Taming Digital Monopolies: A Comparative Account of the Evolution of Antitrust and Regulation in the European Union and the United States

This paper offers a comparative account of how the European Union and the United States surveil dominant internet players in light of recent enforcement efforts by U.S. antitrust agencies and ongoing discussions about regulating digital giants in both jurisdictions. After setting out themes for comparative analysis, the paper turns to the two actions initiated in the United States against Google: the one filed by the Department of Justice (DOJ) is similar in focus to the European Commission’s Android decision and the one led by the State of Texas focuses on advertising markets in a manner similar to the European Commission’s AdSense decision. We observe that while there are similar intuitions about anticompetitive conduct in the manner both jurisdictions address the issues, the framing of the competition problem by the U.S. agencies is more sophisticated in relation to the understanding of the markets, the theories of harm, and the design of forward-looking remedies. The paper then compares the Commission’s proposal for a Digital Markets Act with several Bills proposing platform regulation presently discussed in the United States, examining what the two systems have in common, what they may learn from each other, and what regulatory gaps remain.


I. Introduction
While the European Commission (the Commission) and some national competition authorities in the EU have been pursuing firms that are said to dominate technology markets in the internet era for over a decade, enforcement in the United States has been less forthcoming until late 2020 when the Department of Justice (the DOJ) and several states filed a complaint against Google. 1 Since then, further actions have been commended by the Federal Trade Commission (the FTC) and some state attorneys general. Furthermore, the Biden administration appears eager to step up antitrust and regulatory efforts generally and one of the priorities is addressing the power of big tech firms. 2 This increased attention by U.S. agencies motivates this article. We carry out a comparative account of the EU and U.S. efforts to address the market failures that arise in big tech markets. We consider the application of antitrust first and the emerging regulatory framework second.
The paper is structured in the following manner. In Section II(A), we identify the comparative law discourses that have informed the voluminous scholarship comparing EU and U.S. antitrust. The predominant narrative that emerges from this is that the EU competition law standards should align to those of the United States because the former are too formalistic and risk too many Type I errors. This narrative had a significant impact in the turn to what the Commission referred to as a more economics-based approach to competition law enforcement. 3 In particular, for the purposes of this paper, this narrative led the Commission to rethink its approach to the regulation of dominant firms under Article 102 TFEU, 4 which is the rough equivalent of the monopolization provision found in Section 2 of the Sherman Act. However, we also observe two other comparative law themes that point in a different direction: the first is that globally other jurisdictions appear to apply competition law in a manner more closely aligned to the European Union than the United States and that, on some issues (e.g. refusals to deal), the approach found in U.S. antitrust is an outlier when compared to other systems. Accordingly, while there may have been a degree of methodological convergence between the two jurisdictions under consideration, there remain reasonable disagreements on the application of competition law. The second is that those U.S. scholars who advocate a departure from what they see as the current conservative stance of U.S. antitrust are suggesting reforms that would move U.S. antitrust closer to the EU model. Moving from discourses to cases, in Section II(b), we compare the approach of the EU and U.S. agencies when applying competition law to three cases which have been closed in the United States: Microsoft's exclusionary practices in the browser war, Intel's exclusionary strategy in the market for microprocessor chips, and Google's practices in the search market. In each of these three instances, we compare how the agencies have pursued the cases. This comparative account of antitrust enforcement raises two themes: the first is the substantive standard of assessment that is applied, the second is the manner in which the facts are understood and how the case is focused. This trio of cases reveals that while there is some alignment on substantive standards, there are fairly major differences in how the cases are framed by the two agencies. 5 The resulting appeals in some of these judgments are also relevant in establishing how to apply antitrust law to digital markets and confirm the divergence of the two approaches.
In Section III, we make two comparisons. One consists in looking at the DOJ's case against Google and the Commission's decision in Google Android, which confronts similar issues. The second entails the States' case against Google's practices in the advertising market and the Google AdSense decision by the Commission. 6 Based on the framework in Section II, we show how, even at this preliminary stage, the U.S. agencies are addressing the competition issues in a more holistic manner and we try and account for why this difference of perspectives exists.
In Section IV, we discuss briefly the role of remedies. The main point here is that, while both systems shun the option of structural remedies for different reasons, there is still room to design better behavioral remedies. This leads us to discuss, in Section V, whether the regulatory efforts in the two jurisdictions can prove a more effective means of addressing the market failures that enforcers see in digital markets. In the different legislative proposals, we observe a certain convergence between the two jurisdictions where there is a willingness to go beyond classic antitrust law and tackle unfairness and contestability in digital markets. Having said that, there are some aspects where both systems could learn from each other, such as the designation of platforms covered by the regulations or the configuration of the conducts prohibited therein, in particular, we recommend that the European Union turn its attention to merger control and the United States turn its attention to a different, more regulatory style of enforcement which is inherent in the Digital Markets Act (DMA). 7

A. Comparative Law Discourses
While in the post-war reconstruction of Europe, the United States put some political pressure on European states and the draftsmen of the Treaty of Rome of 1957 which established the European Economic Community to insert provisions regulating cartels and monopolies, the texts that emerged were not carbon copies of the Sherman Act. 8 European Union competition law, since the beginning, had its distinct identity, drawing at times on the legal traditions of the Member States. 9 Early scholarship focused on trying to understand and compare the two systems, but immediately there was a sense that the Europeans had something to learn from the more well-developed jurisprudence of the United States. 10 Indeed, in the first major case to reach the European Court of Justice (ECJ) saw the appellant refer to a judgment of the U.S. Supreme Court to support an argument that not all vertical restraints should be condemned. 11 Since the 1980s, calls to align EU competition law to that of the United States increased. This was largely driven by the greater presence of U.S. firms in the EU internal market. By this time, U.S. antitrust had become much more relaxed and calls for a more business-friendly attitude by the Commission mounted. 12 Furthermore, some clashes began to emerge when the two jurisdictions considered mergers The Advocate General found this judgment helpful in supporting his view that in order to determine whether an agreement restricted competition one had to carry out 'a comprehensive examination of their economic repercussions.' (p. 358). However the ECJ demurred and found that a vertical restraint imposing absolute territorial protection infringed competition law. This created the first cleavage between EU and US antitrust laws: the former concerned about agreements that partitioned national markets, the latter more concerned about the economic consequences of firm conduct. 12. Barry E. Hawk, The American (Anti-trust) Revolution: Lessons for the EEC?, 9 EUR. COMPETITION L. REV. 53 (1988) is one of the early papers. a few years later the same author issued a much more scathing attack on the Commission's approach to vertical restraints: Barry Hawk, System Failure: Vertical Restraints and EC Competition Law, 32 COMMON MKT. L. REV. 973-89 (1995). This paper proved to be influential in shaping the reform of EU competition law. More generally, an epistemic community of lawyers from the US and the EU met regularly to push for greater convergence. that were notified in both systems. The Commission hesitated in authorizing the merger between Boeing and McDonnel-Douglas until political pressure was brought to bear on it, while it blocked the merger between GE and Honeywell despite protestations by U.S. antitrust agencies. 13 However, it is imprudent to view these two examples as symptoms of a greater malaise: for most mergers that are notified in the two jurisdictions, we observe cooperation between the agencies. 14 This was institutionalized by a bilateral agreement to facilitate cooperation in 1991. 15 Twenty years later, the two sides also established a more specific framework for cooperation in the field of mergers to facilitate coordinated enforcement. 16 Such proximity facilitates convergence between the two systems in terms of substantive assessment. Indeed, moving swiftly to the present, the question is not whether the European Union should align to the United States but rather how far there has been convergence between the two systems. 17 However, there are two other comparative law discourses that can be found in the scholarship that have had less resonance in policy circles but which will become key if the U.S. antirust system will evolve away from the current standards of analysis. The first is the observation that the U.S. antitrust stance in many fields means it is the odd one out: most other jurisdictions have rules that are more aggressive than those found in other parts of the world. 18 Indeed, it has also been noted that many jurisdictions outside the European Union and the United States have followed the EU's antitrust model. 19 For all the talk of convergence between the two jurisdictions then, it appears that there remain aspects of antitrust that diverge. We will see specific examples of this in the remainder of the paper, but two might be signaled here. First, the EU recognizes margin squeeze as a discrete example of abuse of dominance, while the U.S. Supreme Court has refused to recognize this as a stand-alone offense. 20 Second, the ECJ's approach to refusals to deal by dominant firms is wider than that which may be found in the U.S. Supreme Court. 21 More generally, as we will see the EU standards tend to be more expansive. The second, and related, type of scholarship (which is not directly comparative in style) consists of attempts to redirect U.S. antitrust along lines that feel more European. 22 Perhaps the most notable policy manifestation of this is a bill in the state of New York, proposing a "Twenty-First Century Anti-Trust Act." 23 This contains two provisions to regulate monopoly power. One replicates section 2 of the Sherman Act, while the other draws on Article 102 TFEU, making in unlawful "for any person or persons with a dominant position in the conduct of any business, trade or commerce or in the furnishing of any service in this state to abuse that dominant position." In a similar vein, some scholars are proposing moving from a consumer welfare standard to a standard-based on considering whether the conduct under antitrust scrutiny harms the competitive process. 24 This stance echoes the approach taken by the EU Courts. In 2009, the Court of Justice of the European Union (ECJ) was confronted head-on with the question whether the litmus test for legality was harm to consumers and it responded by saying that the EU rules are concerned by more than just consumer welfare: there is nothing in the wording of Article [101 TFEU] to indicate that only those agreements which deprive consumers of certain advantages may have an anti-competitive object. Secondly, like other competition rules laid down in the Treaty, Article [101 TFEU] aims to protect not only the interests of competitors or of consumers, but also the structure of the market and, in so doing, competition as such. 25 Granted, this passage is not entirely clear but it signals that the ECJ is not yet completely ready to embrace exclusively the consumer welfare standard. A difficulty that the ECJ has had to confront is how to ensure that, by protecting the competitive structure of the market, one does not end up protecting competitors. As we will explain below, one solution has been to deploy the as-efficient competitor standard.
There are two takeaways from these two less mainstream comparative accounts. The first is that, for all the talk of convergence, there remains room for reasonable disagreement about how the law is to be applied and what purposes it serves. The European Commission might use today similar analytical tools to those found in the United States but results in cases can continue to differ. The second is that, if anything, convergence might in future move in the other direction: perhaps the United States has something to learn from the EU approach. This prospect is particularly so since the Commission has a first mover advantage in certain cases, as we show in Section 3 when considering the challenges against Google brought in the U.S. courts, or when it comes to regulating digital platforms as discussed in Section 5.

B. Comparing Like Cases
We now move to consider three cases where we have seen public enforcement by both the European Commission on one hand and the DOJ or the FTC on the other hand. We start with the Microsoft saga, which was a key case for both agencies in learning about how to apply competition law in high-tech markets which have distinct economic features. 26 We then turn to Intel, which is selected because this was, from the European Commission's perspective, a key test case for its new approach to analyzing unilateral conduct. We move on to consider Google Shopping and the diverging views of the European Union and United States on whether self-preferencing should be controlled. The purpose of this review is not to re-tell the entire story of these cases, but to signal the similarities and differences in the way the two agencies dealt with the cases and how the courts have used these cases to advance the application of antitrust in technology markets. What is intriguing is that there are hardly any public statements by agency officials that reveal the extent of cooperation or coordination between the European Union and the United States in these three case studies; even if the markets under consideration included both jurisdictions, the timing of the investigations was not dissimilar and the underlying facts are the same. Rather, it appears that each agency pursued the cases with their own agenda. These case studies allow us to go deeper in comparing the two systems and, as will be demonstrated, they also serve to show that, even when the two jurisdictions agree on the result, the U.S. system appears to frame the competition analysis in a manner that is more likely to resolve the competition concerns by a more holistic understanding of the relevant issues.

Microsoft.
In Microsoft, the DOJ moved first in 1998 by claiming that Microsoft engaged in a range of exclusionary practices designed to maintain its monopoly in the operating systems market. The DOJ considered that the Netscape navigator browser could become the basis of a software platform that would compete with Microsoft's operating system. 27 After protracted litigation, there was a settlement designed to prevent Microsoft from continuing its conduct and to facilitate competition in ancillary markets. 28 In contrast, the Commission's first Microsoft case (decided in 2004) was concerned with two narrower issues. 29 The first was that firms wishing to compete in the market for work group servers were unable to obtain sufficient information from Microsoft to ensure interoperability between their products and Microsoft's operating systems. This also formed part of the remedy agreed in the final judgment in the United States, with the judge in the United States observing that "in all likelihood, the requirement that Microsoft disclose and license the communications protocols utilized by its PC operating systems to communicate with Microsoft's servers is the most forward-looking provision in the Court's remedy." 30 The second Commission concern was that Microsoft was tying the operating system with Windows Media Player (which was not the main focus of the U.S. case but where the remedy also addressed this by leaving original equipment manufacturers (OEMs) free to install software on the computers they sold). 31 Like in the United States, the theory of harm was foreclosure of rivals but unlike the United States the premise of the case was not that these tactics were designed to protect Microsoft's monopoly. Rather, these practices risked extending the monopoly into other related markets. Ultimately the theories pursued were but a subset of the issues addressed in the U.S. case.
The Commission's interest in securing competition in the browser market emerged later on when it brought a second case against Microsoft arising out of a concern that the firm was tying its web browser Internet Explorer to its dominant client PC operating system. In this decision, the Commission took the view that while it appeared relatively easy for a consumer to switch to an alternative browser, there were a number of barriers to switching, "such as searching, choosing and installing such a competing web browser, which can stem from a lack of technical skills, or be related to the user's inertia. Judged retrospectively, the different angles through which the agencies attacked Microsoft appear to suggest that the U.S. challenge was more well-grounded. The DOJ focused on what might be termed the first browser war, emphasizing the importance of keeping this market open. It has been noted that while the antitrust rules moved too slowly to protect Netscape, the final ruling helped the emergence of new internet firms, including Google. 33 Conversely when it comes to media player, by the time the General Court issued its judgment affirming the Commission's decision, it appeared that the market had become competitive. 34 Moreover, the remedy imposed, which was to require that Microsoft sells two versions of Windows, one with and one without Windows Media Player was not a success as the prices of the two products were the same and very few consumers bought the unbundled version. 35 Conversely, the second Microsoft decision of the European Commission focused on what we might call the second browser war. The lack of interest in the United States in the second browser war is likely explained by the fact that the settlement in the U.S. case already facilitated the entry of browsers and indeed switching browsers was more prominent than the Commission believed. Unofficial statistics indicate that in the United States since 2009, the market share of Internet Explorer began to fall and it was overtaken by Safari and Chrome in 2013. 36 On the other hand, the Commission's focus on work group servers mirrored some of the remedies imposed in the U.S. settlement and while there was no formal cooperation in designing the procedures for access remedies, it has been noted that there were episodes of cooperation between the United States Technical Committee and the European Commission's monitoring Trustee, which had been set up to facilitate the implementation of this remedy. 37 2. Intel. In this case, the FTC and the Commission were more aligned on the competition concerns than in Microsoft. Intel was the world's leading computer chip maker and it used a range of clauses in its distribution agreements to make it difficult for competing chip makers to stay in the market. For example, it offered benefits to computer makers in exchange for their promise to buy chips from Intel exclusively or to refuse to buy chips from others. This foreclosed market access to Intel's major rival, AMD. Both agencies expressed concerns about this, the major difference is that the Commission issued a prohibition decision accompanied by a very large fine while the FTC settled. 38 However, the FTC settlement went further than prohibiting these practices by also observing that there was other conduct that should be controlled: in particular, that Intel should also amend its conduct to facilitate market access for producers of Graphic Processing Units (GPUs). Furthermore, the settlement identified some IP licensing issues between Intel and its main rivals, and the FTC requested that these were resolved. 39 One likely effect of this aspect of the settlement, according to the FTC, is that it would "ensure that these existing competitors can partner with third parties to create a more formidable competitor to Intel." 40 33. Jonathan B. Baker, the antitrust ParadiGm (2019), Ch. Again, in this example, we see that the FTC manages to cover a relatively wider palette of conduct than the Commission. It is not clear what drove the Commission to a narrower scope, so the observations here are speculative. Intel's major competitor in the GPU market was NVIDIA, a firm based in California so perhaps the Commission may have considered that it lacked a particular interest in pursuing this conduct since the FTC was addressing it. Likewise, the licensing issues may have been of greater concern in the United States because this is where the rivals are situated. This might account for the Commission's narrower focus of inquiry. Another reason why the Commission zoomed in on the commercial relations between Intel and its downstream customers may be because this was a key case for the Commission to explain its newly minted approach to assessing these agreements. The Commission's earlier handling of rebates has been one of the most criticized aspects of the application of Article 102 TFEU, leading many commentators to conclude that the Commission was protecting distributors but doing little to safeguard competition in markets. 41 While the Commission was investigating Intel, its officials were also busy re-imagining the approach to rebates. As a result, the decision in Intel is the first one where the Commission carries out an extensive economic analysis to determine whether the rebates granted by Intel were likely to exclude an as-efficient competitor, in line with the approach that it recommended in the Guidance Paper. It may thus be the case that the Commission preferred to focus its resources on the rebates rather than on the conduct of the dominant firm generally. It is worth noting that Intel appealed against the decision, which helped the ECJ reformulate its approach to Article 102 TFEU, as we will discuss in Section 2.2.5 below. However, a final ruling from the General Court is still awaited, more than a decade after the original decision. 42 3. Google Shopping. The first of the three Commission decisions challenging Google was about its exclusionary conduct in the search market. In brief, the theory of harm was that Google leveraged its dominance in the market for general search to exclude rivals in the market for vertical search. It achieved this by demoting vertical search providers in the search results and by promoting its own vertical search by using a more attractive and more prominent display. The intuition is not that different from Microsoft: Google feared a loss of users on its search page, with a concomitant reduction in data that it would be able to gather, and sought to promote its vertical search engine at the expense of rivals to retain the attention of users. To advance this case, the Commission relied on a fairly expansive reading of precedents to establish that Google's product design infringed Article 102 TFEU by marginalizing rivals in vertical search. 43 In contrast, in 2013, the FTC closed a two-year investigation against Google alleging that it had found no evidence supporting search bias allegations. 44 The principal motivating factor was the view that Google's product design benefited consumers and that the FTC should not second-guess a firm's product design decisions in these circumstances. 45  a radically different appreciation of the market by the two agencies. We do not engage with secondguessing the facts here. What matters is that the two decisions reveal a fundamental difference in the appreciation of the conduct of dominant firms. The European Commission's decision places significant emphasis on the overwhelming market power that Google has in the search market and it uses this as a basis for indicating that firms with such high degrees of dominance must be supervised actively. Indeed, the Commission went as far as to state that "[a] system of undistorted competition can be guaranteed only if equality of opportunity is secured as between the various economic operators." 46 The Commission continued by indicating that [c]ustomers and users should have the opportunity to benefit from whatever degree of competition is possible on the market and competitors should be able to compete on the merits for the entire market and not just for a part of it. 47 This reveals that the Commission remains concerned about the competitive process and affording rivals the chance to participate. The Commission also observed that Google had not invented comparison shopping, that it entered this market late, and its product became successful only after it had demoted other vertical search engines. 48 From this perspective, the European Commission also reads the evidence in a manner different from the FTC: the source of innovation is not only Google but also other rivals who compete against it. There is a long-standing commitment of the Commission in its cases against large tech firms that one must facilitate the entry of smaller rivals. This was explained clearly in the first Microsoft decision: Microsoft's research and development efforts are indeed spurred by the innovative steps its competitors take in the work group server operating system market. Were such competitors to disappear, this would diminish Microsoft's incentives to innovate. By contrast, were Microsoft to supply Sun and other work group server operating systems with the interoperability information at stake in this case, the competitive landscape would liven up as Microsoft's work group server operating system products would have to compete with implementations interoperable with the Windows domain architecture. Microsoft would no longer benefit from a lock-in effect that drives consumers towards a homogeneous Microsoft solution, and such competitive pressure would increase Microsoft's own incentives to innovate. 49 This attitude accounts perhaps for the biggest gap between the U.S. and EU systems of antitrust: the former showing greater trust in the procompetitive efforts of a firm with market power as well as circumspection about the law's capacity to remedy such conduct if found illegal, while the latter favoring the competitive process by stimulating market access.

Some Judicial Convergence.
Turning to the appeals against the adverse decisions discussed above, the Microsoft judgment by the Court of Appeals of the District Court of Columba Circuit stands as an authoritative restatement of the legal standard for monopolization and how this is to be applied in hightechnology markets. As we will see when discussing the two Google cases, the principles found in this judgment inform the way plaintiffs have framed their case in the United States. From the perspective of EU Law, the Microsoft, Intel, and Google judgments are also landmark cases that have shaped the way in which the Commission has pursued digital markets. 50  far the two systems have converged on the legal standard applicable to unilateral conduct. It is beyond the scope of this paper to discuss every nuance of EU and U.S. antitrust as applied to unilateral conduct, but three themes that emerge from the U.S. v Microsoft judgment are relevant for a comparative account.
a. The Rule of Reason. The Microsoft judgment's main contribution is to explain the application of the rule of reason in Section 2 cases. This is a three-step analysis, in brief: first, an anticompetitive effect has to be shown, this means a showing that there is harm to the competitive process which thereby causes harm to consumers. The burden of showing this effect is on the plaintiff. Second, if the plaintiff makes a prima facie case, the defendant may offer procompetitive justifications for its conduct (e.g. the conduct is efficient or the conduct makes the product more appealing to consumers). If this is successful, then the burden shifts back to the plaintiff to rebut this claim. Third, if the claim is unrebutted, then the plaintiff must show that the harm caused by the conduct is greater than the benefits. 51 Conversely, the EU case law has generally been less focused on requiring the defendant to show anticompetitive effects. However, the judgment in Intel provides for some convergence by suggesting what, in U.S. parlance, might be described as a quick look assessment. In reviewing the question whether a rebate offered by a dominant firm in exchange for exclusivity the ECJ held that while the Commission could base its claim on a mere showing of exclusivity, the defendant was entitled to rebut this by showing that the rebates are unlikely to foreclose markets. If this showing is made, then the Commission would have to carry out an effects-analysis to establish that the rebates would likely exclude a rival as efficient as the dominant firm. Even when likely anticompetitive effects are shown, the defendant may plead an efficiency defense and the Commission must then examine if the efficiencies outweigh the harm to competition. It remains to be seen whether this approach applies to other types of exclusionary conduct, but it provides some instances where the Commission will be asked for a full effects analysis, providing an analytical framework similar to that found under the rule of reason. 52 b. Specific Guidance on Abuses. Related to the effects assessment, the Court of Appeals also provides some guidance on two specific forms of conduct: exclusivity agreements and tying. For the purposes of exclusivity agreements, the Court clarifies that it is for the plaintiff to show that the scope of the exclusivity agreements forecloses a significant proportion of the market in order for this to be shown to be anticompetitive. 53 For the purposes of tying, the Court explained that the rule of reason was also to be applied on the facts of this case. While tying is sometimes a per se offense, the Court was not convinced that there was sufficient knowledge about software markets to state that the sale of a package of software was conclusively anticompetitive. 54 Conversely, the General Court in Microsoft was less demanding on the Commission's standard of proof in tying cases, indicating that tying could be condemned by showing foreclosure effects on rivals. 55 This explains the reaction of then-Assistant Attorney General for Antitrust Thomas Barnett who said that "the standard applied to unilateral conduct . . . rather than helping consumers, may have the unfortunate consequence of harming consumers by chilling innovation and discouraging competition." 56 Likewise, the judgments in Intel and Google Shopping place more emphasis on the 51 c. Causation. Finally, the U.S. Court of Appeals in Microsoft also clarified the causation standard. It disagreed with the defendant's argument that the plaintiff must show that monopoly maintenance is precisely attributable to the conduct in question. Rather, the standard for causation is as follows: (1) whether as a general matter the exclusion of nascent threats is the type of conduct that is reasonably capable of contributing significantly to a defendant's continued monopoly power and (2) whether Java and Navigator reasonably constituted nascent threats at the time Microsoft engaged in the anticompetitive conduct at issue. 58 This articulation of the standard of causation is important for two reasons: first, it explains a recurring theme in many of the current cases brought against big tech platforms: that these firms thwart the competition of nascent threats to its dominance. The fact pattern identified in Microsoft is analogous to that animating the current litigation. The second is that in a prospective inquiry about the impact of a form of conduct showing the capacity to exclude suffices. This perspective is broadly comparable to the causation standard that the ECJ has crafted in its jurisprudence where what suffices is a showing of likely competitive harm. 59 In sum, the case law reveals a degree of convergence regarding the analytical method to be followed: in both jurisdictions, for many forms of unilateral conduct, it is for the plaintiff to establish likely anticompetitive effects and a showing that the conduct can reasonably be said to cause the harm. However, effects are insufficient without proof of anticompetitive conduct and here some divergences remain, for example, on how to test for illegal tying and possibly on when discrimination against rivals is anticompetitive.

III. The Google Cases
We now turn to two ongoing cases brought in the United States against Google. These are selected because they bear a remarkable similarity to earlier Commission decisions: the DOJ's lawsuit against Google is analogous to the Commission's Google Android case. 60 The case brought by Texas is very closely aligned to the Google AdSense decision. 61 In each section, we sketch the European Commission's decision and then compare this with the available material regarding the proceedings in the United States. For the U.S. cases, our source is principally the statement of complaint, which is obviously framed in a manner that allows the plaintiff's case to be viewed in the best light possible. However, our analysis is not about what the right facts are: we do not second-guess the plaintiffs nor the Commission's factual findings. Our focus is on how the economic and legal issues are framed.

A. Google Android: Similar Premises
In Google Android, the Commission concluded that Google had infringed Article 102 TFEU by engaging in several practices to consolidate its dominant position in the market for general internet search services. The first practice was that Google had tied its Google Search app with the Play Store. This meant that OEMs could preinstall the Play Store on their Google Android devices only if they obtained the license and had preinstalled a bundle of services called Google Mobile Services (GMS), which included the Google Search app. Preinstallation is key for search engines because consumers tend to use what is already loaded on their device. Moreover, it was not possible to uninstall the Google Search app, making switching less likely. Google was dominant in the market for Android app stores, and given that rival search engines could not survive through other distribution channels, the Commission concluded that this granted Google a significant competitive advantage. In addition, Google had also tied the Google Search app and the Play Store with Google Chrome, an internet browser. The Commission considered this practice was also an abuse of dominance. Unless OEMs preinstalled Google Chrome, they could not preinstall the other two apps, and users could not obtain these apps without obtaining Google Chrome simultaneously. Likewise, the Commission concluded that non-OS-specific mobile web browsers could not offset the competitive advantage that Google was gaining, thanks to this practice, which allowed Google to maintain its position in the market for general search services and deterred innovation. Google argued that users could download competing apps, but the Commission was not convinced: the number of downloads of competing apps was far lower than the use of preinstalled apps, and OEMs are also unlikely to preinstall competing apps. 62 The second practice was that Android device manufacturers could not preinstall the Play Store and the Google Search app unless they accepted antifragmentation obligations (AFAs), which prevented then to make use of Android forks. 63 According to the Commission, these forks constituted a credible competitive threat to Google because the investment required to develop an Android fork is much lower that the investment required to develop a completely new OS for smartphones, while apps can adjust more easily to run on Android forks than to run on new OS. Thus, the AFAs prevented developers of Android forks from finding distribution channels that would enable a rapid scaling up of their operations and that could use competing general search engines, which would constitute a serious threat to Google. The Commission concluded that this practice helped Google to maintain its position in the market for general search services and, given that superior Android forks could not be distributed, consumer choice and innovation were also affected.
Third, the Commission examined the exclusivity payments granted by Google, which were based on a revenue share to OEMs and Mobile Network Operators (MNOs) if they preinstalled no competing general search services. 64 The amount paid by Google could not be matched by competitors because it was impossible to offer a sufficient amount of revenue to compensate them for the loss of Google's payments. The Commission concluded that the exclusivity payments deterred innovation because competitors did not have incentives to develop innovative features and therefore infringed Article 102 TFEU.
The case law of Article 102 TFEU allows for undertakings to escape the prohibition if undertakings show that there are objective justifications or efficiencies that justify the unilateral conduct. 65 Google raised some procompetitive arguments but all of them were rejected by the Commission. 66 To mention one example, Google argued that the AFAs were necessary to preserve the Android ecosystem and to prevent free riding. However, the Commission did not consider that Google had provided enough evidence showing that fragmentation could undermine the Android ecosystem. Furthermore, it also observed that Google had offered at some point an alternative agreement to AFAs called "Android Compatibility Commitment" (ACC) that gave OEMs some leeway to produce incompatible Android 62. Google Android supra note 6, Section 11. 63. Id. Section 12. 64. Id. Section 13. 65. E.g. Case C-209/10, Post Danmark A/S v Konkurrenceradet (Post Danmark I), EU:C:2012:172, § 41-42. 66. Supra note 6, Sections 11.5, 12.7, and 13.5. devices for third parties and supply components. This means, thus, that Google was able to offer less restrictive obligations than AFAs and therefore its justification for the conduct could not be sustained. 67 In sum, all these practices constituted a single infringement of Article 102 TFEU, 68 and the DOJ's complaint against Google shows that both agencies share similar concerns: Google's practices have an exclusionary effect that impedes competitors to find distribution channels for their search engine. However, a closer look at the approach of both agencies reveals some significant differences between the two jurisdictions.

B. Google Android: Different Perspectives 1. A More Comprehensive Market Analysis in the U.S. Complaint. While the European Union and United
States converge in considering that the principal harm occurs in the market for general search, the two agencies identify different markets on which to build their analysis. The Commission defines markets in the licensing of smart phones for operating systems, app stores, general search, and non-OS-specific web browsers. The United States focuses instead general search, search advertising, and general search text advertising. 69 As can be seen, there is only one market where the two overlap. This difference then leads the DOJ to design its case in a manner which is quite different from that of the Commission even if the conduct under consideration is similar.
Thus, while the European Commission places the focus on Google's tying practices and tests them against the standards set for tying by the case law, 70 the DOJ puts the emphasis on the two ways by which Google's practices foreclose access to competitors in general search: first, competing search engines are denied access at scale on devices running Android due to, for example, antiforking clauses imposed on OEMs, 71 revenue sharing agreements provided Google search is the default search engine, and the deal with Apple that made entry of a competing search engine on the Apple ecosystem impossible. Second, as a result, competing search engines are denied the capacity to raise revenue via advertising because they cannot grow to such an extent to provide an attractive prospect for advertisers. Noting that the exclusionary conduct is facilitated by the revenue generated in the advertising market because of a lack of competition, the DOJ is able to claim that the monopolization of the search and advertising markets are closely linked. This approach allows the DOJ to explore competitive harm on both sides of the market: the consumer side and the advertising side. This explains the close look the DOJ takes here about how advertising markets work-notably in the European Union, concerns about the impact of platforms on advertising were not central in this case. 72 The question then is why the DOJ adopts what seems a more holistic approach which pays closer attention to the dynamics of the market on both sides. One obvious reason is the result of the Microsoft judgment in the United States: according to that judgment, bringing a tying case in technology markets 67 A difference here is that the DOJ also considers ACC agreements imposed by Google harmful for competition although to a lesser extent than AFA agreements, whereas the European Commission puts ACC agreements aside and focuses on AFA agreements. Nonetheless, the assessment of ACC agreements does not seem to alter the findings of both agencies. 72. Some practices in the advertising market were later considered in Google AdSense (see supra note 6). requires a rule of reason analysis, and the standard for applying this in unclear. 73 As a result the DOJ frames the complaint in a manner that is as comprehensive as possible to ensure success. Similarly, the analysis of exclusivity agreements in the case law and it requires an appreciation of the wider economic context within which the conduct takes place. 74 This however just explains the litigation strategy. Another explanation for the more comprehensive examination of the market might be that the effects analysis differs in the two jurisdictions. The Commission still tends to frame the case so that it fits within the four squares of a particular antitrust category, while in the United States, there is perhaps greater emphasis on the impact of the conduct than on finding the right pigeon-hole for it. In the long term, the U.S. approach is preferable: categories create a veneer of legal certainty but allow parties to circumvent the law easily or create inconsistencies by having different standards for types of conduct that have similar effects. Conversely, placing more emphasis on the impact of conduct allows one to deploy a generally applicable standard to test monopolization irrespective of the specific form that conduct takes. This is particularly helpful in a case such as this where there are multiple types of conduct which vary depending on the counterparty that Google is dealing with.
Another example of the focus on effects is how the Commission and the DOJ approach the analysis of foreclosure. In Google Android, the Commission elaborates its arguments to show the exclusionary effects on the market caused by Google's practices, but the DOJ goes a step further attempting to measure the foreclosure effect by looking at the coverage of Google's agreement. For example, with respect to Apple, it claims that agreements foreclosed market access to an important distribution channel (36% of all general search queries in the United States are affected by the agreements) for a significant period of time. 75 This again is a lesson from the Microsoft judgment which emphasized the importance of measuring foreclosure. 76 2. Digital Markets of the Future. On one hand, both agencies consider to analyze how the conduct in question served to exclude rivals. For example, the Commission presents elaborate research on the fate of Amazon's OS Fire to demonstrate how Google's AFAs can exclude forked versions of Android. The DOJ also mentions this example and constructs a very similar argument in this regard. 77 On the other hand, the Commission makes some reference to the impact that the conduct may have on innovation, but the DOJ's appraisal is more precise on the likely future effects. It considers the likely impact that this conduct could have on potential rivals (e.g. Duck Duck Go which has a different business model that protects user privacy, as well as another new entrant with a subscription-based search engine), 78 and it also looks at the possible impact that this foreclosure can have on the availability of alternative search engines in new devices in the Internet of Things (e.g. smart TVs, digital personal assistants, and automobiles). 79  distribution channels for its search engine Google will try to keep its solid position in the market for search services as these evolve.
From an EU perspective, this difference is interesting because nothing would have impeded the Commission to conduct a similar appraisal. In the end, to apply Article 102 TFEU, the Commission only needs to demonstrate that the unilateral conduct may potentially restrict competition, which provides the enforcer with some leeway about what type of effects should be shown to run a case. The Commission, however, did not go as far as the DOJ in discussing the likely adverse effects of the conduct.
What accounts for this difference? Perhaps the focus of the DOJ on Google's conduct as a whole leads it to look at the expected expansion of Google's practices to nascent distribution channels, while the approach of the Commission to look at the specific form of the conduct places some limits to the scope of the effects that it can discuss. At the same time, the institutional structure of each antitrust system appears to work in favor of the Commission: given that it controls the steps of the investigation and the timing of the decision, nothing impedes the Commission to run another case in the coming years involving the Internet of Things. The DOJ, on the other hand, manages the initial investigation but then it must rely on the timing of a prosecutorial system, so it may make sense to raise a point about the future of digital markets in the present complaint. From this perspective, a Commission decision can afford to be more narrowly focused because it can then take further action against dominant firms in the future with relative more ease and speed than the United States. The latter may be more interested in future-proofing its cases.
More prosaically, the difference might be down to the fact that in 2020, the DOJ had a better understanding of how the market is developing beyond smartphones than at the time the Commission initiated the investigation of Google Android in 2013. See, for example, the evolution of smart voice assistants: Amazon only launched its intelligent personal assistant Alexa incorporated within Amazon's smart speakers called Eco in June 2015. Alexa's main product rival, Google's Home, was released in 2016 and was only available in 2017 in the United Kingdom. 80 At the same time, the difference may reveal a different approach to designing remedies: these have to be sustainable as markets evolve-a high fine might deter but injunctive relief to ensure there is no further monopolization as the market develops is likely a more valuable remedy. The fact that the DOJ is pointing at the future of digital markets might influence the way in which the remedies are designed and stimulate creativity in this regard.

Privacy and Competition Law.
Another aspect of divergence is the connection that the DOJ makes between the reduction of consumer choice and the degree of data protection and privacy for consumers: the fact that Google's services become the only choice for users lowers the degree of privacy and data protection since users are required to accept the terms imposed by their only choice of search engine. Conversely, had there been competition, users would have been able to choose a search engine on the basis which one better matched the value they place on privacy. And had there been the opportunity to compete, new entrants would likely have introduced different products, probably with greater privacy protection embedded in these services. 81 This consideration is linked to the findings of the report of the U.S. Committee on the Judiciary of its investigation of competition in digital markets: companies such as Google that offer low data protection can harvest more data that will allow them to cement their dominant position. 82 This finding helps provide a further demonstration of the anticompetitive effects on consumers that results from the conduct being challenged. Conversely, the Commission's approach to data protection in competition law cases is more conservative, normally assuming that the European Union's data protection Regulation would resolve these issues. 83 Linking data protection concerns to an antitrust cause of action helps provide a better understanding of the risks to consumer welfare. The difference in perspective between the agencies on this point may well be that the European Union has bespoke data protection rules while the European Union does not.
C. Google Adsense 1. The Market for Advertising and the Theory of Harm. The market for online advertising is fiendishly complex. It is however imperative to understand its workings to appreciate the antitrust claims. 84 Starting with the architecture of this market, we have publishers on one side (e.g. the Financial Times) who sell online advertising space to firms wishing to advertise (e.g. a hotel chain). The bringing together of multiple buyers and sellers of advertising space is carried out by three services: (1) publishers utilize ad server tools, which keep an inventory of all advertising space on sale; (2) advertisers use ad buying tools that are informed about the types of audience and publications that the advertisers wish to appear on; and (3) finally ad exchanges bring together ad servers and ad buying tools and serve as a platform where a huge amount of trades are taken place. Every time a user logs on to the Financial Times all three services are activated, the advertisers keenest to reach the user in question will place the higher bids and the one posting the highest bid will ensure that its advertisement is available to the reader who has just logged on. This means that the system is activated every single time we visit a website: the speed at which these transactions occur and volume of transactions are the stuff of science fiction.
One reason for the use of intermediaries between sellers and buyers is to minimize transaction costs: it is cheaper for a single entity to represent an advertiser and transact on their behalf, exchanges too reduce transaction costs. However, there is a second reason for this: data. The publisher records who it is that logs on, and this information is extremely valuable to the advertiser. Is the Financial Times reader a student looking on via a University subscription? Then, this user is less interesting to the peddler of boutique hotels than an executive logging on during the weekend at her country home. These data, which are collected from each of us as we surf the web, are the most valuable commodity in the business end of the internet economy and they shape which advertiser bids, when, and at what price.
With this background in mind, it is arguable that having a large number of intermediaries in each of the three domains described above (i.e. ad server tools for publishers, ad buying tools for advertisers, and ad exchanges where trades occur) would be the competitive outcome: publishers would choose the best performing ad server who, to retain the publisher's custom would look for the best performing ad exchange and might even multihome offering the same advertising space via multiple channels to secure the best price and placement for the publisher. Advertisers too would want their pick of ad buying tools. Given the speed and volume of transactions, customers would be quick to assess the better performers and the competitive process would secure allocative, productive, and dynamic efficiencies.
The EU and U.S. case against Google is premised on similar concerns: that Google dominates all three markets and uses its powers to squeeze rivals in these markets out. Both establish dominance by the same route: observing high market shares, entry barriers, and the firm's use of its power to keep these entry barriers high. A somewhat unique feature of the finding of market power is that both agencies use evidence of Google's exploitation of its power vis-à-vis advertisers and publishers as evidence of dominance. Insofar as EU case law is concerned, this is a rare finding because normally Article 102 TFEU is not used against firms with such established market dominance across several markets. 85 In U.S. antitrust direct evidence of monopoly, power is not normally required. 86 However, the two jurisdictions differ when it comes to analyzing the scope of the competition concerns. In the United States, the case begins by noting how Google first used its market power in the market for ad buying tools to require advertisers to use its ad exchange service, thereby excluding competition there. 87 It claims that this is against the interest of advertisers who would wish to use multiple exchanges to secure the cheapest prices for their advertisements. 88 The claim then moves on to show how, in addition Google then carried out similar exclusionary tactics in the other side of the market, by requiring publishers who used Google's ad services to utilize Google's ad exchange, deploying contractual and technical means to achieve this. 89 Overall, the harm to competition occurs in all three relevant markets. 90 The claim is a challenge to the entirety of the advertising chain, and it also includes a discussion about how Google sought to stifle an innovation that attempted to circumvent its monopoly power (known as header-bidding) as well as a discussion of possible collusion between Google and Facebook to stifle this innovation. The complaint also observes the special position of You Tube and the importance of this as an advertising channel, which is controlled by Google.
Conversely, the Commission's decision is focused more narrowly on the agreements between publishers and Google with regard to the search function embedded in websites. The decision finds that Google tried to foreclose market access to this segment of the website. The findings are threefold. First, that since 2006, publishers were forbidden from showing ads which were not provided by Google's AdSense for search. Second, from 2009, this was changed to a more "relaxed exclusivity" arrangement by demanding that advertisements sold by Google's AdSense were given premium placement. Finally, Google was entitled to approve any changes publishers wished to make to the display of ads sold by its rivals, which risked making these less attractive. The effect of this is that much like with the Android decision, the scope of inquiry is much more restricted, zooming in on a specific commercial relationship and not on the advertising market more generally. One possible explanation for this is that this abuse occurred relatively early on (since 2006) with the procedure commencing in 2010, suggesting that the entire panoply of conduct which forms the basis of the U.S. complaint may not yet have manifested itself. Another explanation is that these cases are resource-intensive and the Commission might prefer selecting a narrow factual issue and examine this sufficiently thoroughly to survive an appeal. The Commission might reasonably expect that if this decision is upheld that then it serves as a general precedent to apply also to other similar exclusionary conduct in the advertising market. Some might doubt whether the level of fines is sufficiently high to serve such general deterrence objectives. Antitrust LJ. 29 (2019). Given that the probability of getting caught and fined for unilateral conduct is lower than for cartel conduct the deterrent effect must be commensurably lower.

The Legal Standard.
Turning to the legal analysis, there are two issues of interest from a comparative perspective. The first is that there is some alignment in the discussion of the anticompetitive effects. The Commission decision identifies five effects: (1) deterring publishers from looking for alternative search ads providers; (2) foreclosing market access to competing intermediaries; (3) deterring innovation; (4) strengthening Google's position; and (5) harming consumers. 92 Two points of convergence may be noted in particular: the adverse effect on innovation and the harm to consumers. The U.S. complaint traces the harm down to consumers: "Because Google's ad server charges supra-competitive prices and depresses publishers' inventory yield, publishers offer consumers less content (lower output of content), lower quality content, less innovation in content delivery, more paywalls, and higher subscription fees." 93 This matches the Commission's understanding of consumer harm. 94 The second issue is of more salience to the EU case and this pertains to the legal standard to be adopted. The Commission insists that it may carry out what we have termed above a quick look assessment: provided it can show that the conduct in question amounted to an exclusive supply obligation and Google is unable to demonstrate that the conduct in question is either objectively justified or that "the exclusionary effect it produced was counterbalanced or outweighed by advantages in terms of efficiency gains that also benefit consumers," then the Commission has established an infringement of Article 102 TFEU. 95 Google counterclaimed that following the Intel judgment discussed above the Commission had to also show that the conduct in question was capable of restricting competition. To this, the Commission gave two answers: the first is that the Intel precedent should be read restrictively and is only applicable in cases where exclusivity is the result of a rebate scheme. The second is that even if the Commission is not required to show the capacity to foreclose competition, the Commission has actually demonstrated such an effect. 96 This response is somewhat problematic because it seems to focus more on rendering the decision appeal-proof than on explaining what the Commission understands to be the law. In other words, if the courts agree with the commission that an exclusivity agreement can be condemned merely by proof of exclusivity, then the Commission wins even without showing effects. Conversely, if the court believes that an effects-analysis is warranted, then the decision already provides for this. However, hedging in this manner means that the Commission may invest more resources than are necessary to bring a case.

Privacy and Competition Law.
As with Android, the U.S. complaint also makes reference to privacy issues as a parameter by which to assess competition harm, while the Commission remains relatively silent on this matter. The complaint makes two references to privacy that are worth considering. The first is the observation that Google secures information on users who go on publisher websites but does not share this information with the publishers. The discussion on this point indicates that if publishers could access these data, they would be able to share it with non-Google ad exchanges and non-Google ad buying tools. 97 It is also claimed that the inability to view these data does not allow the publisher to set the right price for its advertising space, risking a price decrease of 50 percent. 98 The complaint continues by noting that at the time of the Google Doubleclick merger, the antitrust agencies had understood that data would be shared. The plaintiffs also reject a privacy protection defense by observing that while the non-disclosure of data might well protect the individual from having non-Google firms view 92 the data, that very same data are shared in the Google ecosystem. 99 This discussion is interesting because it points to one of the entry barriers in this market: data. A new entrant in any of the ad markets will suffer a competitive disadvantage unless it has access to data that allows them to quickly establish a service which maximizes the utility of either publishers or advertisers.
Second, the complaint is also forward-looking in considering Google's ongoing project known as the privacy sandbox. 100 Google is planning to eliminate cookies and replace them with a different system of collecting data. Under the present system, advertisers and publishers can install cookies to track users. This allows both to understand the value of what the buy or sell and adjust their commercial strategies accordingly. Google plans to remove these cookies in favor of a tracking scheme that will give advertisers and publisher aggregated data instead. From a privacy perspective, one might see less people looking at personal data, but on the other hand, the complaint alleges that Google will still view the individual data, making its ad businesses more competitive than those of rivals. Thus, the user will continue to be tracked, while at the same time, the firm's monopoly power will become greater. 101 This points to a second feature of digital markets that risks creating a market failure: a conflict of interest when one firm is present in multiple, related, markets where its incentives are to favor is businesses rather than to act for the benefit of the customer. This points to a structural concern in digital markets.

IV. Remedies
It has been said that a wise plaintiff will start a case by thinking about remedies first. 102 However, the law of antitrust remedies and the role of remedies have remained under-studied. 103 It may be said that mainstream antitrust is not regulatory in character, so the expected remedy is a negative injunction. In the European Union, for example, most cases merely require the defendant to put an end the conduct in question. However, in digital markets, this general principle may prove exceptional: the Microsoft decisions discussed above as well as the remedies offered by Google in Android and Shopping provide for a detailed remedial design. In the United States, a case like that brought against the State of Texas on the advertising market seems to provide a good setting for the discussion of structural remedies, while a case like Android suggests that a good compliance pathway could be designed with an adequate mandatory injunction. From this perspective, both complains are lacking, listing a range of possible remedies sought by the plaintiffs without considering what would be most suitable.
A further point to reflect upon when considering remedies is that these should be prospective. Damages can serve to compensate for past harms suffered, administrative fines can serve to deter future conduct, but an important role for remedies which is recognized in both jurisdictions is to restore competition. 104 An even more complicated question arises when the dominant firm's conduct harms the 99. Id. at paragraphs 156 and 157. Furthermore the complaint also suggests that Google has weak privacy protections and is hungry for more data, see paragraphs 158 to 169. 100. Id. at paragraph 265-70. 101. Id. at paragraph 271. 102. Thomas O. Barnett, Section 2 Remedies: What to Do After Catching the Tiger by the Tail (June 4, 2008), https://www.justice.gov/atr/speech/section-2-remedies-what-do-after-catching-tiger-tail 'a bad section 2 remedy risks hurting consumers and competition and thus is worse than no remedy at all. That is why it is important to consider remedies at the outset. . . Doing so has a number of benefits. As an initial matter, having found a section 2 violation, you want to craft an effective remedy. It is not easy to craft what is typically a behavioral remedy that will achieve its desired objectives, avoid unintended harm, and be administrable. entry of potential competition-what remedy is appropriate to restore nascent competition? 105 Add to these the speed at which online businesses evolve and the slowness of antitrust intervention that it is unlikely that good remedies are imposed unless one is willing to go beyond the specific offense and prescribe a course of conduct which is designed to open markets to competition, an option which seems to be open in the United States. 106 Turning briefly to structural remedies, both jurisdictions are reluctant to impose divestitures, for similar reasons. In Microsoft, the Court of Appeals explained that one reason for this is the logistical difficulty of undoing an enterprise and the concomitant risk of creating a less efficient market structure. 107 Second, the Court took the view that one consideration that mattered in deciding in favor of structural remedies is the closeness of the causal link between conduct and the creation or maintenance of market power. 108 This was fatal in Microsoft because as we observed before, the plaintiff was only able to show that the conduct was capable of harming competition. The logic seems to be that if there are possibly other reasons why competition is harmed other than the defendant's antitrust violation, then it would be disproportionate to impose a structural remedy. In the European Union, the role of proportionality is more prominent: the Commission may demand a structural remedy only having established that a behavioral remedy would not solve the competition concerns. 109 A further consideration when it comes to structural remedies is the importance of selecting the right case to allow for divestiture. From this perspective, the question that follows is whether the U.S. enforcers in particular are picking the right cases to go for stronger remedies. In DOJ v Google, the issue is about exclusivity agreements so no structural remedy seems appropriate. The case against Google's advertising market may well be more suited to structural remedies. However, structural remedies come with the caveat that the benefits of users that currently enjoy from interoperability between the two companies may be gone. These interoperability benefits can however be recovered if the U.S. Bill on interoperability (discussed below) is finally adopted. Thus, it is arguable that the choice of a structural remedy might hinge upon there bring an adequate regulatory framework to ensure that the market works well after break-up.
Turning to behavioral remedies, the aftermath of the Commission's first Microsoft decision and the Google Shopping case reveal weaknesses in the design of remedies. As we discussed above in Microsoft, one remedy (selling a version of Windows without the media player) proved wholly ineffective. The more prescriptive and regulatory access remedy caused significant discussion and required further litigation to resolve. This remedy, which was analogous to the access regime designed in the United States, was seen as doing too little by some and going too far by others. 110 Google Shopping, it will be recalled, raised concerns that Google was favoring its vertical search engine at the expense of rivals. The compliance mechanism that the firm devised (absent formal guidance from the Commission) was to allow rivals to bid to be included in a prominent place on the website (the so-called Shopping Unit) but critics have indicated that this does not solve the Commission's competition concern which was that the algorithm used by Google in performing search undermined rivals. 111 In response to this criticism, the Commissioner for Competition indicated that the remedy appeared to be working: To date, the take-up of the compliance mechanism by competitors of Google Shopping, who could not have displayed their offers in the Shopping Unit before the decision of June 2017, has substantially increased. In June 2018, only one third of Shopping Units included at least one competitor, and around 6% of clicks in the Unit went to competitors. In November 2019, 81% of Shopping Units include at least one offer of a competitor, and 45.9% of clicks go to these competitors. 112 This, however, appears to undermine the theory of harm upon which the decision was predicated: true a person can pay a search engine for prominence but the concern in the case was also that for a non-paying search function Google was favoring itself. Furthermore, the difficulty for outsiders observing compliance is that there appears no formal and transparent monitoring mechanism.
In Google Android, it seems that the Commission reverted to the approach it adopted in the second Microsoft decision, indicating that Google should provide a choice screen so that a user does not have a preset search engine but is asked to select one actively. 113 Thus, Google announced in 2019 that it would implement a choice screen in new Android devices to give users the same opportunity to choose the default search engine. This choice screen will incorporate the winners of an auction. 114 However, early signs suggest that the design of the choice screen may not yet achieve the results the Commission hoped for, and Google's rivals have sent a joint letter to the Commission asking to reconsider the remedies. 115 The choice screen appears to be a good idea at first sight, but its effectiveness may depend on how the auction is designed so that it favors bids by search engines which are more likely to be installed by users. 116 In the summer of 2021, the remedy has been further fine-tuned. Among the changes is that participation in the choice screen is to be free for some providers. 117 As with the other decisions, the dominant undertaking's compliance pathway evolves as the Commission and rivals respond to the firm's efforts. This iterative approach to remedies raises a question about how to design an appropriate procedure to secure compliance. Presently in both Commission decisions discussed here, there is little transparency in the manner in which remedies are designed. Furthermore, a larger question is the extent to which these remedies, even if they succeed, yield an improved market for consumers. It is in part as a result of dissatisfaction with antitrust that both legal systems may be turning to regulation.

V. The Regulatory Turn
In Dec. 2020, the European Commission unveiled its draft for a DMA. This is a proposal for a regulation that would provide for ex ante regulation of a number of businesses that are said to be gatekeepers, targeting many of the online services offered by Google, Facebook, Amazon, Apple, and Microsoft. Interestingly enough, in the summer of 2021, several Bills were proposed in the United States to address the conduct of these firms as well. 118 These proposals are set to undergo major revisions and some may not be enacted. Consequently, in this section, we focus our comparative exercise on some common themes which emerge from these legislative proposals rather than comparing these line by line.

A. Motivation and Objectives
It is relatively easy to understand the desire in the United States to press for legislative change. The principal reason is that U.S. antitrust agencies have generally appeared unwilling until very recently to challenge conduct in digital markets. Suffice it to recall the 2013 decision of the FTC to close the investigation concerning Google's self-preferencing practices after an extensive review of the matter, arguing that Google's display of its own content could be seen as an improvement for users. 119 And, even if they had brought more cases to courts, they would have faced a defendant-friendly court. 120 Furthermore, as observed earlier, the application of the rule of reason requires showing anticompetitive effects and a discussion of countervailing efficiencies, which are resource-intensive. These concerns, which have been present for some time in academic scholarship, have now reached the ears of U.S. policymakers and legislators. 121 However, in the European Union, given the greater activism of the Commission and some national competition authorities in handling digital markets cases, understanding why the Commission proposed a DMA is less easy. As we saw above, legal doctrine is also more conducive to enforcement than the U.S. case law. Furthermore, an expert report commissioned specifically to consider the role of competition law in digital markets recommended that the Commission make incremental changes to competition law rather than develop specific legislation. 122 The main motivation in the European Union appears to be that regulation can be quicker and more comprehensive than completion law intervention: as we observed, in Microsoft, Android, and Adsense, the Commission could only pick on some elements of the conduct at play.
A further question to ponder is what the U.S. and E.U. legislators seek to achieve with these legislative efforts. Is the purpose to provide an accelerated enforcement of antitrust law, or is it something more? As we will see in more detail below, it is not unreasonable to conclude that the legislation is just an abridged version of antitrust: both sets of proposals contain an abridged market power test and a long list of prohibitions and obligations imposed on platforms that are found to have significant market power. As some commentators have noted, the list of obligations corresponds to some extent to anticompetitive types of conduct that competition agencies and courts have tackled in the last decade. 123 This is particularly interesting for the United States as it suggests a shift from a rule of reason to a per se illegality standard, since enforcement is mostly conceived in the Bills to occur ex post. It is sense, thus, it seems that the legislative proposals in both jurisdictions constitute an accelerated enforcement of antitrust law.
However, the answer to this question changes if we zoom in on the DMA. The Commission proposed two objectives: fairness and contestability. Fairness relates to the relationship between businesses and dominant platforms and seeks to implement a principle of a balanced exchange between the two sides. For example, the DMA will allow businesses to multihome and offer their services via more than one platform and, at the same time, the platform cannot give preference to its services at the expense of those of rival businesses operating through that same platform. 124 Securing fairness is sometimes achieved using Article 102 TFEU (which, unlike Section 2 of the Sherman Act, also applies to exploitative abuses by which the monopolist takes advantage of its market power to secure rents from customers) but the use of this provision for such purposes is limited. 125 Of more importance for the future of digital markets is the DMA's reference to contestability. This is patently not a reference to the notion of contestable markets which emerged in the literature of the 1980s. 126 While the Commission's legislative proposal does not provide an explicit definition of contestability, it can be interpreted as seeking to achieve three outcomes. 127 One is to allow the emergence of competitors to challenge the position of the dominant platform. The second outcome is to achieve platform disintermediation, so that a user may opt to use some of the platform's services but may also in parallel utilize some of the services of another. For example, consumers might use a dominant operating system but then use an app store operated by another firm. Or businesses might sell via one platform but seek to receive payment using a service provided by another firm. The third outcome is to prevent leveraging, that is to say, forbidding the platform from extending its market power into adjacent markets. Reading these three ambitions as a package, together with the goal of fairness, it appears that the DMA goes beyond antitrust law by providing an appropriate response to the main policy concerns we find in digital markets: that of a small number of dominant platforms that expand their market power across a range of different services and stifling the emergence of competitors. However, it is not entirely clear whether similar policy considerations animate the U.S. legislator. As we discuss in Section 5.3, some of the Bills seek to achieve contestability in a fairly radical manner.
A final, brief observation pertains to the role that antitrust rules play in markets that are covered by regulation. In the European Union, nothing prevents the application of Article 102 TFEU or national competition law to forms of conduct that will be already regulated under the DMA. 128 This is because EU primary law (the Treaty) may not be displaced by secondary law (a Regulation). In the past, the primacy of EU law has served to correct weak regulatory efforts. 129 Thus, the DMA provides a floor: those platforms designated as gatekeepers have to comply with its minimum prohibitions and obligations, but more may be imposed case-by-case by national authorities enforcing competition law and by the legislators of Member States if the new rules pursue other legitimate public interest. 130 The issue that might arise is that, given the relatively narrow scope of some of the obligations and prohibitions listed in the DMA, it is likely that further intervention using competition law powers may be needed to fill in the gaps of the DMA. It is true that the DMA establishes a channel for the Commission to update the list of obligations and prohibitions if it learns that gatekeepers are engaging in new practices that limit the contestability of digital markets or that are unfair. However, while this aims to make the DMA future-proof, it is worth noting the procedure for such update can be lengthy. 131 Conversely, the U.S.
Bills are silent on their relationship with the Sherman Act. Questions will thus arise about whether the U.S. Acts are "an effective steward of the antitrust function" 132 such that antitrust need not apply. This suggests that the United States is potentially taking a greater risk than the European Union in choosing to regulate because, if regulation does not deliver the results hoped for and the courts deem the Sherman Act inapplicable, conduct may escape control altogether.

B. Scope of the Regulations
Both legislators seek to regulate a number of online services but with different terminology and somewhat different policy choices. The DMA identifies a set of core platform services where regulation is called for if any one of these services is provided by a platform designated as gatekeeper. In the United States instead, the regulatory framework applies to covered online platforms under the relevant Bills.
On one hand, in the European Union, the following are identified as core platform services: (1) online intermediation services, which includes marketplaces and app stores; (2) online search engines; (3) online social networking services; (4) video-sharing platform services; (5) number-independent interpersonal communication services, such as WhatsApp; (6) operating systems; and (7) cloud computing services. 133 Furthermore, advertising services are included if they are provided by any one of the other core platform services. In the United States, on the other hand, an online platform is a website, online or mobile application, operating system, digital assistant, or online service that (A) enables a user to generate content that can be viewed by other users on the platform or to interact with other content on the platform; (B) facilitates the offering, sale, purchase, payment, or shipping of goods or services, including software applications, between and among consumers or businesses not controlled by the platform; or (C) enables user searches or queries that access or display a large volume of information. 134 Note, thus, that there is considerable overlap between the two approaches. The U.S. proposal is forward-looking by including emerging forms of digital assistants and, at the same time, the European Union includes cloud computing which may be outside the scope of the U.S. Bills-although one might argue that cloud computing is not a market which seems to raise many concerns that require ex ante regulation. Moreover, the inclusion of the advertising side of these markets in the DMA proposal is instead valuable because, as we have seen in our discussion of Google Adsense, this is the market where two of the platforms with which policymakers are concerned make their money. Browsers and not included in the DMA proposal and it is not clear if they are included in the U.S. proposals, but they should be included, given the competitive significance of this service.
Another similarity concerns the personal scope of the proposals, as not every provider is regulated. In the European Union, the focus is on providers of core platform services designated as gatekeepers and, in the United States, on covered platforms. The standards used to determine these are analogous. In the United States, a covered platform is one which (1) has at least 50,000,000 US-based monthly active users on the online platform, or at least 100,000 US-based monthly active business users on the platform; (2) is owned or controlled by a person with net annual sales, or a market capitalization greater than US$600,000,000,000; and (3) is a critical trading partner for the sale or provision of any product or service offered on or directly related to the online platform. 135 In the European Union, the standard is qualitative at first. The Commission is required to show that the core platform service (1) has a significant impact on the internal market; (2) operates a core platform service which serves as an important gateway for business users to reach end users; and (3) enjoys an entrenched and durable position in its operations or it is foreseeable that it will enjoy such a position in the near future. However, the first two requirements are presumed on the basis of quantitative indicators. 136 Significant impact is presumed where the firm to which the service belongs has an annual EEA turnover equal to or above EUR 6.5 billion in the last three financial years, or where the average market capitalization or the equivalent fair market value of the undertaking to which it belongs amounted to at least EUR 65 billion in the last financial year, and it provides a core platform service in at least three Member States. The gateway role is presumed where it provides a core platform service that has more than 45 million monthly active end users established or located in the Union and more than 10 000 yearly active business users established in the Union in the last financial year. 137 Having said that, the key point for comparison is that the U.S. designation is much more automatic than the European Union's where the parties might contest the presumption. 138 Conversely, the Commission is also able to apply the DMA to firms below these numerical thresholds. This means that the Commission can designate a platform as gatekeeper by looking at some elements, including the entry barriers derived from network effects and data-driven advantages, users lock-in effects, or other structural characteristics of the market. 139 Politically, this approach might be surprising since the platforms that will be targeted by the DMA are well-known. However, there are other reasons that may justify this choice. From a legal perspective, the EU legislator is constrained by the principles of proportionality and non-discrimination. This implies that the proposal must be designed in more functional terms to include within the scope of the DMA those entities that are more relevant for the functioning of market. From the perspective of future-proofing of the regulation, including a flexibility clause for the purposes of designating gatekeepers gives more room for the Commission to adjust the application and scope of the DMA. In this sense, the quasi-automatic designation of covered platforms in the U.S. proposals may make the regulation in the United States more rigid than in the European Union.
Another important comparator is that the DMA proposal in the European Union seems to require a two-sided market: the gateway must have both a given number of business users and consumers. The U.S. proposal only requires to meet the threshold regarding only one side of the market when designating covered platforms. In theory, it seems that the different choices in the European Union and the United States could lead to a scenario where a platform is covered by the regulation in one jurisdiction but not in the other. In practice, however, this may well prove to be less significant than it seems since the target firms are the same in both jurisdictions.

C. Regulatory Approaches
The proposed legislation in the two jurisdictions contains three regulatory approaches: command-andcontrol, regulatory dialogue, and structural separation. The first approach is designed in a highly imperfect manner; the second one is present but plays a small role; and the third approach is potentially the most radical and thus likely to face the greatest political resistance. 136. DMA draft supra n 7, Article 3(1) 137. Id. Article 3(2). 138. Id. Article 3(4). 139. Id. Article 3(6).
The command-and-control approach is at the heart of the DMA and the American Choice and Innovation Online Act. In both, we find lists of conduct that is forbidden or conduct which is required of the regulated firm. It is beyond the scope of this paper to review the two lists, although both proposals address similar types of conduct. However, there are two notable differences that reveal divergent regulatory philosophies.
The first is that the obligations in the DMA are much more precise and often address specific conduct by a specific platform service. Consider, for example, self-preferencing. Under the DMA, the prohibition specifically forbids the gatekeeper from treating more favorably in ranking the services or products of rivals. 140 This seems to largely focus on the factual matrix in Google Shopping. In contrast, the American Choice and Innovation Online Act provides that it is unlawful to advantage the covered platform's own products, services, or lines of business over those of another user. 141 This nondiscrimination obligation is much more flexible in application. The difference between the two can be readily explained. In the European Union, the legislator was keen to provide a list of obligations which were very easy to comply with ex ante. This helps accelerate the effect of the regulation as firms know what it takes to comply and to address the concern that antitrust actions arrive too late. One might even argue that, short of naming names of the firms and services that must comply, the draftsman sought to provide a clear list of forbidden conduct. The U.S. legislator instead does not appear to be in a hurry: as we observed above, the main concern is that in antitrust, the plaintiff tends to lose so providing for a rule that may require further litigation to be specified precisely is less of a drawback.
The second difference is that the prohibitions in the European Union must always be complied with, while a defense is available in the United States. The defendant in the United States, thus, may justify the clause either using a rule of reason type approach and showing that the conduct "would not result in harm to the competitive process by restricting or impeding legitimate activity by business users" or by invoking reasoning akin to an ancillary restraint: the conduct "was narrowly tailored, could not be achieved through a less discriminatory means, was nonpretextual, and was necessary to prevent a violation of, or comply with, Federal or State law; or protect user privacy or other non-public data." 142 Note the difference with the DMA that only allows to freeze its application temporarily if the gatekeeper demonstrates that complying with the obligations would endanger its economic viability or if there are grounds on public morality, public health, or public security. 143 The divergence, again, stems from the greater focus on speedy intervention. Furthermore, the European Union can justify the absence of a defense by reference to the narrow scope of the prohibitions, while the U.S. justification for a defense is precisely because the obligations are much broader.
There is however a second regulatory style which we find traces of. In the DMA, the Commission distinguished the prohibitions in two lists: the Article 5 list contains prohibitions that should be selfexecuting (e.g. a gatekeeper must allow a business user to sell the same goods or services via alternative platforms at different prices), 144 while the Article 6 list contains those which may require further specification on part of the Commission (e.g. allowing the competing business user the capacity to interoperate with the hardware or software used by the gatekeeper). 145 The process of specification, described as regulatory dialogue, is discussed in Article 7 of the DMA proposal. 146 It is true, however, that the Article itself gives more prominence to the Commission in the dialogue than to the gatekeeper; although some might advocate of a procedure where the Commission and the gatekeeper are able to explore in greater detail how to best comply with the obligation while not delaying the application of the obligations in Article 6. Moreover, unlike what now happens with competition law behavioral remedies, the process could be rendered transparent and opened up for comment by third parties. 147 This approach is also found in the Augmenting Compatibility and Competition by Enabling Service Switching Act of 2021 (the ACCESS Act). This is designed, on one hand, to allow a consumer to take the data which is controlled by the covered platform to another provider and, on the other hand, it is designed to facilitate access to the covered platform by competing businesses by imposing an obligation to devise and maintain interfaces that allow interoperability with competitors and potential competitors. 148 Similar obligations are also found in the DMA. 149 However, the U.S. legislator realized that these onerous duties require a different regulatory stance. The Bill thus provides for the creation of a technical committee that assists the FTC in devising appropriate standards. The committee includes businesses using the covered platform as well as representatives of competition and privacy organizations and the representatives of the covered platform. The task of these committees is complex, for they are asked to balance across a range of interests: to stimulate competition but at the same time to secure data security and privacy while also preventing fraud and other malicious conduct and all this also having regard to the legitimate interests of the covered platform by limiting the right of businesses to make too many requests and the possible need for the covered platform to be remunerated. 150 Many of the obligations which policymakers are eager to impose on gatekeepers are of a similar nature. It may thus be worth considering expanding this regulatory approach across all obligations imposed on covered platforms and gatekeepers.
The structural approach is found most prominently in the Ending Platform Monopolies Act. This provides for a simple prohibition: a covered platform cannot own or control or have a beneficial interest in any other line of business which (1) utilizes the covered platform for the sale of product or services (e.g. Amazon cannot sell its goods via the Amazon website); (2) offers a product or service that the covered platform requires a business user to purchase as a condition of access to the covered platform (e.g. tying a search engine to a browser); or (3) gives rise to a conflict of interest. 151 The latter is defined as a situation where the platform has the incentive to favor its lines of business at the expense to those of rivals. 152 No efficiency defense is proposed, however. Thus, this Bill could at one stroke break up many of the present-day internet giants. If we consider past experiences with deconcentration, it is unlikely that this ambitious project will see the light of day. 153 In contrast, the DMA opts for structural remedies only as a last resort: if the gatekeeper has systematically failed to comply with its obligations and thereby strengthened its market power, then structural or behavioral remedies may be imposed. 154 However, as with general EU antitrust law, structural remedies are only available if behavioral remedies will not resolve the concerns about fairness and contestability. 155 Moreover, as most of the gatekeepers are likely firms based outside the European Union, it is unlikely that the Commission will press for structural remedies unless these are coordinated with the firms' home jurisdiction. 147. Giorgio Monti, The Digital Markets Act: Improving Its Institutional Design, 5(2) Eur Compet Regul Law Rev. 90 (2021) developing this further. 148. ACCESS Act supra note 118, sections 3 and 4. 149. DMA proposal supra note 7 Articles 6(1)(h) and (f) respectively. 150. ACCESS Act supra note 118, section 7(c). 151. Id. section 2(a). 152. Id. Section 2(b). 153. The closest example we can think of are proposals by the FTC in the 1970s to break up certain industries that were too concentrated. This is discussed in Giorgio Monti

VI. Conclusion
Antitrust scholars have hardly ever enjoyed a time where so much enforcement and regulatory activity is focused on one segment of the economy. Given the ongoing process of litigation and regulation, this article can only offer a snapshot of present developments. However, the contribution we seek to make here is to use ongoing proceedings to take a step back and reflect, using a comparative method, on what kinds of policy choices are made by EU and U.S. competition authorities, plaintiffs, and courts.
From the perspective of EU competition law, we observe that on one hand, the Commission is placing more emphasis than it might have done in the past in constructing a theory of harm and in establishing harm to the competitive process and to consumers. In particular, in the Android decision efforts are devoted to showing consumer harm even when this is strictly speaking not required based on the case law. This points to some convergence with the U.S. standard. On the other hand, the Commission also insists that there are certain forms of unilateral conduct where it is able to condemn without a full effects-based analysis. The space for such a quick look approach will be determined by ongoing appeals against its Google decisions. In the recent Google Shopping judgment, the General Court appears to have accepted that a quick look approach is available when the conduct of the platform is prima facie suspect. 156 This aspect of the EU approach reflects some divergence with the U.S. standards. Conversely, considering the plaintiffs' arguments in the two Google cases in the United States, we observe the long shadow cast by the Microsoft judgment, the plaintiffs certain that a rule of reason approach must be used to frame the analysis and anticipating the efficiency claims of the defendant.
A consideration of several recent cases in this article reveals that another point of divergence is that often the U.S. agencies frame the competition problem in a more comprehensive manner: we observed this in Microsoft and Intel where the Commission focused on discrete abuses, while the DOJ and the FTC, respectively, focused on describing and addressing a general exclusionary strategy. The same applies to the claims about Google's conduct in the advertising market: the Commission zooming in on a specific practice, the U.S. plaintiffs examining a wider range of conduct across the whole ad tech value chain. It is not clear whether this is motivated simply by strategic litigation choices (e.g. the Commission might find it easier to win if it approaches a smaller set of issues) or whether this is the result of limited resources requiring greater focus, or if it is down to the more formalistic approach for which the Commission is often criticized.
Comparing remedies proved somewhat harder, given the modest findings of the U.S. agencies so far. However, experience to date suggests that specific remedies which are supervised by third parties (e.g. in Microsoft) can be laborious to implement but may prove to be more effective than a simple ceaseand-desist order which allows the firm to decide how to comply with relative independence. As we observed in both of the Commission Google decisions (Shopping and Android), a nontransparent process of review and discussion with the Commission results in gradual adjustment of the firm's compliance efforts but this could be ameliorated by a more structured discussion among competition agencies, firms, and interested stakeholders. This approach is found in commitment decisions and nothing prevents the same to be replicated in infringement decisions.
More generally, policymakers on both sides of the Atlantic are convinced that regulation is required to control digital platform markets and that antitrust will not suffice: the US basing this on the consideration that precedent and judicial attitudes favor a hands-off approach, the Commission concerned more about the duration of proceedings. One concern with respect to both legislative initiatives is about the end-point of regulation: what are the legislators seeking to achieve? In liberalizing telecommunications markets in the past, the regulatory vision was more clear: to open several segments of the market to competition and facilitate the entry of new service providers, while observing that there may well be some nonreplicable assets to which rivals must have access. 157  the legislators' ambition is. As others have commented before, not all the GAFAMs have the same business model, and thus, regulatory intervention that is overly sweeping may not be optimal. 158 On the other hand, the DMA and U.S. Bills cannot just promise a more speedy route to achieve the goals of antitrust law. The legislator should spell out a vision of what the regulated markets should look like and explore how far the obligations it is imposing will lead to that outcome. In this respect, there is one key issue which has been the leitmotif of technology development for centuries: the nascent competitor and the incumbent's desire to quash it. 159 Someone, somewhere, will see a gap in the market or an inefficient incumbent and will seek to enter. In this article, we have seen from the first Microsoft case of the 1990s that the dominant player's greatest fear is disruptive innovation: some new technology that sweeps away the basis of the incumbent's monopoly. To a degree, the regulatory initiatives discussed in this article appear to be directed at this by safeguarding a process by which new entrants can participate in the market, and where consumers can easily port their data away from the incumbent to them. If this is a plausible policy direction, then there is a major gap in the European Union's current proposals, and this is the absence of merger rules to regulate the acquisition of nascent competitors. 160 Finally, in addition to specifying objectives, both legislators should give more thought to how to regulate. Using the comparison with telecommunications regulation again, it might be that a more participatory form of regulation (which is hinted at episodically in the DMA with reference to a regulatory dialogue and in the ACCESS Act with respect to data portability and interoperability) merits greater prominence than the deterrence-based approach which characterizes the bulk of the current regulatory efforts.