The judgment of the Court of Justice of the European Union (CJEU) in Case C‑233/23 Alphabet and Google v AGCM (Android Auto) (the ‘Judgment’) may seem like the only possible outcome in light of recent case law on Articles 101 and 102 of the Treaty on the Functioning of the European Union (TFEU), as had already been anticipated by Advocate General Medina in her elegant opinion (the ‘Opinion’—discussed in a previous post). However, the way in which the Italian Council of State referred the questions, which had to be reformulated (Judgment, paras 33-36; Opinion, para 26), and, for instance, the draft guidelines on exclusionary abuses (particularly paras 96 or 163, discussed here and here), and the ensuing discussion at the 13 February workshop for that matter, show that surprising hesitation still remains in embracing the analytical framework developed in case law. This opportunity might have been taken by the Court of Justice to raise the bar for finding a refusal to deal to be prima facie legitimate.
On the one hand, over a decade of case law on Article 101 TFEU, from Allianz Hungary (C‑32/11) and Cartes Bancaires (C‑67/13 P), on how the analysis of anticompetitive effects is unnecessary if agreements and concerted practices reveal a sufficient degree of harm as to be restrictive by object came full circle in Servier (C‑176/19 P), Krka (C‑151/19 P), and BNP/BIC Português (C‑298/22). These rulings explain that the cursory analysis of the anticompetitive object is merely a preliminary check of whether the nature of the conduct—i.e., its objective features (content and economic and legal context, as opposed to subjective considerations or intent)—is enough to demonstrates that its objective logic or ‘objective aims’ is to cause competitive harm. If this epistemological but objective element is present, the resulting competitive harm (the lessening of competition inherent in then effects of agreements and concerted practices—for all, Dole Food, C-286/13 P, para 126) can be imputed to the practice through causation in law.
Causation in law means that there is no need to isolate the practice as the sole cause of competitive harm through counterfactual analysis; it suffices that the practice could have been objectively foreseen to cause competitive harm due to its nature (object) or effects by the company, thereby making competitive harm the plausible restrictive objective logic behind the company’s engaging in it. This plausibility standard of proof entails the possibility of rebuttal by presenting one alternative explanation for the practice (for all, Google Shopping, C-48/22 P, para 227).
That is why counterfactual analysis is required only to establish the appreciable effects’ instrumentality in causing competitive harm in cases of restrictions by effect—i.e., when the plausible restrictive objective logic cannot be inferred from the anticompetitive ‘objective aims’ (MasterCard, C‑382/12 P, paras 165-170). If effects are appreciable and instrumental in causing competitive harm, then competitive harm is objectively foreseeable, and the plausible restrictive objective logic of the agreement or concerted practice may be imputed solely on the basis of effects when the anticompetitive object is ruled out. This is possible under Article 101 TFEU because the effects of agreements and concerted practices inherently entail a departure from normal competition (a reduction in competition). Consequently, under Article 101 TFEU, object (or nature) and effects serve as alternative substantive methods of objectifying the restrictive logic of the practice in order to impute competitive harm based on causation in law. The proof of this is that the lack of effects is not a defence for practices deemed to have an anticompetitive object (Amory, Gerard, Smuha, van de Walle, 2017).
On the other hand, since Generics (C‑307/18) and Slovak Telekom (C‑165/19 P), and arguably since TeliaSonera (C-52/09) and Deutsche Telekom (C-280/08 P), all the way to SEN (C-377/20), Unilever (C-680/20), and Google Shopping (C-48/22 P), the case law on Article 102 TFEU has made clear that, unlike under Article 101 TFEU, the plausible restrictive logic of unilateral conduct must be objectively inferred simultaneously from its nature and effects. This is because, unlike agreements and concerted practices, the effects of unilateral rent-seeking conduct are usually normal competition, as they purge the market of inefficient competitors and reward innovation with windfall profits. Therefore, only if the objective logic of the conduct is to foreclose equally or more efficient competitors or to profiteer from their absence does the resulting competitive harm become restrictive (Solano Díaz, 2024a).
Thus, exclusionary conduct is only abusive if it is plausible (subject to rebuttal) that a hypothetical equally efficient competitor cannot simultaneously (i) obtain a comparable advantage by engaging in the same conduct (nature or ‘deviation from competition on the merits’) or (ii) offset the advantage by adopting a different strategy (effects or ‘capability to foreclose’). Otherwise, the conduct does not make anticompetitive sense for the dominant company to engage in (Solano Díaz, 2024b). This includes the ‘killer abuse’ scenario, where the conduct prevents less efficient competitors from reaching the scale required to become hypothetical equally efficient competitors (Superleague, C-333/21, para 129; Google Shopping, C-48/22 P, para 165).
As a result, the nature and effects of conduct are interconnected in exclusionary abuses: the more easily its restrictive objective logic can be inferred from its nature, the less evidence is needed to demonstrate a potential effect to the plausibility standard. This explains why there are instances of so-called ‘abuses by object’ (improperly because at least the capability of producing effects must be shown unlike under Article 101 TFEU) where the CJEU has found that the ‘abusive nature’ of conduct gives it an ‘intrinsic capacity […] to foreclose competitors which are at least as efficient as the dominant undertaking’ (Intel, C‑413/14 P, paras 126, 140, and 142; Google Android, T‑604/18, para 795). At the other end of the spectrum, there is conduct such as prima facie legitimate refusals to deal, whose nature is not artificial and can therefore only be abusive ‘by effect’. That is their restrictive objective logic can only be inferred from particularly harmful effects, as in the exceptional circumstances of the essential facilities doctrine (elimination of competition due to the indispensability of dealing).
Against this background, the Court of Justice has clarified in the Judgment that refusals to deal are subject to a preliminary assessment of whether they depart from the logic inferred from the content and context of the dominant company’s conduct (Judgment, paras 44-47; Opinion, paras 35-39). This is equivalent to the cursory analysis used under Article 101 TFEU to establish the artificial nature (or object) of agreements and concerted practices so that proof of appreciable and foreseeable effects is unnecessary. Conversely, under Article 102 TFEU, the purpose is to rule out such artificial nature in order to subject the refusal to a full effects analysis under the essential facilities doctrine. This approach strikes a balance between the economic interests in short-term competition benefits and long-term incentives to innovate, as well as between the legal interests in free competition and the freedom to conduct business and the right of property (Judgment, paras 39-43; Opinion, paras 42-43).
The solution resembles that applied to reconciling Member States’ tax sovereignty with the avoidance of distortions caused by tax state aid: Member States are free to set up a tax framework pursuing the objectives that they see fit. However, once they do, any differences in treatment between taxpayers that is not justified by the logic of that framework would be a derogation therefrom and thus selective (for all, British Aggregates, C-487/06 P, para 88; or Commission v Netherlands, C 279/08 P, para 62). By the same token, once the dominant company defines its overarching dealing policy, any subsequent case-by-case refusal, obstacle or delay that deviates from it would not be considered a legitimate change of that policy but to have an artificial logic precisely because it is a derogation from the general policy, especially if adopted against a potential competitor (Judgement, paras 46 and 51).
In casu, the content of the conduct is Google’s refusal to develop, upon Enel’s request, a template application that was necessary to access a car’s integrated display through Google’s mobile application Android Auto. The template was necessary for Enel to display its charging station mapping mobile application on car’s integrated displays but not strictly essential for marketing it for use on handheld devices (Judgement, para 36; Opinion, para 21). The context of the conduct is that Google’s ‘economic model which applied to the development of [Android Auto]’ was to facilitate third-party access by making it deliberately open to be freely shared, and that the one exception is made in relation to an application that happened to potentially compete with Google’s applications Google Maps and Waze (Judgement, paras 7, 46 and 51; Opinion, paras 8, 38 and 56).
Therefore, Google’s conduct was not considered the exercise of its freedom to change its general commercial policy in relation to Android Auto but rather a specific derogation from that commercial policy and one that was questioned by leveraging concerns. In those circumstances, the artificial nature of the refusal cannot be ruled out in the preliminary assessment, and thus the essential facilities doctrine does not apply (Judgement, 52; Opinion, para 56). Instead, the refusal is subject to the general two-fold test that requires confirming its artificial nature (or deviation from competition on the merits) and its potential effect based on its aptitude to foreclose equally efficient competitors (Judgement, para 54).
The emphasis that the Court of Justice devotes to insisting on the well-established case law that the mere capability to foreclose is enough to meet the requirement to prove potential effects is noteworthy (Judgement, paras 55-58; Opinion, paras 52-53). This may be a hint at the abovementioned interconnected character of the nature-effect duality in inferring the restrictive objective logic of the conduct under Article 102 TFEU (unlike under Article 101 TFEU) which is crucial in the Google Android appeal (C-738/22 P). This shortcut, allowing for so-called ‘abuses by object’, may make sense where the restrictive objective logic of the conduct finds solid ground in its artificial nature. It could be particularly useful to avoid complex counterfactual analysis where the theory of harm is the elimination of competitors before they grow equally efficient (‘killer abuse’), as seems to be what the Court of Justice had in mind in the case at stake (Judgment, para 60).
The rest of the Judgment focuses on the possibility for the dominant company to invoke an objective justification (or efficiency defence), which is obviously available regardless of whether the essential facilities doctrine or the general two-fold test is applied. Indeed, the presumption of innocence requires that any plausible alternative explanation prevail over the plausible restrictive objective logic of the conduct, thereby breaking the causation in law. The narrow scope of the objective justification in this case mirrors the fact that the refusal was a derogation from a system that Google had designed to be open. Therefore, if Google’s general policy is to make application templates available, then not creating one within a reasonable timeframe and charging the corresponding fees would hardly be an alternative explanation to a restrictive objective logic (Judgement, para 74; Opinion, para 65). Finally, the question about the need to define precisely a downstream market is answered in the negative by reference to the case law on prima facie legitimate refusals to licence intellectual property rights preventing the emerge of new products (for all, IMS Health, C-418/01), which matches the factual situation of a ‘killer abuse’.
In conclusion, I welcome the CJEU’s clarification that a preliminary Article 101-like assessment of the content and context of a refusal to supply can determine whether its artificial nature can be ruled out so that it is subject to a full effects analysis under the essential facilities doctrine. Otherwise, the general two-fol test based on relative efficiency applies. The exceptional circumstances of the essential facilities are also based on the relative efficiency of hypothetical competitors because it is obvious that no competitor, regardless of how efficient it is, can derive a comparable advantage to that obtained from refusing indispensable dealing or offset it by other means if the result is potential elimination of competition. Such a structured analytical framework based on a relative plausibility standard does not only find full legal support in case law but is also a good candidate for an optimal legal rule from the economic perspective (Gal, Padilla, Piccolo, 2025).
However, we can question whether the bar for determining whether the refusal is prima facie legitimate is set too high. I tend to think it would be appropriate if a decision by Google to reserve the whole Android Auto platform to its own apps would have been prima facie legitimate and subject to the essential facilities doctrine as an ‘abuse by effect’. There may be cases where the regulatory framework or the fact that the dominant company did not privately develop the assets from its use form a context where the dominant company does not have the possibility to freely define its dealing policy (Slovak Telekom, C‑165/19 P, paras 46 and 58; Deutsche Telekom, C-152/19 P, paras 46 and 58; Lietuvos geležinkeliai, C‑42/21 P, para 75; Commercial Solvents, 6 and 7-73, para 24; Sot. Lélos kai Sia, C-468/06 to C-478/06, para 49). In any other case, the dominant company’s decision to change its dealing policy should be subject to the essential facilities doctrine unless it is a specific derogation from it. This clarification is unfortunately missing in the Judgement.
