The European Court of Justice’s (ECJ) Android Auto judgment, delivered in late February, could mark a radical shift in how courts interpret the European Union’s essential-facilities doctrine, as well as the legal standard applied to “refusal to deal” cases. My colleague Giuseppe Colangelo has a great working paper analyzing the decision and its potential consequences.
The case raises a number of important questions, including the relevant market definition (is Android Auto a monopoly?) and the theory of harm (was anyone excluded from the market? Did this affect consumers?). This post will focus on what I see as the judgment’s internal inconsistencies, and how these are likely to leave European antitrust policy confused with regard to market incentives to create, innovate, and properly maintain digital platforms (or any productive asset, for that matter).
Background on the Case
The case arose from a dispute between Google and Enel X, a division of the Italian electricity company Enel S.p.A. focused on innovative energy solutions. Enel X offers users the JuicePass App to allow them to find Enel X recharging stations for electric vehicles. Android Auto offers a system whereby users can sync their Android smartphones with their cars’ infotainment systems in order to access some apps more conveniently via their dashboard screens, rather than by using a smartphone.
In 2018, Enel X requested Google to ensure that JuicePass would be interoperable with Android Auto. Google refused, saying that multimedia and messaging apps were the only third-party apps it sought to make interoperable with Android Auto. Google then refused a second Enel X request in January 2019, invoking security concerns, as well as the need to rationally allocate resources to create a new template.
Enel X filed a complaint before the Italian Competition Authority (AGCM) that claimed Google’s refusal constituted an abuse of dominant position under Article 102 of the Treaty on the Functioning of the European Union (TFEU). In April 2021, the AGCM ruled that, in obstructing the JuicePass app’s availability on Android Auto, Google’s conduct impeded competition. The Italian authority imposed a €102 million fine on Google and mandated that it publish a template for electric-vehicle-charging apps.
As the London School of Economics’ Pablo Ibañez Colomo has observed, the AGCM decision departs from European competition law precedent with regard to so-called “essential facilities” in that it did not require “indispensability.” As Ibañez Colomo put it, the decision instead appeared “to mark the comeback of a doctrine of ‘convenient facilities.’” It was also unprecedented in that it required Google to do more than simply provide access. The company would instead have to adapt its platform to ensure that JuicePass is compatible, which could entail costly changes to the platform’s design.
Google contested the AGCM decision before a regional administrative court (the Tribunale Amministrativo Regionale per il Lazio), which dismissed the action. It then appealed to the state council (Consiglio di Stato), which decided to refer the case to the ECJ regarding the proper interpretation of Article 102 TFEU. Specifically, the Consiglio di Stato asked:
i) whether Article 102 TFEU must be interpreted as meaning that a refusal of an owner of a platform to provide access to an app is capable of constituting an abuse of a dominant position even though that platform is not indispensable, but rather “convenient” to make that app more attractive to consumers; and,
ii) whether Article 102 TFEU must be interpreted as meaning the owner of a dominant platform may be required to adapt its platform (to develop that template”) to accommodate the needs of the app requesting access.
The Consiglio di Stato has not yet rendered a final decision in the case and could still affirm or reject the AGCM’s findings, but the Italian court is bound by the ECJ’s interpretation of Article 102 TFEU.
The ECJ Judgment and ‘Indispensability’
Rather than provide an interpretation of the concept of “indispensability” under the Bronner precedent, the ECJ’s judgment reinterprets the AGCM question and seeks to distinguish the Android Auto case from the established case law:
… where a dominant undertaking has developed infrastructure not solely for the needs of its own business but with a view to enabling third-party undertakings to use that infrastructure, the condition laid down by the Court of Justice in paragraph 41 of the judgment of 26 November 1998, Bronner, C?7/97, EU:C:1998:569), relating to whether that infrastructure is indispensable for carrying on the business of the entity applying for access, in that there is no actual or potential substitute for that infrastructure, does not apply.
45 In such a situation, neither the preservation of the freedom of contract and the right to property of the undertaking in a dominant position nor the need for that undertaking to continue to have an incentive to invest in developing high-quality products or services justify limiting a refusal to provide access to the infrastructure in question to a third-party undertaking being classified as abusive, as provided for in Article 102 TFEU, to cases in which that refusal makes it impossible for that third-party undertaking to carry on its business, by developing a viable offer on a neighbouring market.
46 It is sufficient to state, in that regard, that, where the cost of developing such infrastructure has been assumed by the undertaking in a dominant position not solely for the needs of its own business but with a view to that infrastructure being able to be used by third-party undertakings, the fact of requiring the undertaking in a dominant position to provide access to that infrastructure to a third-party undertaking does not fundamentally alter the economic model which applied to the development of that infrastructure. (Android Auto, Paragraphs 43-46, emphasis added).
This assessment of incentives is both static and fundamentally incomplete. Just because a company has developed infrastructure for third-party use does not mean that mandated sharing of that infrastructure with specific parties, in particular ways and timeframes, would not impact its ability to monetize that asset. Developing a platform is costly, and keeping it functional and attractive for users over time requires permanent investment in innovation, maintenance, and user-engagement strategies (marketing, cross-subsidies, etc.).
Furthermore, contrary to what the ECJ seems to assume, the decision to open a platform is not binary; it’s not simply a choice between being open or closed for “own use.” There exists a wide spectrum of strategic decisions to consider when opening a platform: How many users on each side of the platform will have access, under what conditions, at what price, and what will be the quality level of the services or products the platform offers? Platform owners (“gatekeepers” in today’s parlance) play a curation role that improves not only the platform’s value, but also increases consumer welfare (by, for instance, reducing the cost of finding and choosing good quality services or products).
Jonathan Barnett has analyzed at length the dilemma that digital platforms (“hosts”) face. They need to offer incentives for certain types of users (e.g., app developers) to join their platform in order to reach scale and in turn be attractive to other types of users (consumers). One common strategy is to grant open (free) access.
But developers also know that, once the platform reaches scale, it may have incentive to charge a higher price. To address that dilemma, as Barnett explains, “host entities will tend to implement hybrid structures that reflect a mix of open-access elements (to promote platform adoption) and closed-access elements (to recover costs).”
Think, for instance, of a shopping mall, a classic example of a two-sided platform. For the mall to thrive, it must appeal to both the retailers seeking to rent space and the consumers who choose to shop there. Key factors in achieving this balance include location, accessibility, amenities (such as parking and children’s playgrounds) and, most critically, a well-curated tenant mix. Some of these amenities would be granted for free (parking), while others come at a cost. Some retailers will pay fixed rents, while others may pay a combination of rent and a share of revenue.
Shopping centers often aim to attract “anchor” stores—large department stores that draw significant foot traffic or a movie theather—alongside a diverse array of shops that cater to a wide spectrum of consumer preferences. Therefore, a successful mall typically features a blend of retailers specialized in clothing, cosmetics, sporting goods, toys, electronic devices, jewelry, and dining options. It’s essential to strike the right balance; a mall cannot thrive with 40 clothing stores and no restaurants. Instead, it must identify the optimal combination that maximizes consumer visits.
Reaching this ideal mix is not always straightforward, and consumer preferences are continually evolving. The convenience of online shopping, for instance, has shifted demand for certain products. This dynamic environment demands that mall owners remain flexible and able to make ongoing decisions regarding tenancy. Occasionally, it will be necessary to phase out underperforming stores and replace them with retailers that better meet current consumer demands. In some cases, this might even mean eliminating a successful retailer if a new entrant shows greater potential to attract foot traffic and make more sales.
Something similar happens with digital platforms—even if they make multihoming more easy and congestion is less likely. Owners of digital platforms should be able to curate their “real estate” to make it attractive in the first place, and to keep it attractive over time. In this context, that means to keep it secure, functional and yes, profitable.
If it becomes more the general rule than the exception that competition authorities can force sharing, such interference with a company’s ability to monetize its assets could obliterate platforms’ incentives to build, conduct proper maintenance, and innovate. Empirical evidence on the impact of net-neutrality regulations, for instance, suggests that these sorts of regulations have had a “significant and strong negative impact on fiber investments.”
The ECJ used to understand this (to some extent). Previous decisions like Bronner and Magill employed a standard that sought to balance competition concerns with incentives to invest. Sadly, Android Auto embodies a trend in the European Union in which the legal standard for getting (forced) access to another company’s assets is becoming more lax with every decision.
This is the first contradiction that I wanted to underscore. The ECJ recognizes in Android Auto that:
… it is generally favourable to the development of competition and in the interest of consumers to allow a company to reserve for its own use the infrastructure which it has developed for the needs of its business. If access to a production, purchasing or distribution facility were allowed too easily, there would be no incentive for competitors to develop competing facilities. In addition, a dominant undertaking would be less inclined to invest in efficient facilities if it could be bound, at the mere request of its competitors, to share with them the benefits deriving from its own investments… (Paragraph 42, emphasis added).
And even more to the point, the judgment also recognizes the “need for undertakings in a dominant position to continue to have an incentive to invest in developing high-quality products or services, in the interest of consumers.” (Paragraph 43). But by arbitrarily construing that “the needs of its business” is limited to exclusive use by the platform owner itself, the ECJ is depriving of all protection the incentives it appears to take into account.
Incentives Created by Android Auto and the DMA
The second contradiction the judgment creates concerns the incentives it provides to make platforms more (or less) open. The imposed fine and remedies, of course, are a direct incentive for Google to open its platform to Enel X; JuicePass is now available on Android Auto and has been since at least 2024. Other digital platforms also now face a powerful incentive—the threat of similar fines—to be more careful in analyzing access requests from application developers.
But Android Auto also creates an unintended incentive: if an initial decision to partially open a platform makes it possible for enforcers to mandate complete openness, then digital platforms now have an incentive to remain closed.
How is this consistent with the Digital Markets Act (DMA)? In order to fulfill its goal to make digital markets more “contestable and fair,” the DMA introduced several obligations for “gatekeepers” to make their platforms more open. According to Article 5 of the DMA, for instance, once a platform is labeled as a “core platform service” (CPS), its owner “shall not prevent business users from offering the same products or services to end users through third-party online intermediation services at prices or conditions that are different from those offered through the online intermediation services of the gatekeeper.” This means, for example, that Apple and Android have to allow developers to offer alternative channels to subscribe to their services outside the platforms’ app stores, which affects the commissions the platforms can charge.
Gatekeepers also “shall allow business users, free of charge, to communicate and promote offers, including under different conditions, to end users acquired via its core platform service or through other channels, and to conclude contracts with those end users, regardless of whether, for that purpose, they use the core platform services of the gatekeeper.” This amounts to a taking of advertising space. Moreover, it diminishes the incentives to create and maintain the platform, even if it would—in the short term, at least—facilitate entry of new actors. Alas, the Android Auto judgment provides incentives for platforms to remain closed.
There’s a reasonable counterpoint here: closed business models are rarely successful. As I have described elsewhere, the initial iPhone operated within a completely closed ecosystem, denying users the ability to add new applications. Apple recognized the limitations of this approach and, in 2008, launched the App Store, albeit without fully embracing openness. This strategic pivot made sense, as broader access stimulates application development, leading to a greater volume and higher quality of apps. Therefore, markets would counterweight the incentives created by Android Auto and no platform is likely to adopt a fully closed model.
But as we have explained, identifying the optimal mix of open and closed usually requires a platform to experiment through multiple iterations. Android Auto makes such iteration much more difficult, as the decision to open a platform to any degree is effectively irreversible and may lead to more (forced) openness than the platform desires. In other words, Android Auto pushes companies toward inefficient platform-design decisions, particularly when they are faced with evolving consumer preferences.
Consider a hypothetical example. Although still a niche market, there’s increasing demand for minimalist and focus-oriented “dumb phones” or “minimal phones.” Imagine a company launches “MiniPhone,” a sleek, closed ecosystem featuring only a few essential in-house apps—say, an e-reader, telephony, and email client. Despite their preference for simplicity, even minimalist users may appreciate the convenience of accessing vast music libraries, thus prompting MiniPhone to consider allowing Spotify on the device.
From a user-experience perspective, this would be a sensible and efficient move. But if MiniPhone opens up to Spotify, would it then be legally obligated to offer access to Apple Music, YouTube Music, Deezer, and Tidal as well? Would such openness actually benefit consumers on a device with limited memory and intentionally restricted functionality? What if “MiniPhone Music” is a key source of revenue that helps subsidize hardware costs?
Perhaps users also want access to Google Maps. Would that trigger the need to support Apple Maps, Bing Maps, and Waze, too? In fact, the Android Auto ruling might even go one step further: platforms may legitimately wonder whether opening up to music apps would require the host to accept any type of apps, with all of the cost, performance, usability, security, and price risks and consequences that would entail.
In practice, MiniPhone’s legal team would likely raise concerns about the risks of selectively allowing access. The Android Auto ruling encourages a binary choice—fully open or fully closed—when, in reality, the most efficient outcome may lie somewhere in between. This legal uncertainty could push companies toward overly cautious, inefficient decisions that ultimately harm both innovation and the consumer experience.
These risks are particularly acute in the EU, because of the exceedingly narrow market definitions drawn in competition cases by the European Commission (often unchallenged by the courts). Indeed, the “MiniPhone ecosystem” itself could be considered the entire market and found dominant, even though it is competing to offer an alternative to Apple, Android, and other device makers.
Conclusion
The Android Auto decision highlights a deepening paradox in European antitrust enforcement: the ambition to “tame” dominant digital players risks undermining the very incentives that fuel innovation and consumer benefit in digital markets.
By loosening the “indispensability” requirement in refusal-to-deal cases, the ECJ is hindering the incentives to build, maintain, and curate high-quality digital infrastructure. This not only destabilizes established legal principles in EU competition law that sought to balance competition with incentives to innovate, but also sends a confusing signal to digital-platform operators: openness may now be penalized, rather than rewarded.
This legal trajectory also clashes with the EU’s broader digital-policy goals, particularly those articulated in the DMA. While the DMA aims to encourage more open and interoperable ecosystems, the Android Auto ruling paradoxically encourages firms to keep platforms closed.
Of course, openness is not inherently good. While “openness,” “interoperability,” and “portability” are now commonly asserted as goals of various public-policy initiatives, their proponents often omit an important part of the story. These are proposals for forced openness, interoperability, or portability, which ignores that markets already provide a good amount of all three.
