ICLE Issue Brief

Clouded Judgment: AWS, Azure, and the DMA’s Gatekeeper Test

Executive Summary

On June 25, 2026, the European Commission reached the preliminary view that Amazon Web Services (AWS) and Microsoft Azure should be designated as gatekeepers under the Digital Markets Act (DMA). The decisions would be the first to rely entirely on the DMA’s qualitative criteria because cloud computing satisfies neither of the law’s user-number thresholds. The Commission must prove that each undertaking is an “important gateway” with an “entrenched and durable position.”

This issue brief argues that the Commission’s preliminary position conflates gatekeeping with size. Cloud services are eligible for designation, but Article 3(8) requires the Commission to establish each statutory condition independently and without a presumption.

Evidence of scale, investment, or economic importance may show that AWS and Azure matter. They will not show that they intermediate between business users and end users, which is the core of the gateway inquiry. Nor does the current record establish entrenchment. Quality-adjusted prices have fallen, market shares have shifted, multihoming is widespread, switching costs have declined, and artificial intelligence (AI) demand has encouraged entry.

The Commission also relies on the firms’ combined position even though each service must qualify individually, and it has not explained why Google Cloud falls outside the inquiry. Any prediction that AI, data gravity, or sovereign-cloud requirements will entrench AWS and Azure must satisfy the General Court’s demand in Meta Platforms for specific, contemporaneous reasoning.

Proportionality presents a final problem. The United Kingdom examined the same firms on a fuller record and chose voluntary commitments. The European Union has also enacted a purpose-built switching regime in the Data Act, while the Commission is still studying whether the DMA’s obligations fit cloud services.

The proposed designations will show whether “gatekeeper” remains a distinct legal category or becomes a synonym for “big.”

I. Introduction: Cloud Gatekeepers Without the Numbers

On June 25, 2026, the European Commission reached the preliminary view that Amazon Web Services (AWS) and Microsoft Azure should be designated as gatekeepers under the Digital Markets Act (DMA).[1] The announcement followed market investigations opened in November 2025[2] and started a demanding timetable. The companies are expected to submit written observations in the autumn, the Commission must issue final decisions by November 2026, and, if it confirms the designations, both providers will have six months to bring their cloud services into compliance with the DMA.[3]

The proposed designations rely on an expansive interpretation of the DMA’s qualitative criteria. If upheld, that interpretation would broaden the regulation’s material scope substantially, as some early commentary anticipated and encouraged. Section II.C surveys that literature.

Article 3(8) allows the Commission, after a market investigation, to designate a company that does not satisfy the DMA’s quantitative thresholds. The Commission has used that authority once before. In April 2024, it designated Apple’s iPadOS as a core platform service (CPS), even though the operating system fell below the end-user threshold.[4]

The iPadOS designation remained closely tied to the statutory presumptions. Apple exceeded the business-user threshold by roughly elevenfold, its end-user numbers approached the statutory threshold and were expected to rise, and iPadOS formed part of a consumer-facing system already covered by the DMA.[5]

The proposed cloud designations differ in kind. AWS and Azure satisfy neither user-number threshold because cloud computing is overwhelmingly a business-to-business service. Its ultimate users are largely indirect and do not resemble the end users contemplated by the statutory presumption.[6]

For the first time, the DMA’s qualitative criteria must support a designation without assistance from the numerical thresholds. The Commission must establish that each service has a “significant impact on the internal market,” serves as an “important gateway,” and holds an “entrenched and durable position.” And it must do so both without a numerical presumption and without a consumer-facing side of the market that could support one.[7]

Until now, the Commission has made most designations through the quantitative presumptions in Article 3(2). Even when it relied on qualitative criteria, as with iPadOS, those criteria remained closely connected to the numerical thresholds. The AWS and Azure investigations therefore test whether gatekeeper status identifies firms that control genuine competitive bottlenecks or merely large firms operating in contested markets.

On the public record, the Commission cannot sustain its preliminary position without conflating gatekeeping with size. This issue brief advances five claims.

First, Article 3(1)(b) contains two distinct requirements. A service must be a gateway, and that gateway must be important. The two-sided structure of the Article 3(2)(b) presumption ordinarily supports that inference by measuring both business users and end users. That proxy fails for business-to-business infrastructure services because both sides of the statutory test reflect the same mismatch between the service and the threshold. Evidence of scale or economic importance cannot establish the separate proposition that a service controls access between two groups of users.

Second, Article 3(1)(c)’s requirement that a firm hold an “entrenched and durable position” amounts to a test of market contestability. A qualitative designation under Article 3(8) must therefore rest on evidence showing that rivals cannot discipline the firm or displace its position.

Third, the cloud investigations rely on a duopoly framing that is difficult to reconcile with the DMA’s requirement that each service qualify individually. The Commission has also offered no adequate explanation for excluding Google Cloud. Because Article 3(8) gives the Commission discretion, these distinctions expose the decisions to an equal-treatment challenge that designation through a statutory presumption would largely avoid.

Fourth, any forward-looking finding of entrenchment based on AI or “data gravity,” meaning the tendency of large datasets to attract related applications and services, must satisfy the reasoning standard the General Court applied when it annulled the Facebook Marketplace designation. The Commission must connect its prediction to evidence in the decision. A bare assertion that AI will entrench current market positions cannot meet that standard.

Fifth, the Commission’s approach differs sharply from that of the United Kingdom’s Competition and Markets Authority (CMA). The CMA examined the same two firms on a fuller evidentiary record and concluded that voluntary commitments offered a more proportionate response than designation under the United Kingdom’s comparable ex ante regime.

That comparison reinforces the proportionality concern. The European Union has already adopted a purpose-built cloud-switching regime through the Data Act. The Commission is also conducting a parallel inquiry into whether the DMA’s existing obligations fit cloud services at all. Those proceedings suggest that the Commission is pursuing designation before establishing that the resulting obligations address the conduct identified in the investigation.

II. Gatekeeper Designation and the Cloud Problem

This section develops the legal and economic framework for assessing the proposed cloud designations. Section II.A examines Article 3’s three cumulative conditions, the quantitative presumptions that ordinarily establish them, and the Commission’s heavier evidentiary burden when it proceeds under Article 3(8). Section II.B explains the economic theory behind gatekeeper status and the distinction between firms that are merely large and services that control genuine gateways. It also shows how the Commission’s designation practice has begun to enforce that distinction. Section II.C turns to cloud computing. Although the DMA expressly makes cloud services eligible for designation, their business-to-business structure prevents them from satisfying the user thresholds and fits uneasily with the regulation’s model of two-sided platform intermediation. That mismatch does not foreclose designation, but it requires the Commission to prove through evidence that AWS and Azure function as competitive bottlenecks rather than rely on their scale or economic importance.

A. The DMA’s Designation Framework

Article 3(1) of the DMA makes gatekeeper designation contingent on three cumulative conditions. The undertaking must have “a significant impact on the internal market,” provide a CPS that serves as “an important gateway for business users to reach end users,” and hold “an entrenched and durable position” or be likely to attain one in the near future.[8]

Article 3(2) establishes rebuttable presumptions for each condition. Financial thresholds and operations in at least three member states create a presumption of significant market impact. At least 45 million monthly active end users and 10,000 yearly active business users in the European Union create a presumption that the service is an important gateway. Meeting those user thresholds in each of the previous three financial years creates a presumption of an entrenched and durable position.[9]

Two corrective mechanisms complete the framework. Article 3(5) allows a firm that meets the thresholds to rebut the presumptions with arguments that “manifestly call into question” its designation. The General Court has interpreted that standard strictly, requiring a “high degree of plausibility.”[10] Article 3(8), by contrast, allows the European Commission to designate a firm that does not meet the thresholds after conducting a market investigation under Article 17.[11]

As Friso Bostoen and Giorgio Monti explain, Articles 3(5) and 3(8) form an error-correction pair. Article 3(5) addresses false positives produced by the quantitative presumptions, while Article 3(8) addresses false negatives.[12] Under Article 3(2), designation is mandatory once the thresholds are met unless the firm rebuts the presumptions. Under Article 3(8), no presumption applies. The Commission must affirmatively establish all three statutory conditions based on the evidence.[13]

Because the conditions are cumulative, the Commission must establish each one separately. Evidence relevant to one condition cannot simply be reused to establish another. For example, evidence of turnover, market capitalization, or investment may establish that a company has a significant impact on the internal market. But those measures do not, without more, show that its service acts as an important gateway or that its position is entrenched and durable.[14]

B. The Economic Logic of Gatekeeper Status

The DMA’s designation criteria reflect a specific economic diagnosis. The expert reports that shaped the regulation offered a common account of why some digital markets resist ordinary competitive pressure. Extreme economies of scale, strong network effects, and data-driven feedback loops can cause a platform market to tip toward a single intermediary. Once that occurs, even more innovative rivals may struggle to challenge the incumbent.[15]

The DMA’s gateway concept translates that diagnosis into law. In theory, the regulation targets a particular market structure rather than size alone. It focuses on platforms that mediate between distinct user groups and may become durable bottlenecks when tipping dynamics reinforce an early lead.[16]

Under this account, the Article 3(2)(b) presumptions are not arbitrary. The dual user thresholds serve as a proxy for two-sided intermediation. Their significance lies less in the absolute size of either group than in each group’s reliance on the platform to reach the other.[17]

Several commentators warned before the DMA’s enactment that these proxies could sweep more broadly than the economic diagnosis justified. Damien Geradin, a prominent supporter of the DMA, cautioned that the proposed criteria could capture firms that were large but did not control genuine gateways. He urged lawmakers to define the regulation’s target more clearly.[18] Pinar Akman likewise argued that the designation framework poorly approximated the competitive concern it was meant to address.[19]

Pablo Ibáñez Colomo focused on the institutional consequences of separating designation from market-power analysis.[20] Other scholars examined the resulting tension between the DMA and Article 102 of the Treaty on the Functioning of the European Union (TFEU).[21] Although the two regimes formally protect different legal interests, they inevitably confront overlapping economic questions.

Defenders responded that avoiding case-by-case effects analysis was central to the DMA’s design. In their view, Article 102 enforcement had proved too slow and uncertain for digital markets, and bright-line proxies offered the administrability that traditional antitrust lacked.[22]

Three years of practice have lent support to both positions. The Commission has designated seven gatekeepers and several dozen CPSs, almost all through the Article 3(2) presumptions.[23] Yet its decisions also show that Article 3(1)(b) imposes an independent requirement connected to an undertaking’s gateway status.

The Commission declined to designate X’s online social-networking service even though the company met the turnover thresholds because the service was not an important gateway.[24] It also accepted rebuttals for Gmail and Outlook.com despite user numbers above the statutory thresholds.[25] The same pattern appears in several other non-designations[26] and successful rebuttals.[27]

These decisions confirm that a service can be large without serving as an important gateway. They provide the baseline against which the proposed cloud designations must be assessed, as Section III.A explains.

C. Cloud Computing’s Uneasy Fit

Cloud computing has generated a substantial and fast-growing literature. Regulators have examined the sector through market studies and investigations by the Office of Communications (Ofcom) and the CMA in the United Kingdom; the Autorité de la concurrence in France; the Authority for Consumers and Markets (ACM) in the Netherlands; the Japan Fair Trade Commission (JFTC); and the U.S. Federal Trade Commission (FTC).[28] The Organisation for Economic Co-operation and Development (OECD) has also synthesized much of this work. Concurrences devoted a multipart On-Topic series to cloud competition in 2025, covering issues that ranged from interoperability mandates to AI partnerships.[29]

The economic literature has focused on customer lock-in, economies of scale, and the relationship between the Data Act and the DMA. A 2024 report by the Centre on Regulation in Europe (CERRE) examined each of these issues and urged the European Commission to apply proportionality before designating cloud services, given their differences from other CPSs.[30] Doctrinal scholarship has questioned whether the European Union’s digital rules fit cloud computing at all.[31]

The policy literature divides over the proper response. Some authors argue that the Data Act provides the purpose-built instrument for improving cloud contestability and should be given time to operate before the Commission invokes the DMA.[32] Others contend that qualitative designation is especially well suited to an industry organized around integrated groups of complementary services.[33] A separate public-utility strand would regulate cloud providers more extensively than the DMA does.[34] At the time of writing, no full-length analysis has examined the proposed AWS and Azure designations themselves.[35] This paper fills that gap by testing the proposed designations against the framework developed in the designation literature.

One possible argument is that cloud computing falls outside the DMA because an infrastructure service that does not interact directly with consumers cannot serve as an “important gateway for business users to reach end users.” Whatever force that argument may have as an economic description, the statutory text makes it difficult to sustain as a legal claim.

Article 2(2)(i) expressly lists cloud computing as a CPS.[36] Recital 14 explains that the listed services have “the capacity to affect a large number of end users and business users, which entails the risk of unfair business practices,” and should therefore fall within the regulation’s scope.[37] The recital refers to the capacity to “affect” end users, rather than requiring providers to have a direct relationship with them. That wording indicates that the legislature contemplated services that reach end users indirectly through the businesses built on top of them. Reading the recital otherwise would make the qualitative gateway criterion in Article 3(1)(b) largely redundant for cloud services.

Some scholars nevertheless argue that the DMA, like the Digital Services Act and perhaps the Data Act, was drafted around consumer-facing platforms and fits the technical and commercial features of cloud computing poorly. Others reach the opposite conclusion.[38] Reconstructing the Commission’s impact assessment, they argue that lawmakers included cloud because of its role within integrated groups of complementary services. On that account, qualitative designation is especially appropriate for a business-to-business industry whose competitive significance the user thresholds may understate.[39]

We take a middle position. Cloud’s inclusion in Article 2 settles its eligibility for designation, but eligibility does not establish that AWS or Azure satisfies the substantive criteria. The feature that justified including cloud—the capacity to affect end users without directly intermediating with them—also demands closer scrutiny. Cloud services do not operate like traditional multisided, consumer-facing platforms. The Commission must therefore show that the remaining evidence identifies a competitive bottleneck rather than merely a large and economically important firm.

A finding of a genuine bottleneck would accord with the DMA’s stated goal of promoting contestability. A finding based primarily on size would support the criticism that the regulation singles out large digital companies without establishing gatekeeper power.

The point at which the quantitative presumptions fail is instructive. Amazon and Microsoft satisfy the financial thresholds in Article 3(2)(a) at the undertaking level. That is undisputed. Their cloud services fall short under Article 3(2)(b), which measures active business users and end users. That user-number test is the regulation’s principal proxy for intermediation rather than size.[40]

The missing presumption is therefore the one directed most closely at the gateway requirement. Under Article 3(8), the Commission must replace that presumption with evidence, and it bears the burden of proving the statutory condition.

The first condition, significant impact on the internal market, is not seriously contested for AWS or Azure. The dispute concerns the remaining conditions. The Commission must prove that each service is an “important gateway” and holds an “entrenched and durable position.” The following sections address those requirements in turn.

III. Testing the Cloud Gatekeeper Case

With eligibility established and the burden of proof allocated, this section tests the European Commission’s preliminary position against the substantive conditions in Article 3(8). Section III.A argues that an “important gateway” must both intermediate between business users and end users and hold sufficient importance. Evidence that cloud services are large, critical inputs addresses only the latter requirement. Section III.B treats an “entrenched and durable position” as a question of market contestability. Falling quality-adjusted prices, shifting market shares, widespread multihoming, lower switching costs, and AI-driven entry weigh against entrenchment.

Section III.C examines the Commission’s reliance on the combined position of AWS and Microsoft Azure, even though the DMA requires each service to qualify individually. It also considers whether the unexplained exclusion of Google Cloud can survive equal treatment review under Article 3(8)’s discretionary framework. Section III.D addresses the possible claim that AI workloads, data gravity, and sovereign-cloud requirements will entrench AWS and Azure in the near future. Any such prediction must satisfy the General Court’s demand for specific, contemporaneous evidence rather than assumptions about technological change or European industrial policy goals.

A. Gateway Power Requires Intermediation

The DMA’s gateway concept rests on intermediation. A CPS sits between business users and the end users they seek to reach. As the General Court held in ByteDance:

[I]n order to consider that business users of a CPS [core platform service] “depend” on it in order to reach their end users, it is not necessary for that CPS to be the only channel through which those undertakings can reach those users. It is sufficient for it to be an important channel for that purpose, which those business users can access only if they have an account on that CPS.[41]

The economics of two-sided platforms support this conception. Intermediation power arises when a platform coordinates distinct user groups whose demands depend on one another. The platform’s position on one side then affects which users businesses can reach on the other side and at what price. In economic terms, the users on each side cannot reallocate the platform’s pricing decisions among themselves.[42]

Taken alone, the first and third conditions in Article 3(1) describe nothing distinctively digital or specific to gatekeeping. Many firms have a significant economic impact and hold long-established positions. The second condition distinguishes a gatekeeper from a company that is merely large and durable. Article 3(1)(b) targets services that control a bottleneck and determine how business users reach end users. Without that requirement, the DMA would impose special obligations on companies largely because they have remained big for three consecutive years.

This conception readily fits marketplaces, social networks, and app stores.[43] It fits raw computing and storage services poorly. A business that hosts an application on AWS does not reach its customers “through” AWS in any way those customers experience. Cloud computing is an input into the business’s product, not a channel to its audience in the sense that a marketplace or app store is.

The European Commission’s preliminary position blurs that distinction. Its announcement describes AWS and Microsoft Azure as “an important gateway between businesses and their customers.” It also observes that more than half of European Union businesses rely on cloud services, which it calls “a prerequisite for AI.”[44]

Reliance does not establish intermediation. Electricity, telecommunications transit, commercial real estate, and payroll software are inputs on which many businesses depend. None necessarily stands between a business and its customers in the sense contemplated by the gateway condition. Treating criticality as sufficient would deprive that condition of independent force. Every essential business-to-business input supplied on a large scale could qualify, turning the DMA into a statute governing important inputs and economic dependence throughout the economy rather than important gateways.[45]

Article 3(1)(b) contains two distinct requirements. The service must be a gateway, and the gateway must be important. The presumption that ordinarily establishes this condition, Article 3(2)(b)’s thresholds of 45 million monthly active end users and 10,000 yearly active business users, performs both functions.

The magnitude of those figures supports the importance inference. A service used on that scale clearly matters to the internal market. Requiring large numbers of both business users and end users supports the separate gateway inference because it identifies a service connecting two user groups.

For cloud services, both parts of the proxy fail for the same reason. People who use an application generally have no relationship with the infrastructure on which it runs. They therefore do not count as the cloud provider’s end users under Article 3(2)(b). Cloud infrastructure also lacks a second side in the conventional platform sense and does not exhibit the same cross-group network effects. The structural inference that many users on both sides depend on the service to reach one another cannot be drawn.

That mismatch required the Commission to proceed under Article 3(8), the proper legal route for services that do not meet the thresholds. But Article 3(8) requires the Commission to prove importance and gateway status separately, without a presumption.

Evidence of revenue, capacity, investment, or widespread use as an input may establish importance. It does not, however, establish that the service intermediates between business users and end users. A decision that relies only on measures of scale proves one half of Article 3(1)(b) while omitting the half that distinguishes a gatekeeper from a company that is merely large, significant, or important.

The Commission’s prior decisions apply this distinction. It declined to designate X even though the undertaking met the DMA’s turnover thresholds because the service was not an important gateway.[46] It accepted that Gmail and Outlook.com were not gateways despite their millions of business users and end users.[47] It also declined to designate iMessage after finding that the service was not important enough, even though it met the quantitative thresholds.[48]

The Commission’s only previous designation under Article 3(8), involving iPadOS, remained closely tied to the quantitative framework. Apple exceeded the business-user threshold by roughly elevenfold. Its end-user numbers approached the statutory threshold and were expected to rise. The designation also filled a gap within a consumer-facing group of complementary services whose related services were already designated.[49]

The iPadOS decision was therefore a gap-filling designation. Designating AWS and Azure would create a new category. It would mark the first designation in which neither side of the statutory intermediation proxy exists even approximately.

That prospect creates a dilemma. The Commission may be applying a more permissive gateway standard to infrastructure services than it applied to X, Gmail, Outlook.com, and iMessage, which would create the equal-treatment problem addressed in Section III.C. Alternatively, it may be treating “gateway” as another word for “big.”

A narrower interpretation can preserve the gateway condition while keeping cloud services eligible for designation. Under that interpretation, “gateway” describes the market structure that motivated the DMA. A platform intermediates between distinct user groups in a market where network effects, data-driven feedback loops, and extreme economies of scale can cause tipping and entrenchment.[50] That was the diagnosis developed in the Crémer report and Furman Review and reflected in the regulation’s design.

Cloud designation remains possible under this interpretation, but the Commission must show that the market’s structure makes tipping and entrenchment likely. That inquiry leads to Article 3(1)(c)’s requirement of an “entrenched and durable position,” where the evidence can be examined directly.

B. Entrenchment Requires Limited Contestability

Article 3(2)(c) presumes that an undertaking enjoys an entrenched and durable position where the user thresholds in Article 3(2)(b) were met in each of the previous three financial years. The presumption affects the undertaking, but its trigger is a particular service. Thus, Article 3(2)(b) is satisfied only where the undertaking provides a CPS that itself clears the user thresholds.

Entrenchment is therefore not a free-floating attribute that an undertaking carries into every service it ever launches. The General Court confirmed as much in Meta, holding that where some of an undertaking’s CPSs meet the thresholds and others do not, the Commission may open a market investigation under Article 17(1) in order to assess whether those services meet the requirements of Article 3(1), entrenchment included. Under the DMA’s own structure, entrenchment means that gateway status persists over time. When the user thresholds do not apply, the European Commission must fill that evidentiary gap with proof that the market itself is difficult to contest.

The case law points in the same direction. The DMA deliberately dispenses with Article 102 TFEU’s dominance test. The General Court has confirmed that designation requires neither a relevant-market definition nor a finding of market power.[51] Yet the two inquiries address related economic conditions.[52]

Dominance asks whether rivals can discipline a firm that raises prices, restricts output, or allows quality and innovation to decline. The General Court defines an “entrenched and durable” position in similar terms. It refers to circumstances in which “the contestability of that position is limited” and examines “the stability of that position over time.”[53]

In economics, a market is contestable when entry or expansion by rivals constrains an incumbent’s conduct, regardless of current concentration.[54] The General Court’s formulation therefore makes Article 3(1)(c) an economic inquiry. Without the quantitative presumptions, the Commission must establish actual impediments to contestability.

This point has received too little attention in the designation debate. As Chad Syverson explains, concentration “is an outcome, not an immutable core determinant of how competitive an industry or market is.” Indeed, on that score, “we cannot even generally know which way the barometer is oriented.”[55]

The same principle applies to the DMA’s qualitative test. Outside the statutory presumptions, an entrenched and durable position cannot be inferred solely from a snapshot of market structure. It results from limited contestability. The relevant question is whether competitive pressure can discipline incumbents over time.

Market shares alone therefore cannot establish entrenchment, particularly in an industry where rivals are actively gaining and losing business. The cloud industry displays vigorous competition along at least five dimensions.

First, market shares continue to move. AWS’ worldwide share of cloud-infrastructure services declined from roughly 32% in 2021 to about 28% in early 2026. Microsoft’s share rose to roughly 21%, and Google’s reached 14%. Oracle and a growing group of AI-focused “neoclouds,” which provide specialized computing capacity for AI workloads, have also gained ground. Five neocloud providers now rank among the top 30, and the segment accounts for about 5% of a market whose revenue increased 35% year over year to $129 billion in the first quarter of 2026.[56] Growth at that rate attracts entry, and the entry is occurring.

Second, quality-adjusted prices have fallen sharply, particularly as competition intensified. David Byrne, Carol Corrado, and Daniel Sichel’s hedonic price indexes, which account for changes in service quality, show rapid declines in the prices of AWS computing, database, and storage services between 2009 and 2016. Those declines accelerated to double-digit annual rates after 2014, when Microsoft and Google had reached sufficient scale to offer competitive prices.[57] Industry price histories show a similar pattern in storage, where list prices fell by more than 80% during the early 2010s.[58]

It is true that nominal prices for some services have changed little since the mid-2010s. That fact does not erase the earlier evidence of competitive discipline. AWS accelerated its price reductions when rivals attained scale. Quality improvements, expanding free-service tiers, and steep negotiated discounts also weaken any inference drawn from stable list prices.

Third, the economics of switching costs cut against designation. The standard literature predicts that firms in markets with switching costs compete aggressively ex ante for customers who may later become partly locked in. This “bargains-then-ripoffs” dynamic can dissipate ex post rents through ex ante discounts and concessions.[59]

Cloud credits and committed-spend discounts may reflect that competition for customers. Their prevalence suggests that providers must compete for business before customers become attached to a particular service. The CMA and Ofcom likewise documented intense competition for new customers and workloads.[60]

Fourth, the practices most often identified as lock-in mechanisms have weakened. In 2024, Google, AWS, and Microsoft eliminated or sharply reduced data-egress fees worldwide for customers leaving their platforms.[61] Data-egress fees are charges for transferring data out of a provider’s cloud.

The changes may have reflected regulation, including the EU’s Data Act,[62] although their worldwide scope makes that explanation less convincing. Competition may also have played a role. Either way, the friction most commonly cited as a cloud lock-in mechanism is receding. Providers that reduce the switching costs said to trap customers offer evidence that competitive pressure still constrains them.

Fifth, multihoming is widespread. Flexera’s 2026 survey of 753 cloud decision-makers found that 83% of respondents run some or significant workloads on AWS and 79% do so on Azure, with both providers used in some capacity by 88%. Among enterprises, AWS leads active workloads at 84% and Azure follows at 82%. Usage shares of that magnitude cannot obtain unless most organisations run both.[63] The CMA and Ofcom correctly caution that “multicloud” often means placing different workloads on different clouds rather than moving the same workload freely among providers.[64] Migrating an established workload can remain costly.

That caveat does not resolve the contestability inquiry. Competition often operates at the margin. New and expanding workloads, especially AI workloads, are awarded provider by provider. Competition for those workloads can constrain prices and terms for existing customers.

Using several providers also reduces the cost of returning to a provider and preserves a credible alternative during contract renewals. The relevant question is whether incumbents could worsen price or quality without losing marginal workloads. Moving market shares, falling quality-adjusted prices, and aggressive customer-acquisition discounts suggest that they could not.

The strongest counterargument comes from the CMA’s institutional findings. After a full market investigation, the CMA concluded that AWS and Microsoft each possessed significant unilateral market power in U.K. cloud-infrastructure services. It identified high barriers to entry and adverse effects on competition associated with egress fees, technical switching barriers, committed-spend discounts, and Microsoft’s software-licensing practices.[65] Ofcom’s referral study reached similar conclusions.[66]

Those findings deserve serious consideration, but they do not establish that AWS and Azure are entrenched and durable gateways under the DMA.

First, market power in a differentiated oligopoly does not necessarily establish an entrenched and durable gateway. The CMA’s record also documents price competition and active rivalry for new business. Its inquiry group therefore considered calibrated conduct remedies rather than structural intervention.

Second, the CMA’s eventual response is instructive. As Section IV.A explains, the authority that made the market-power finding ultimately concluded that voluntary conduct commitments were proportionate.

Third, the best-supported concern in the record involves Microsoft’s licensing terms. Those terms make it materially more expensive to run Windows Server and SQL Server on rival clouds than on Azure. This is firm-specific conduct involving an adjacent software market.[67]

Such conduct may warrant enforcement. Google withdrew its Article 102 complaint concerning those practices in November 2025, after the Commission opened the cloud investigations,[68] and the CMA has opened a separate investigation into Microsoft’s business-software offerings. But Microsoft’s licensing practices cannot support an infrastructure-wide designation of cloud computing as a CPS, and they provide no basis for designating AWS.

Article 3(8) does not allow the Commission to avoid this evidence. Once the quantitative thresholds fall away, entrenchment must be established for the services under investigation on their own facts, which means proving limited contestability directly. The available record describes a market in which prices fall, market shares move, new providers capture growing demand, and incumbents reduce switching costs to retain customers.

C. Individual Designation and Equal Treatment

The European Commission also faces an aggregation problem. The cloud statistic cited most often is the combined market share of AWS and Microsoft Azure, which exceeds 50% in the European Union. But the DMA applies to individual undertakings and their services. Designation attaches to each CPS separately: Article 3(9) directs the Commission to list only those services that “individually” are an important gateway, and the General Court confirmed in Meta that the Commission must appraise each undertaking and each CPS on its own facts, without reference to other undertakings or services.[69]

The DMA also contains no counterpart to collective dominance under Article 102 TFEU. A combined market share therefore says little about whether AWS or Azure individually holds an entrenched position. It joins the shares of two firms that compete intensely with each other, while Google Cloud and Oracle add further competitive pressure.

This problem also complicates the Commission’s decision to investigate AWS and Azure while excluding Google Cloud.[70] The Commission must identify an objective distinction between Amazon and Microsoft, on one hand, and Google, on the other.

If the distinction rests on user numbers, the designation inquiry returns to size. If it rests on a line between Google’s roughly 14% worldwide share and Microsoft’s roughly 21% worldwide share, the DMA provides no such threshold. Amazon and Microsoft can therefore argue that the Commission has offered no legitimate basis for treating Google differently that does not reduce to the kind of user counting that Article 3(8) was meant to replace when Article 3(2)(b)’s presumptions did not apply.

The General Court rejected an equal-treatment claim in ByteDance. When a service meets the statutory thresholds, designation is mandatory, and one undertaking cannot challenge its own designation by pointing to the Commission’s treatment of another.[71] That reasoning fits Article 3(2), where the thresholds dictate the result. It does not, however, transfer easily to Article 3(8).

Subthreshold designation under Article 3(8) depends on Commission discretion. The Commission chooses which services to investigate and which to designate. Equal-treatment principles carry particular force when an institution exercises that kind of discretion. Comparable situations may not be treated differently unless an objective justification supports the distinction.[72]

Article 3(8) therefore reopens the issue that ByteDance closed. Because subthreshold designation results from an exercise of discretion rather than the automatic application of numerical thresholds, the Commission’s choice of whom to designate—and whom to exclude—must withstand equal-treatment review.

The point is not that Google Cloud should also be designated. The exclusion instead reveals the criteria driving the Commission’s decisions. If the line between designation and non-designation rests on relative user counts below the statutory thresholds, then the Commission is relying on an unstated measure of size. The numbers are not high enough to trigger Article 3(2), yet the Commission treats them as high enough under Article 3(8). That approach replaces the DMA’s express quantitative thresholds with an undisclosed set of lower ones.

D. Foreseeable Entrenchment Requires Evidence

The European Commission may argue that the competitive problem is forward-looking. Article 3(1)(c) permits designation when it is “foreseeable” that an undertaking will attain an entrenched and durable position “in the near future.” The Commission’s public framing points in that direction. It describes cloud computing as “a prerequisite for AI,” suggests that AI tools and partnerships influence procurement decisions, and presents the investigations in terms of European technological sovereignty.[73]

On that account, the Commission could concede that AWS and Microsoft Azure face meaningful competition today while predicting that AI workloads, “data gravity,” and sovereign-cloud requirements will entrench them tomorrow. Data gravity refers to the tendency of large datasets to attract related applications, services, and computing resources because moving the data becomes costly or impractical.

But a prediction of foreseeable entrenchment remains a claim about market contestability, even when it looks ahead. It must therefore satisfy the reasoning standard the General Court applied in Meta Platforms. The Court annulled the Facebook Marketplace designation on two grounds. First, the legality of a designation had to be assessed “on the basis of the facts and the law as they stood at the time when the measure was adopted.” Second, the Commission’s treatment of a material factual change was “vague and hypothetical” and lacked any “specific analysis.”[74]

The Court applied the ordinary standard governing statements of reasons under European Union law:

[T]he statement of reasons required by … Article 296 TFEU … must disclose in a clear and unequivocal fashion the reasoning followed by the institution which adopted the measure … to enable the persons concerned to ascertain the reasons for it and to enable the court having jurisdiction to exercise its power of review.[75]

By the time of the judgment, the Commission had already removed Marketplace’s designation after its user numbers fell below the thresholds. The ruling nevertheless matters because of its ex tunc effect and the evidentiary standard it establishes.[76]

A finding that cloud services are large and AI will lock in their positions would be conclusory under Meta Platforms. To survive review, the Commission would need to identify the mechanism by which AI demand will convert today’s contested positions into entrenched ones. It would also need to define the relevant time horizon, because the “near future” cannot mean an open-ended period, and provide evidence available at the time of adoption showing that competition will cease to constrain AWS and Azure.

The current AI record points in the opposite direction. CERRE’s 2025 assessments conclude that competition in cloud computing and AI remains strong, both in present conditions and over time, and caution against broad ex ante intervention.[77] Model developers commonly use several cloud providers, and the leading cloud–AI partnerships have generally remained nonexclusive.[78]

The CMA’s work on foundation models and the Autorité de la concurrence’s opinion on generative AI identify access to computing capacity as a possible future concern, while documenting substantial current competition.[79] The FTC’s Section 6(b) study of cloud–AI partnerships raises questions about spending commitments and vertical relationships, but it offers no evidence that cloud-infrastructure markets are tipping.[80]

Entry also weakens the claim that AI will deepen incumbent moats. AI demand has helped Oracle gain share and has supported the emergence of specialized “neocloud” providers, which now account for roughly 5% of the market and are expected to continue growing rapidly.[81] Cloud computing has also lowered barriers to AI adoption by giving startups access to advanced models and computing resources through services such as Amazon Bedrock and Azure OpenAI Service.[82] The development the Commission portrays as a future source of entrenchment has so far expanded access and encouraged entry.

The broader political context also warrants attention. The Commission proposed the Cloud and AI Development Act (CADA) at the same time as the preliminary designations. CADA is an industrial-policy measure expressly motivated by the decline in EU providers’ market share from roughly 29% in 2017 to 15% in 2022. Read alongside CADA and Mario Draghi’s account of European dependence on non-European computing capacity, the cloud investigations appear partly connected to technological-sovereignty goals.[83]

Those goals may support separate industrial-policy measures. They do not, however, alter the legal test under the DMA. Dependence on foreign providers does not make a company a gatekeeper, and technological sovereignty cannot substitute for evidence that AWS or Azure holds, or will soon hold, an entrenched and durable position.

IV. Proportionality and the Available Alternatives

The preceding section assessed whether the proposed designations satisfy Article 3(1). This section asks a separate question. Even if the European Commission could establish those conditions, is designation a proportionate response to the competitive frictions documented in the cloud market?

Two comparisons guide the analysis. Section IV.A examines the United Kingdom, where competition authorities reviewed the same firms on an extensive evidentiary record and chose voluntary commitments rather than designation under the country’s comparable ex ante regime. Section IV.B turns to the European Union’s Data Act, a purpose-built switching regime whose central prohibition has not yet taken effect. It also considers the Commission’s parallel inquiry into whether the DMA’s existing obligations can address cloud-specific concerns.

Both comparisons point toward narrower measures. Designation is the most intrusive option available, yet the Commission is pursuing it before testing the tailored remedies already adopted or determining whether the obligations it would trigger fit the conduct at issue.

A. The United Kingdom Chooses a Lighter Touch

The United Kingdom examined the same market and the same two firms under a regime whose gatekeeper logic closely resembles that of the DMA. It reached a different conclusion.

In 2023, Ofcom referred cloud-infrastructure services to the CMA. Ofcom found that AWS and Microsoft held a combined U.K. market share of 70% to 80%. It identified egress fees, technical-interoperability barriers, and committed-spend discounts as impediments to switching, while also documenting strong competition for new customers.[84]

The CMA completed its market investigation in July 2025. It concluded that AWS and Microsoft each held significant and entrenched unilateral market power, protected by high barriers to entry. The CMA also identified adverse effects on competition arising from switching frictions and Microsoft’s software-licensing practices. Its inquiry group recommended that the CMA Board prioritize Strategic Market Status (SMS) investigations into both companies’ cloud businesses under the Digital Markets, Competition and Consumers Act 2024 (DMCCA).[85]

The Board reviewed the same record and declined to follow that recommendation. In March 2026, it accepted a package of voluntary commitments addressing egress and interoperability, which the CMA cannot itself enforce, including the extension of European Union Data Act switching standards to U.K. customers. The CMA said it would review progress in six months’ time.

The Board directed the concern it considered serious enough for possible designation—Microsoft’s software-licensing practices—into an SMS investigation of Microsoft’s business-software offerings rather than cloud infrastructure itself.[86]

The decision drew criticism. Practitioners argued that the Board had departed from the inquiry group’s recommendation without offering a fully persuasive explanation, and coalitions representing challenger providers objected strongly.[87] Nor did the Board foreclose a future SMS investigation into cloud services.

Still, the authority with the most extensive evidentiary record assembled on AWS and Microsoft concluded that voluntary commitments offered a proportionate response in a market that remains competitive and continues to change. The decision reflects the risk that broad intervention in a market central to AI-driven growth could condemn competitive conduct or impede investment. Negotiated remedies, by contrast, can address identified practices while reducing that risk.

The European Commission is pursuing a markedly different course under the DMA, although the comparison is not exact. The DMCCA requires findings of substantial and entrenched market power and strategic significance, while the DMA formally disclaims a market-power test.[88]

Yet the qualitative routes under the two regimes converge on the same economic question, as Section III.B explains. Both require an assessment of contestability. The U.K. authority made an express finding of market power but still chose targeted commitments. The Commission proposes broader intervention on a thinner basis.

The proposed designations would be the first to cover services that satisfy neither user threshold. They would also subject AWS and Azure to the DMA’s full set of obligations in a market that the CMA recently concluded could be addressed through narrower, voluntary measures.

B. The Data Act and the Proportionality Problem

Proportionality presents a final obstacle to designation. The European Union has already enacted a purpose-built regime for the specific frictions documented in the cloud market.

Chapter VI of the Data Act applies to all providers of “data processing services,” rather than only designated gatekeepers. It requires providers to remove commercial, technical, contractual, and organizational barriers to switching. The law caps switching charges, including data-egress fees, at cost during a transitional period and prohibits them entirely beginning Jan. 12, 2027. It also requires functional equivalence and open interoperability specifications, supported by European Commission-endorsed standard contractual clauses.[89]

Scholars have examined the economics and legal structure of this regime in detail.[90] The national competition authorities that studied cloud services most closely also identified the Data Act as the appropriate remedial instrument. In 2023, France’s Autorité de la concurrence concluded that existing competition-law tools and the proposed Data Act could address most of the risks it identified. The Netherlands Authority for Consumers and Markets likewise pointed to the Data Act as the proper means of addressing cloud lock-in.[91] As Section IV.A explains, the United Kingdom’s remedial package also incorporates Data Act switching standards.

Against that background, several commentators argue that the Commission should explain why the tailored regime is inadequate before invoking the DMA. That requirement would follow from proportionality and the need for coherence across the European Union’s digital laws.[92] Designation would impose the DMA’s broadest obligations on services already governed by a specialized switching regime whose central prohibition has not yet taken effect.

The Commission appears to recognize the mismatch. On the same day it opened the designation investigations, it launched a separate investigation under Article 19 of the DMA into whether the regulation’s existing obligations can address the practices identified in cloud services. Those practices include interoperability barriers, restrictions on data access, tying and bundling, and imbalanced contract terms.

The Commission must issue its Article 19 report within 18 months. The report could support delegated acts under Articles 12 and 49.[93] It is expected around May 2027, six months after the November 2026 deadline for the designation decisions.

The sequence is backward. The Commission proposes to designate AWS and Azure before determining whether the obligations triggered by designation fit the cloud industry.[94] Commentators have already observed that none of the DMA’s current obligations directly addresses the concerns documented in cloud markets.

Designation without suitable obligations would impose substantial compliance costs while producing little improvement in contestability. A proportionate response should connect the identified harm to a remedy capable of addressing it. The Commission’s parallel investigation suggests that this connection has not yet been established.

The Commission’s first DMA review report, published in April 2026, presents the cloud investigations as evidence that the regulation’s “future-proofing tools” are working and concludes that no legislative amendment is needed.[95] But the investigations cannot validate the DMA’s flexibility before the Commission has established that designation is lawful and that the resulting obligations are appropriate. Treating the proposed designations as proof of the framework’s success assumes the very proposition the Commission must still demonstrate.

V. Conclusion: When Size Becomes Gatekeeping

The arguments developed here do not guarantee that the proposed cloud designations will fail. Article 3 permits two competing interpretations. Under the narrower reading, the qualitative criteria filter out firms that are large but do not control genuine gateways. Under the broader reading, a large company that provides a CPS may qualify as a gatekeeper largely because the DMA assumes that size, durability, and integration in digital markets tend to produce gatekeeper power.

The General Court’s first two DMA judgments suggest how the courts may police that boundary. ByteDance and Meta both involved the statutory presumptions. The Commission prevailed when the numerical thresholds supported the qualitative criteria.[96] When the Court annulled the Facebook Marketplace designation, it focused on the quality of the Commission’s reasoning. It required the Commission to assess the facts as they existed when the decision was adopted and to provide a concrete analysis of material developments.[97]

The courts need not import Article 102 TFEU’s market-power test into the DMA. They can protect the same economic substance by requiring the Commission to prove each element of a subthreshold designation separately. The pending ByteDance appeal may further define how closely the courts will review that reasoning.[98]

The AWS and Azure investigations therefore present a consequential test. In every previous designation, a quantitative presumption largely carried the substantive inquiry. Article 3(8) offers no such shortcut. The Commission must prove that each cloud service is an important gateway under Article 3(1)(b), and that the undertaking enjoys, or foreseeably will enjoy, an entrenched and durable position in its operations under Article 3(1)(c).[99]

The current record makes that task difficult. Quality-adjusted prices have fallen, market shares have shifted, multihoming is widespread, providers have reduced switching costs, and new entrants are capturing demand associated with AI. The Commission has also excluded Google Cloud without explaining the distinction, while relying at times on the combined position of two firms that compete against each other. Meanwhile, the Data Act’s tailored switching rules have not yet taken full effect, and the United Kingdom chose voluntary commitments after examining the same market on a fuller record.

If those facts nevertheless suffice for designation, the DMA’s qualitative criteria will have done little more than repackage size and commercial importance. “Gateway” will mean indispensability as an input rather than intermediation between users. “Entrenched and durable” will describe scale rather than limited contestability. Forward-looking claims about AI and data gravity will substitute for evidence that rivalry is likely to weaken.

The alternative is to give each statutory condition independent force. “Gateway” should require intermediation, not mere criticality. “Entrenched and durable” should require evidence that rivals cannot discipline the incumbent. Predictions about future entrenchment should satisfy Article 296’s demand for specific and contemporaneous reasoning. Proportionality should also require the Commission to explain why the Data Act and targeted enforcement cannot address the documented concerns.

Under that reading, the Commission should not convert its preliminary position into a final designation. If it does, the courts should require the evidence and reasoning that Article 3(8) demands.

[1] Press Release, Eur. Comm’n, Commission Reaches Preliminary Position that Amazon’s and Microsoft’s Market-Leading Cloud Services Should Be Designated Under the Digital Markets Act (June 25, 2026), https://ec.europa.eu/commission/presscorner/detail/en/ip_26_1444 [hereinafter Preliminary Position]; see Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on Contestable and Fair Markets in the Digital Sector and Amending Directives (EU) 2019/1937 and (EU) 2020/1828 (Digital Markets Act), 2022 O.J. (L 265) 1 [hereinafter DMA].

[2] Eur. Comm’n, Commission Launches Market Investigations on Cloud Computing Services Under the Digital Markets Act (Nov. 18, 2025), https://digital-markets-act.ec.europa.eu/commission-launches-market-investigations-cloud-computing-services-under-digital-markets-act-2025-11-18_en. The Commission simultaneously opened a third investigation under Article 19 of the DMA to determine whether the regulation’s existing obligations can effectively address practices in the cloud sector. See infra Section IV.B.

[3] Preliminary Position, supra note 1 (reporting that final decisions are due by November 2026 under the 12-month deadline for market investigations); DMA, supra note 1, arts. 3(10), 17 (providing a six-month compliance window following designation).

[4] Commission Decision of 29.4.2024, Apple—iPadOS, Case DMA.100047, https://ec.europa.eu/competition/digital_markets_act/cases/202427/DMA_100047_5491.pdf [hereinafter iPadOS].

[5] Id.; see also Press Release, Eur. Comm’n, Commission Designates Apple’s iPadOS Under the Digital Markets Act (Apr. 29, 2024), https://ec.europa.eu/commission/presscorner/detail/en/ip_24_2363 (“Apple’s business user numbers exceeded the quantitative threshold elevenfold, while its end user numbers were close to the threshold and are predicted to rise in the near future.”).

[6] Lazar Radic, Cloudy Logic: The DMA’s Search for a Gatekeeper, Truth on the Mkt. (Apr. 14, 2026), https://truthonthemarket.com/2026/04/14/cloudy-logic-the-dmas-search-for-a-gatekeeper.

[7] DMA, supra note 1, art. 3(1). The three criteria and their accompanying presumptions are discussed infra Section II.A.

[8] DMA, supra note 1, art. 3(1)(a)–(c).

[9] Id. art. 3(2)(a)–(c).

[10] Case T-1077/23, ByteDance Ltd. v. Comm’n, ECLI:EU:T:2024:478, ¶ 71 (July 17, 2024) [hereinafter ByteDance] (holding that arguments submitted under Article 3(5) must “manifestly call into question” the presumptions—a standard requiring a high degree of plausibility). An appeal is pending before the Court of Justice. See infra note 98.

[11] DMA, supra note 1, arts. 3(8), 17.

[12] Friso Bostoen & Giorgio Monti, The Rhyme and Reason of Gatekeeper Designation Under the Digital Markets Act, J. Antitrust Enf’t (advance article 2025), jnae054, https://doi.org/10.1093/jaenfo/jnae054 [hereinafter Bostoen & Monti].

[13] DMA, supra note 1, recital 23 (authorizing designation based on a qualitative assessment conducted through a market investigation).

[14] Cf. Bostoen & Monti, supra note 12 (documenting the evidence used to assess each criterion under Article 3(1)).

[15] Jacques Crémer, Yves-Alexandre de Montjoye & Heike Schweitzer, Competition Policy for the Digital Era (2019) (report prepared for the Eur. Comm’n); Digit. Competition Expert Panel, Unlocking Digital Competition (2019) (U.K.) [hereinafter Furman Review]; Stigler Comm. on Digit. Platforms, Final Report (2019).

[16] Nicolas Petit, The Proposed Digital Markets Act (DMA): A Legal and Policy Review, 12 J. Eur. Competition L. & Prac. 529 (2021).

[17] Luís Cabral et al., The EU Digital Markets Act: A Report from a Panel of Economic Experts (Joint Rsch. Ctr. 2021).

[18] Damien Geradin, What Is a Digital Gatekeeper? Which Platforms Should Be Captured by the EC Proposal for a Digital Markets Act? 20 (Feb. 2021) (unpublished manuscript) (“The methodology used to designate gatekeepers cannot be divorced from the issue(s) that ex ante regulation seeks to address, as otherwise the DMA might end up regulating the wrong set of companies.”).

[19] Pinar Akman, Regulating Competition in Digital Platform Markets: A Critical Assessment of the Framework and Approach of the EU Digital Markets Act, 71 Int’l & Comp. L.Q. 85 (2022).

[20] Pablo Ibáñez Colomo, The Draft Digital Markets Act: A Legal and Institutional Analysis, 12 J. Eur. Competition L. & Prac. 561 (2021).

[21] See, e.g., Giuseppe Colangelo, The European Digital Markets Act and Antitrust Enforcement: A Liaison Dangereuse, 47 Eur. L. Rev. 597 (2022); Viktoria H.S.E. Robertson, The Complementary Nature of the Digital Markets Act and the EU Antitrust Rules, 12 J. Antitrust Enf’t 325 (2024).

[22] See, e.g., Fiona Scott Morton & Cristina Caffarra, The European Commission Digital Markets Act: A Translation, VoxEU (Jan. 5, 2021).

[23] Bostoen & Monti, supra note 12 (surveying the 34 designation decisions adopted at the time of publication and analyzing how each applied the designation criteria, as well as the type I- and type II-error-correction functions of Articles 3(5) and 3(8)).

[24] Eur. Comm’n, Commission Concludes that Online Social Networking Service X Should Not Be Designated Under the Digital Markets Act (Oct. 16, 2024), https://digital-markets-act.ec.europa.eu/commission-concludes-online-social-networking-service-x-should-not-be-designated-under-digital-2024-10-16_en.

[25] Commission Decision of 5.9.2023, Alphabet, Case DMA.100011, https://ec.europa.eu/competition/digital_markets_act/cases/202344/DMA_100011_147.pdf (finding that Gmail was not an important gateway despite meeting the user-number thresholds); Commission Decision of 5.9.2023, Microsoft, Case DMA.100023, https://ec.europa.eu/competition/digital_markets_act/cases/202344/DMA_100023_115.pdf (reaching the same conclusion for Outlook.com).

[26] Press Release, Eur. Comm’n, Commission Closes Market Investigations into Microsoft’s and Apple’s Services Under the Digital Markets Act (Feb. 2024), https://ec.europa.eu/commission/presscorner/detail/en/mex_24_785 (addressing iMessage, Bing, Edge, and Microsoft Advertising).

[27] See Press Release, Eur. Comm’n, Digital Markets Act: Commission Designates Six Gatekeepers, IP/23/4328 (Sept. 6, 2023) (declining to designate Samsung because its browser was not an important gateway); Bostoen & Monti, supra note 12 (discussing the accepted rebuttals concerning TikTok Ads and X Ads).

[28] Autorité de la concurrence, Opinion 23-A-08 of 29 June 2023 on Competition in the Cloud Computing Sector, https://www.autoritedelaconcurrence.fr/en/press-release/cloud-computing-autorite-de-la-concurrence-issues-its-market-study-competition-cloud; Autoriteit Consument & Markt, Market Study Cloud Services (Sept. 5, 2022), https://www.acm.nl/system/files/documents/public-market-study-cloud-services.pdf; Japan Fair Trade Comm’n, Report Regarding Cloud Services (June 2022); OECD, Competition in the Provision of Cloud Computing Services (2025); Press Release, Fed. Trade Comm’n, FTC Seeks Comment on Business Practices of Cloud Computing Providers that Could Impact Competition and Data Security (Mar. 2023).

[29] See generally On-Topic: Cloud and Competition Policy, Concurrences No. 8-2025 (Aug. 2025) (Parts I–VIII, examining regulation in the age of hyperscalers, interoperability, policy responses to competition concerns, and cloud–AI partnerships).

[30] Antonio Manganelli & Daniel Schnurr, Competition and Regulation of Cloud Computing Services: Economic Analysis and Review of EU Policies (Ctr. on Regul. in Eur., Feb. 2024), https://cerre.eu/publications/competition-and-regulation-of-cloud-computing-services-economic-analysis-and-reviewbrof-eu-policies [hereinafter CERRE 2024]; see especially id. Recommendation 2 (urging proportionality when designating cloud services under the DMA because they differ from other core platform services).

[31] Konstantina Bania & Damien Geradin, The Regulation of Cloud Computing: Why the European Union Failed to Get It Right, 33 Info. & Commc’ns Tech. L. 1 (2024), https://doi.org/10.1080/13600834.2023.2260687.

[32] Selçukhan Ünekbas, Why Use the Digital Markets Act for Cloud?, Oxford Bus. L. Blog (May 4, 2026), https://blogs.law.ox.ac.uk/oblb/blog-post/2026/05/why-use-digital-markets-act-cloud.

[33] Kalpana Tyagi, Can Europe’s Digital Markets Act and Data Act Rein in Cloud Hyperscalers?, Tech Pol’y Press (Feb. 3, 2026), https://www.techpolicy.press/can-europes-digital-markets-act-and-data-act-rein-in-cloud-hyperscalers.

[34] Vanderbilt Pol’y Accelerator, How to Regulate the Cloud: A Blueprint to Address the Market Failures and National Security Risks of Cloud Computing (2025), https://cdn.vanderbilt.edu/vu-URL/wp-content/uploads/sites/412/2025/09/18140135/How-to-Regulate-the-Cloud.pdf; cf. Tejas N. Narechania & Ganesh Sitaraman, An Antimonopoly Approach to Governing Artificial Intelligence, 43 Yale L. & Pol’y Rev. 95 (2024) (treating cloud computing as an infrastructure layer that warrants regulation).

[35] See, e.g., The EU Commission Preliminarily Finds that 2 Leading Cloud Computing Providers Should Be Designated as Gatekeepers Under the DMA Despite Falling Below the Quantitative Thresholds (Amazon/Microsoft), e-Competitions (June 2026); Selçukhan Ünekbas, The EU’s Facebook Marketplace Decision: The Gatekeeper That Wasn’t, Truth on the Mkt. (June 4, 2026), https://laweconcenter.org/resources/the-eus-facebook-marketplace-decision-the-gatekeeper-that-wasnt (drawing implications from the Meta ruling for qualitative designations).

[36] DMA, supra note 1, art. 2(2)(i).

[37] Id. recital 14.

[38] Bania & Geradin, supra note 31 (arguing that the obligations imposed by the Digital Markets Act, Digital Services Act, and Data Act fit poorly with the technical and commercial characteristics of cloud services).

[39] Tyagi, supra note 33 (reconstructing the impact assessment’s ecosystem rationale for including cloud services and arguing that qualitative designation is better suited to innovation-intensive, business-facing markets).

[40] Preliminary Position, supra note 1.

[41] ByteDance, supra note 10, ¶ 210.

[42] Jean-Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 1 J. Eur. Econ. Ass’n 990 (2003); Mark Armstrong, Competition in Two-Sided Markets, 37 RAND J. Econ. 668 (2006).

[43] Even these canonical platform markets do not fit neatly within the two-sided-market framework. See Dirk Auer & Nicolas Petit, Two-Sided Markets and the Challenge of Turning Economic Theory into Antitrust Policy, 60 Antitrust Bull. 426 (2015).

[44] Preliminary Position, supra note 1; see also Eur. Comm’n, Commission Reaches Preliminary Position that Amazon’s and Microsoft’s Market-Leading Cloud Services Should Be Designated Under the Digital Markets Act, Digit. Strategy (June 25, 2026), https://digital-strategy.ec.europa.eu/en/news/commission-reaches-preliminary-position-amazons-and-microsofts-market-leading-cloud-services-should (quoting Executive Vice Presidents Teresa Ribera and Henna Virkkunen).

[45] The firms facing designation make nearly identical arguments. See Amazon, Why Applying the DMA to Cloud Would Regulate Away EU Competitiveness and Resiliency (June 2026), https://www.aboutamazon.eu/news/policy/why-applying-the-DMA-to-cloud-would-regulate-away-eu-competitiveness-and-resiliency (arguing that cloud services involve no intermediation, cross-side network effects, or gateway function). We cite the paper as a party submission, not as independent authority.

[46] See Eur. Comm’n, Commission Concludes that Online Social Networking Service X Should Not Be Designated Under the Digital Markets Act, supra note 24.

[47] See Commission Decision of 5.9.2023, Alphabet, supra note 25; Commission Decision of 5.9.2023, Microsoft, supra note 25.

[48] See Eur. Comm’n, Commission Closes Market Investigations into Microsoft’s and Apple’s Services Under the Digital Markets Act, supra note 26.

[49] iPadOS, supra note 4; see also Bostoen & Monti, supra note 12 (analyzing the decision’s reasoning and reliance on projected end-user growth).

[50] Crémer, de Montjoye & Schweitzer, supra note 15; Furman Review, supra note 15.

[51] ByteDance, supra note 10, paras 45-46 (rejecting arguments that designation requires market definition or proof of market power).

[52] See Robertson, supra note 21; Georgios Gryllos, The New Digital Landscape: Interaction Between the DMA and Rules of National and EU Law Governing the Conduct of Gatekeepers, Concurrences No. 1-2024, Art. No. 116827, at 40 (2024).

[53] ByteDance, supra note 10, ¶¶ 296–297.

[54] William J. Baumol, Contestable Markets: An Uprising in the Theory of Industry Structure, 72 Am. Econ. Rev. 1 (1982); William J. Baumol, John C. Panzar & Robert D. Willig, Contestable Markets and the Theory of Industry Structure (1982).

[55] Chad Syverson, Macroeconomics and Market Power: Context, Implications, and Open Questions, 33 J. Econ. Persps. 23 (2019).

[56] Synergy Rsch. Grp., Cloud Market Annual Revenue Run Rate Topped Half a Trillion Dollars in Q1 as Growth Surge Continues (Apr. 2026), https://www.srgresearch.com/articles/cloud-market-annual-revenue-run-rate-topped-half-a-trillion-dollars-in-q1-as-growth-surge-continues (reporting first-quarter 2026 market shares of 28% for Amazon Web Services, 21% for Microsoft, and 14% for Google, with neoclouds accounting for 5% of the market and five providers ranking among the top 30); Synergy Rsch. Grp., Cloud Market Share Trends: Big Three Together Hold 63%, While Oracle and the Neoclouds Inch Higher (2025), https://www.srgresearch.com/articles/cloud-market-share-trends-big-three-together-hold-63-while-oracle-and-the-neoclouds-inch-higher (reporting that Amazon Web Services’ share declined from roughly 32% in 2021).

[57] David Byrne, Carol Corrado & Daniel E. Sichel, The Rise of Cloud Computing: Minding Your P’s, Q’s and K’s (Nat’l Bureau of Econ. Rsch., Working Paper No. 25188, 2018), https://www.nber.org/papers/w25188, reprinted in Measuring and Accounting for Innovation in the Twenty-First Century (Carol Corrado et al. eds., 2020).

[58] See Wasabi, Cloud Storage Fee Inflation, Wasabi Blog (Sept. 29, 2022), https://wasabi.com/blog/cost-optimization/cloud-storage-fee-inflation (reporting that Amazon S3 list prices fell by more than 80% in the early 2010s).

[59] Paul Klemperer, Competition When Consumers Have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics, and International Trade, 62 Rev. Econ. Stud. 515 (1995); Joseph Farrell & Paul Klemperer, Coordination and Lock-In: Competition with Switching Costs and Network Effects, in 3 Handbook of Industrial Organization 1967 (Mark Armstrong & Robert Porter eds., 2007).

[60] Ofcom, Cloud Services Market Study: Final Report (Oct. 5, 2023), https://www.ofcom.org.uk/siteassets/resources/documents/consultations/category-3-4-weeks/244808-cloud-services-market-study/associated-documents/cloud-services-market-study-final-report.pdf [hereinafter Ofcom Report] (documenting strong competition for new customers); Competition & Mkts. Auth., Cloud Services Market Investigation: Summary of Final Decision (July 31, 2025), https://assets.publishing.service.gov.uk/media/688b20e6ff8c05468cb7b120/summary_of_final_decision.pdf [hereinafter CMA Final Decision].

[61] See Google Cloud, Eliminating Data Transfer Fees When Migrating Off Google Cloud (Jan. 2024), https://cloud.google.com/blog/products/networking/eliminating-data-transfer-fees-when-migrating-off-google-cloud; Amazon Web Servs., Free Data Transfer Out to Internet When Moving Out of AWS (Mar. 2024), https://aws.amazon.com/blogs/aws/free-data-transfer-out-to-internet-when-moving-out-of-aws; Microsoft Azure, Now Available: Free Data Transfer Out to Internet When Leaving Azure (Mar. 2024), https://azure.microsoft.com/en-us/updates?id=now-available-free-data-transfer-out-to-internet-when-leaving-azure.

[62] Regulation (EU) 2023/2854 of the European Parliament and of the Council of 13 December 2023 on Harmonised Rules on Fair Access to and Use of Data (Data Act), 2023 O.J. (L 2854) 1, arts. 23–31 [hereinafter Data Act]. For a more detailed discussion, see infra Section IV.B.

[63] Flexera, State of the Cloud Report (2026 ed.), https://info.flexera.com/CM-REPORT-State-of-the-Cloud

[64] CMA Final Decision, supra note 60; Ofcom Report, supra note 60 (both distinguishing multicloud architectures from workload-level switching).

[65] CMA Final Decision, supra note 60 (finding that Amazon Web Services and Microsoft each held U.K. market shares by value of up to 30%–40% and recommending that the CMA Board prioritize Strategic Market Status investigations into both providers’ cloud activities).

[66] Ofcom Report, supra note 60.

[67] CMA Final Decision, supra note 60 (finding materially higher effective prices for running Windows Server and SQL Server on rival cloud platforms than on Azure); see also CISPE, CISPE and Microsoft Agree Settlement in Fair Software Licensing Case (July 11, 2024), https://www.cispe.cloud/cispe-and-microsoft-agree-settlement-in-fair-software-licensing-case (settling the association’s 2022 complaint on terms that excluded Amazon Web Services, Google, and Alibaba).

[68] See Google Files EU Antitrust Complaint Accusing Microsoft of Stifling Cloud Competition, CNBC (Sept. 25, 2024), https://www.cnbc.com/2024/09/25/google-files-eu-antitrust-complaint-accusing-microsoft-of-stifling-cloud-competition.html. Google withdrew the complaint on Nov. 28, 2025, after the European Commission opened its cloud-market investigations. See Caroline Donnelly, Google Cloud Withdraws Complaint with European Commission over Microsoft’s Cloud Licensing Tactics, Computer Weekly (Dec. 1, 2025), https://www.computerweekly.com/news/366635494/Google-Cloud-withdraws-complaint-with-European-Commission-over-Microsofts-cloud-licensing-tactics.

[69] DMA, supra note 1, art. 3(9); Case T-1078/23, Meta Platforms, Inc. v. Comm’n, ECLI:EU:T:2026:357, paras 64 and 174  (Gen. Ct. June 3, 2026) [hereinafter Meta] (upholding Messenger’s designation as a distinct core platform service that individually constitutes an important gateway).

[70] See European Commission Lines Up Amazon and Microsoft for Cloud Gatekeeper Status, The Register (June 25, 2026), https://www.theregister.com/legal/2026/06/25/european-commission-lines-up-amazon-and-microsoft-for-cloud-gatekeeper-status/5262127; cf. Article 19, Investigations into Cloud Computing Under DMA a Welcome Move (Nov. 18, 2025), https://www.article19.org/resources/europe-investigations-into-cloud-computing-under-dma-a-welcome-move (noting, from a pro-designation perspective, the puzzle posed by Google Cloud’s exclusion).

[71] ByteDance, supra note 10 (rejecting the claim that the designation breached the principle of equal treatment).

[72] See, e.g., Case C-127/07, Arcelor Atlantique et Lorraine, ECLI:EU:C:2008:728, ¶ 23 (holding that comparable situations must not be treated differently, and different situations must not be treated alike, unless the difference is objectively justified).

[73] Preliminary Position, supra note 1 (describing cloud computing as “a prerequisite for AI”); see also sources cited supra note 45.

[74] Meta, supra note 69; see also Ct. Just. Eur. Union, Press Release No. 77/26 (June 3, 2026), https://curia.europa.eu/site/upload/docs/application/pdf/2026-06/cp260077en.pdf.

[75] Meta, supra note 69, ¶ 53.

[76] See Alba Ribera Martínez, The Chicken or Egg Dilemma: The General Court Partially Annuls the European Commission’s DMA Designation Decision Against Meta (Case T-1078/23), Kluwer Competition L. Blog (June 8, 2026), https://legalblogs.wolterskluwer.com/competition-blog/the-chicken-or-egg-dilemma-the-general-court-partially-annuls-the-european-commissions-dma-designation-decision-against-meta-case-t-107823 (discussing the legal-interest analysis and the probative-effort question).

[77] Zach Meyers & Marc Bourreau, A Competition Policy for Cloud and AI (Ctr. on Regul. in Eur., June 2025), https://cerre.eu/wp-content/uploads/2025/06/A-Competition-Policy-for-Cloud-and-AI_FINAL.pdf; Zach Meyers & Marc Bourreau, What Policy Interventions for a Competitive AI Sector? (Ctr. on Regul. in Eur., July 2025), https://cerre.eu/publications/what-policy-interventions-for-a-competitive-ai-sector.

[78] Christophe Carugati, The Competitive Relationship Between Cloud Computing and Generative AI (Bruegel, Working Paper No. 19/2023), https://www.bruegel.org/system/files/2023-12/WP%202023%2019%20Cloud%20111223.pdf (mapping largely nonexclusive cloud–AI partnerships and multihoming by model developers).

[79] Competition & Mkts. Auth., AI Foundation Models: Initial Report (Sept. 2023); Competition & Mkts. Auth., AI Foundation Models: Update Paper (Apr. 2024); Autorité de la concurrence, Opinion 24-A-05 of 28 June 2024 on the Competitive Functioning of the Generative Artificial Intelligence Sector.

[80] Fed. Trade Comm’n, Partnerships Between Cloud Service Providers and AI Developers: FTC Staff Report on AI Partnerships & Investments 6(b) Study (Jan. 2025), https://www.ftc.gov/system/files/ftc_gov/pdf/p246201_aipartnerships6breport_redacted_0.pdf.

[81] Synergy Rsch. Grp., Neocloud Market Forecast to Approach $400B by 2031, Driven by Surging AI Infrastructure Demand (2026), https://www.srgresearch.com/articles/neocloud-market-forecast-to-approach-400b-by-2031-driven-by-surging-ai-infrastructure-demand; see also sources cited supra note 56.

[82] Comments of the Int’l Ctr. for L. & Econ., In re Cloud Computing Request for Information, Fed. Trade Comm’n (June 2023), https://laweconcenter.org/resources/icle-response-to-the-ftcs-cloud-computing-rfi.

[83] Eur. Comm’n, Proposal for a Regulation Establishing a Framework of Measures for Strengthening Europe’s Cloud and AI Ecosystem (Cloud and AI Development Act), COM (2026) 502 final (June 3, 2026), https://digital-strategy.ec.europa.eu/en/policies/cloud-and-ai-development-act (noting that EU providers’ market share fell from roughly 29% in 2017 to 15% in 2022); Mario Draghi, The Future of European Competitiveness (Sept. 2024).

[84] Ofcom Report, supra note 60 (reporting a combined U.K. market share of 70%–80% and identifying frictions from egress fees, technical-interoperability barriers, and committed-spend discounts).

[85] CMA Final Decision, supra note 60.

[86] Press Release, Competition & Mkts. Auth., CMA Announces Package of Actions on Business Software and Cloud Services (Mar. 31, 2026), https://www.gov.uk/government/news/cma-announces-package-of-actions-on-business-software-and-cloud-services. The Strategic Market Status investigation into Microsoft’s business-software ecosystem began in May 2026, with a designation decision expected around February 2027.

[87] Christophe Humpe & Greg Dowell, CMA Decision to Shelve Cloud Services SMS Investigations Raises Questions, Macfarlanes (Apr. 2026), https://www.macfarlanes.com/insights/102mqfd/cma-decision-to-shelve-cloud-services-sms-investigations-raises-questions-around; see also Antony Adshead, CMA to Launch Strategic Market Status Investigation into Microsoft; Amazon Web Services Off the Hook, Computer Weekly (Mar. 2026), https://www.computerweekly.com/news/366640828/CMA-to-launch-strategic-market-status-investigation-into-Microsoft-Amazon-Web-Services-off-the-hook (collecting reactions from challenger providers).

[88] Digital Markets, Competition and Consumers Act 2024, c. 13, §§ 2, 5–6 (UK).

[89] Data Act, supra note 62, ch. VI, arts. 23–31, 34–35; see also Eur. Comm’n, Data Act Explained, Digit. Strategy, https://digital-strategy.ec.europa.eu/en/factpages/data-act-explained (explaining that switching charges are capped at cost during a transitional period and prohibited beginning Jan. 12, 2027, and that standard contractual clauses were published in 2025).

[90] Daniel Schnurr, Switching and Interoperability Between Data Processing Services in the Proposed Data Act, in Data Act: Towards a Balanced EU Regulation (Ctr. on Regul. in Eur. 2023), https://cerre.eu/wp-content/uploads/2023/03/230327_Data-Act-Book.pdf; Leonie Ott & Yifeng Dong, Clouds Connecting Europe: Interoperability in the EU Data Act, JIPITEC (2025), https://www.jipitec.eu/jipitec/article/download/435/434/2202.

[91] Autorité de la concurrence, Opinion 23-A-08, supra note 28; Autoriteit Consument & Markt, supra note 28.

[92] Ünekbas, supra note 32; see also Consolidated Version of the Treaty on European Union art. 5(4), 2012 O.J. (C 326) 13 (requiring proportionality in Union action).

[93] Eur. Comm’n, Commission Launches Market Investigations on Cloud Computing Services Under the Digital Markets Act, supra note 2; see also Eur. Comm’n, Commission Hosted Stakeholder Roundtable on Cloud Computing Services Under the Digital Markets Act (July 1, 2026), https://digital-markets-act.ec.europa.eu/commission-hosted-stakeholder-roundtable-cloud-computing-services-under-digital-markets-act-2026-07-01_en (describing the Article 19 investigation as “distinct and separate” from the designation investigations).

[94] See Alba Ribera Martínez, Generative AI in Check: Gatekeeper Power and Policy Under the DMA, Kluwer Competition L. Blog (Nov. 19, 2024), https://legalblogs.wolterskluwer.com/competition-blog/generative-ai-in-check-gatekeeper-power-and-policy-under-the-dma.

[95] Eur. Comm’n, Report on the First Review of the Digital Markets Act, COM (2026) 178 final (Apr. 28, 2026), https://digital-markets-act.ec.europa.eu/system/files/2026-04/DMA%20Review%20Report_COM_2026_178_1_EN.pdf; see also Megan Kirkwood, What the EU’s First Digital Markets Act Review Actually Changes, Tech Pol’y Press (Apr. 30, 2026), https://www.techpolicy.press/what-the-eus-first-digital-markets-act-review-actually-changes.

[96] See ByteDance, supra note 10; Meta, supra note 69 (upholding Messenger’s designation).

[97] Meta, supra note 69.

[98] Case C-627/24 P, ByteDance Ltd v. Comm’n (appeal lodged Sept. 26, 2024; argued May 12, 2026; judgment pending).

[99] Meta, supra note 69, paras 168, 219.