that the “practices at issue were closely linked to Google’s dominant position” in general search and were “directed in reality to the users of, and national markets for, general search services, on which Google held a dominant position.”
Under the MADA, Google requires OEMs that preinstall Play also to preinstall the Google Search app and Chrome. The Commission concluded that the MADA contained two unlawful tying arrangements, as explained above.
On appeal, the Court accepted that “close examination of the actual effects or further analysis […] was required before it could be concluded that the tying in question was harmful to competition.” An effects-based analysis was required to “reduce the risk that penalties may be imposed for conduct which is not actually detrimental to competition on the merits,” and to “clarify the gravity of the conduct.”
The Court agreed with Google that “it is easy for users to obtain general search or browser apps competing with those that are subject to tying,” which the Commission also recognized. It nonetheless agreed with the Commission that preinstallation of the Google Search app and Chrome under the MADA was capable of having—and did have—anticompetitive effects, as rival general search services and browsers were unable to offset the anticompetitive advantage that Google achieved through preinstallation.
- First, the Court accepted that OEMs were free under the MADA to preinstall apps competing with the Google Search app and Chrome. In the case of browsers in particular, they often did. It concluded, however, that even though certain findings in the Decision were unsupported by the evidence, preinstallation of competing apps alongside Google’s apps “did not happen for much of the infringement period” due, at least in part, to the “combined effects of the MADAs, the RSAs and the AFAs.”
- Second, the Court confirmed that “users can easily download general search service or browser apps competing with the Google Search app or Chrome.” Downloading could “in principle offset the advantage that would be conferred by pre-installation.” However, rival apps “are not downloaded in practice, or in any event are downloaded for an insufficient proportion of the devices concerned.” The Court’s reasoning on this point creates some questionable consequences. Under its logic, preinstallation of low-quality apps is not abusive, because users will download rivals to those apps (as the Commission had accepted, for example, with respect to messaging apps). But preinstallation of high-quality apps (as the Court accepted Google Search is) can be problematic because users may not then seek out rivals of those apps.
- Third, in respect of Google’s argument that users could access competing search services through their browsers, the Court held that “users do not in practice access other general search services through browsers and only rarely change the default settings of those browsers.” It did not, though, explain what stopped users from accessing their preferred search services through browsers, even if those services are not set as default.
The Court rejected Google’s proffered objective justification that the MADA preinstallation conditions were necessary to recoup its investments in maintaining the free Android platform. Google, the Court concluded, “has always been in the position of having significant sources of revenue to finance those investments,” including through revenues generated by the Play Store.
AFAs and ACCs
The Commission had found that Google’s practice of conditioning licenses to Play and the Google Search app (under the MADA) on anti-fragmentation obligations contained in the AFAs/ACCs constituted an abuse of dominance. The anti-fragmentation obligations, which set a minimum baseline of compatibility across a signatories’ whole portfolio of Android devices, allegedly inhibited the development of incompatible Android “forks” (i.e., versions of open-source Android that did not meet the compatibility requirements). Incompatible Android forks could in turn have supported the distribution of rival general search services.
The Commission accepted that Google could enforce its compatibility requirements in respect of Android devices that preinstalled Google’s proprietary apps and services. The parties contested the lawfulness of the requirements as they applied to other Android devices.
Google argued that, as steward of the Android ecosystem, it needed to protect the Android ecosystem from the fragmentation inherent in open-source license models. It also argued that the Commission failed to establish that the lack of success of incompatible Android forks (in particular Amazon’s Fire OS) was attributable to the AFA/ACC, rather than their lack of access to the attractive apps and services that users demanded.
The Court upheld the abuse, concluding that Google’s conduct “deprived potential or existing competitors of Google of any market, strengthened Google’s dominant position on the markets for general search services and deterred innovation.” It did so for four main reasons: (i) Google knowingly implemented the practice with the aim of limiting market access of incompatible Android forks; (ii) Google failed to establish that incompatible Android forks could not have constituted a credible competitive threat to Google’s approved Android implementation; (iii) the AFAs/ACCs in fact inhibited the development of incompatible Android forks (it not mattering if there could have been other reasons for the failure of Fire OS, such as lack of access to Google’s proprietary apps and services); and (iv) Google had not demonstrated that its conduct was objectively justified.
Under Google’s portfolio-based RSAs, it paid OEMs and MNOs a portion of its search advertising revenue if the Google Search app was the sole preinstalled search app across an agreed portfolio of Android devices. The Commission found that these payments constituted exclusivity payments that were capable of foreclosing as-efficient competitors to Google. Google moved from portfolio-based RSAs to “device-based” RSAs in 2013-2014. The Commission did not challenge device-based RSAs, which left OEMs and MNOs free to choose the devices on which they wanted to promote Google apps, and so did not put OEMs or MNOs before an all-or-nothing choice.
The Court annulled the Commission’s finding of abuse on two grounds:
- First, the Court held that the Commission was required, following Intel, to assess the coverage of the challenged practice. The Court criticized the Commission’s coverage assessment as failing to consider the entire relevant markets, rather than a segment of them. The Court was persuaded by Google’s coverage assessment, which showed that the portfolio-based RSAs covered less than 5% of the relevant markets. The Court went further in suggesting that coverage of [10-20]% would be insufficient. The Commission therefore failed to show that the coverage of the challenged practice was significant.
- Second, the Court criticized the Commission’s as-efficient competitor (“AEC”) analysis. The AEC test can be useful to establish the “intrinsic capacity of a practice to foreclose competitors which are at least as efficient as the dominant undertaking.” When the Commission carries out an AEC test, the Court held that “it must be conducted rigorously.” The Court found that the Commission’s AEC test contained several errors of reasoning and lacked the required rigour, and could not therefore support the Commission’s finding of abuse with respect to Google’s portfolio-based RSAs.
Rights of Defence
Google submitted that the Commission had infringed its rights of defence during the administrative procedure on two grounds:
- First, Google alleged that the Commission failed to respect its right of access to the file by neglecting to provide notes of meetings with third parties that were adequate to safeguard Google’s rights of defence.
- Second, Google argued that the Commission failed to respect its right to be heard by not adopting a supplementary statement of objections and thus denying Google an oral hearing regarding essential aspects of the Commission’s case on portfolio-based RSAs and the AEC test.
On the first sub-plea, the Court concluded that the interview notes in question were “too late and too cursory to be capable of constituting the record of an interview in the sense prescribed by Article 19(1) of Regulation No 1/2003.” Google had not, however, established that it would have been better able to ensure its defence had the procedural errors not been made. The procedural error therefore had no impact of the Commission’s findings of abuse.
The Court also upheld Google’s second sub-plea. It accepted Google’s argument that, instead of setting out its analysis of Google’s portfolio-based RSAs in two letters of facts (which did not afford Google the opportunity for an oral hearing), the Commission should have issued a supplementary statement of objections instead. The Commission’s findings in its letters of facts “substantially supplement[ed] the substance and scope of the objection” on portfolio-based RSAs, and “significantly alter[ed] the evidence.” This was the case in particular with the Commission’s AEC test, which “played an important role” in the Commission’s assessment of the portfolio-based RSAs. Google could have developed its defence “more easily orally” and would have had this opportunity had a supplementary statement of objections been issued instead. This procedural error further vitiated the portfolio-based RSA abuse.
The Court upheld the Commission’s finding of a single and continuous infringement, despite its annulment of the portfolio-based RSA abuse. The Court held that, “irrespective of the characterization of the RSAs in terms of competition law, the combined effects of the practices established by Google gave it the benefit, as regards Google Search, of exclusive pre-installation” on a significant proportion of Android devices in the EEA. Google’s RSAs—including device-based RSAs, which were not challenged––were correctly considered as “elements of the factual context” for the remaining aspects of the single and continuous infringement. The Court also rejected Google’s arguments on lack of intention or negligence, errors in the calculation of the fine, and the novelty of the abuses.
Ultimately, the Court reduced the fine by €200 million to €4.1 billion. In an apparent departure from the Commission’s Fining Guidelines, the Court made a calculation based on a new methodology that seems, in light of its annulment of the portfolio-based RSA abuse, to have increased the fine for the infringements the Court maintained.
The Court’s judgment largely upholds the Commission’s decision, but its approach to and subsequent annulment of the portfolio-based RSA abuse raises several questions.
First, the judgment demonstrates the requirement to establish anticompetitive effects in both pricing and non-pricing abuse of dominance cases. On the facts, the Court was persuaded that the Commission had established actual anticompetitive effects of Google’s tying arrangements by reference to evidence of OEM and user behavior regarding Google’s actual rivals. In other words, it was sufficient for the Court that, in practice, OEMs did not preinstall and users did not download rival search engines. For the exclusivity abuse, the Court took a different approach. It demanded evidence of anticompetitive effects on hypothetical as-efficient competitors and concluded that the Commission had failed in this exercise.
Second, the judgment raises questions of causation. The Commission is required to show that any alleged anticompetitive effects are attributable to the conduct under investigation. In the recent Qualcomm judgment, for example, the Court annulled a Commission decision because Qualcomm’s incentive payments to Apple did not cause Apple to choose Qualcomm’s chips. In upholding the MADA abuse, however, the Court relied on the “combined effects” of the MADAs and RSAs, even though the latter were absolved (in the case of portfolio-based RSAs) or unchallenged (in the case of device-based RSAs) and therefore any effects that arise from RSAs must be presumed to result from competition on the merits. Similarly, the Court held that it did not matter whether incompatible Android forks failed for reasons other than the AFAs/ACCs, such as their inability to license Google’s proprietary apps.
Third, the judgment creates a situation where both portfolio-based exclusive preinstallation and device-based exclusive preinstallation (under the RSAs) were not abusive, but device-based non-exclusive preinstallation (under the MADA) was said to be abusive. The recent trend in the Court’s case law seems to be to bless exclusivity pricing practices (traditionally viewed as the most egregious form of conduct), while outlawing non-pricing behaviour that is on its face more benign and potentially beneficial.
Finally, the judgment underscores the Commission’s obligation to conduct investigations free from procedural defects. The Court advised the Commission with respect to meeting minutes that “[i]n the future, it would be useful and appropriate for the record of each interview conducted by the Commission with a third party for the purpose of collecting information relating to the subject matter of an investigation to be made or approved at the time when that interview is held or shortly afterwards so as to be added to the file as quickly as possible to enable the person accused of an infringement, when the time comes, to acquaint himself or herself of it for the purpose of exercising the rights of defense.”
This is the third General Court judgment this year annulling a finding of abuse due to procedural errors or the Commission’s failure to establish anticompetitive effects, following Intel and Qualcomm.
 Judgment of September 14, 2022, Google LLC and Alphabet, Inc. v. European Commission, T-604/18, ECLI:EU:T:2022:54.
 Judgment, ¶¶139–142.
 Judgment, ¶147.
 Judgment, ¶¶148–152.
 Judgment, ¶180.
 Judgment, ¶¶182–205.
 Judgment, ¶196.
 Commission decision of February 13, 2012, Case COMP/M.6381, Google/Motorola, ¶83 (“[T]he mobile OS plays a crucial role in the success of a smart mobile device because the most important other features of smart mobile devices that are appealing to consumers are dependant on (or must be compatible with) the mobile OS.”).
 Judgment, ¶¶211–215.
 Judgment, ¶223.
 Judgment, ¶250. See also ¶270 (“The Google ‘ecosystem’, characterised by the relationship between the Android OS and the Play Store, would thus have been in competition with the Apple ‘ecosystem’, characterised by the relationship between iOS and the App Store.”).
 Judgment, ¶¶249–254. At ¶252, the Court holds that “[l]ogically, it is not possible that a competitive constraint exercised by the App Store on the Play Store would differ in intensity from that exercised by iOS on Android.”
 Judgment, ¶259.
 Judgment, ¶260.
 Judgment, ¶261.
 Judgment, ¶295.
 Judgment, ¶292.
 Judgment, ¶425.
 Judgment, ¶439.
 Judgment, ¶537.
 Judgment, ¶547.
 Judgment, ¶557.
 Judgment, ¶562.
 Judgment, ¶614.
 Judgment, ¶¶608–609.
 Judgment, ¶¶810; 828.
 Judgment, ¶892.
 Judgment, ¶¶837–841.
 Judgment, ¶¶843–847.
 Judgment, ¶850.
 Judgment, ¶¶866–891.
 Judgment, ¶688.
 Judgment, ¶684–685 (describing Google’s calculation as “plausible”).
 Judgment, ¶692.
 Judgment, ¶¶700–799.
 Judgment, ¶641.
 Judgment, ¶644.
 Judgment, ¶¶798–799.
 Judgment, ¶932.
 Judgment, ¶939.
 Judgment, ¶981.
 Judgment, ¶982.
 Judgment, ¶999.
 Judgment, ¶1029.
 Judgment, ¶1019.
 Judgment, ¶1018.
 Judgment, ¶¶1037–1051; 1061–1071; 1101–1106.
 Judgment, ¶1101.
 European Commission, 2006/C 210/02, Guidelines on the method of setting fines imposed pursuant to Article 23(2)(a) of Regulation No 1/2003 (September 1, 2006).
 See, e.g., Judgment of October 6, 2015, Post Danmark A/S v. Konkurrencerådet, Case C-23/14, ECLI:EU:C:2015:651, ¶47.
 Judgment, ¶451.
 Judgment, ¶933.
 Judgment, ¶933.