The European Commission wants to treat cloud computing as a gatekeeper market. That is the wrong diagnosis, and it would lead to the wrong cure.
The Commission’s preliminary view that Amazon Web Services (AWS) and Microsoft Azure should be designated as Digital Markets Act (DMA) gatekeepers for cloud-computing services is more than another skirmish in Brussels’ long campaign against large technology companies. It is a test of whether the European Union will use competition policy to protect competitive rivalry or to administer digital markets by regulatory fiat.
The Commission says AWS and Azure are, respectively, the largest and second-largest cloud providers in the EU and that each serves as an “important gateway.” Yet no cloud provider satisfies the DMA’s ordinary quantitative thresholds. A formal decision is expected later in 2026, reportedly in December. The intervening months should not become a procedural pause. They should be used to ask the question the preliminary finding largely assumes away: Do enterprise cloud services fit the gatekeeper model at all?
The answer is no. The Commission’s cloud theory would extend an already defective ex ante regime into a market defined by enterprise procurement, multi-cloud strategies, rapid technological change, falling unit costs, and substantial ongoing investment. It also overlaps with the EU’s own cloud-specific Data Act, which already addresses switching, portability, interoperability, and egress fees. If the Commission’s goal is to make Europe more competitive and innovative, these proposed designations would move policy in precisely the wrong direction.
The Bureaucrat as Market Referee
The DMA was sold as a competition measure, but it is not competition law in the traditional sense. Classic antitrust, at its best, asks whether challenged conduct harms the competitive process, weighs efficiencies, and tailors remedies to proven harms. The DMA instead imposes categorical duties once a company and service receive the gatekeeper label. As I argued recently here at Truth on the Market, the regime replaces case-by-case proof of competitive harm with obligations triggered by administrative classification. That shift matters. It makes regulatory error cheap for the state and costly for markets.
The problem is not that all large platforms are harmless, or that policymakers should ignore exclusionary conduct. It is that ex ante intervention in dynamic markets carries high false-positive costs. Rules that ban or redesign integrated product features can destroy consumer benefits before anyone measures them. Access and interoperability mandates can reduce security, weaken investment incentives, or turn differentiated products into regulated utilities. Rules that force a firm to make rivals’ business models easier can protect competitors rather than competition. In digital markets, where value often comes from scale, integration, learning, reliability, and rapid iteration, these costs are not incidental. They are the main event.
Early DMA implementation already counsels caution. Some studies and market evidence report consumer frustration and business harm from compliance-driven product changes. The Nextrade Group survey of EU consumers and the European Centre for International Political Economy (ECIPE) study on consumer experience emphasize increased friction and limited visible benefits. In travel search, Mirai reported sharp declines in Google Hotel Ads clicks and bookings in DMA-affected markets, with intermediaries gaining visibility at the expense of direct hotel channels. These findings remain contested and do not settle the full welfare question. But they do puncture the tidy assumption that DMA compliance always ratchets toward consumer benefit.
For that reason, the EU should not expand the DMA into cloud unless the economic case is especially strong. The Commission’s preliminary case is not. It relies on size, investment capacity, AI partnerships, alleged lock-in, and cloud’s importance to Europe’s digital economy. Those facts may show commercial significance. They do not establish gatekeeping in the relevant economic sense.
Where’s the Gate?
The strongest objection to the cloud designation is conceptual. The DMA’s core platform-service framework was built for services that sit between businesses and consumers: search engines, app stores, social networks, operating systems, advertising services, and other digital chokepoints that can influence which firms reach end users. Cloud computing is different. It is business-to-business infrastructure. Cloud providers supply computing power, storage, databases, networking, security, and software tools. Their enterprise customers then use those resources to build products, applications, and internal systems. The cloud provider generally does not determine whether those customers reach end users, how users find them, or what prices they charge.
That is why International Center for Law & Economics (ICLE) Director of Competition Policy Dirk Auer argues that applying the DMA to AWS and Azure “stretches the act into a market it was never built to reach.” In its June 2026 assessment, ICLE noted that no cloud provider meets the DMA’s quantitative thresholds, that AWS and Azure hold estimated EU market shares of roughly 28% and 21%, respectively, and that Google Cloud accounts for about 14%. It also found that roughly 70% of customers use more than one cloud provider. Those are not the hallmarks of a tipped—or tipping—market. They are the hallmarks of vigorous competition among differentiated suppliers.
Lazar Radic makes the same point more fundamentally in his commentary “Cloudy Logic”: cloud has “no gate” in the relevant DMA sense. Enterprise customers routinely use multiple providers, negotiate contracts, benchmark performance, repatriate workloads, and combine hyperscale services with on-premises systems, European providers, telecommunications companies, specialized AI clouds, and open-source tools. A large bank, manufacturer, airline, or startup that relies on AWS for some workloads and Azure or Google Cloud for others bears little resemblance to a consumer locked into a single app store or search engine.
The Commission appears to treat switching costs as presumptive evidence of anticompetitive conduct. That is a mistake. In complex infrastructure markets, switching costs often reflect productive integration, customization, security choices, regulatory compliance, data architecture, employee training, and customer-specific investments. A sophisticated customer may deepen its relationship with one provider because it values reliability, speed, technical support, or specialized features. The fact that such a relationship is expensive to unwind does not prove exclusion. It may simply show that cloud services are differentiated products rather than interchangeable commodities.
Two Rulebooks, One Market
The overlap with the Data Act further weakens the Commission’s case. The EU’s own explanation of the law says Chapter VI is designed to make switching between data-processing services, including cloud and edge-computing services, “free, fast and fluid.” It requires providers to remove barriers to switching and multi-cloud use through measures such as open interfaces, machine-readable data exports, and functional-equivalence efforts for infrastructure services. It also abolishes switching charges, including data-egress fees, beginning Jan. 12, 2027.
That targeted framework already addresses the concerns driving the Commission’s cloud investigation: interoperability, portability, switching costs, and data access. Meanwhile, traditional EU competition law, particularly Article 102 of the Treaty on the Functioning of the European Union (TFEU), remains available to address specific exclusionary conduct by dominant firms. The added value of DMA designation is therefore difficult to see. The added costs are not: duplicative obligations, conflicting interpretations, compliance uncertainty, and litigation over which regime governs.
Regulatory overlap is not costless. Every new layer changes investment incentives. It encourages risk-averse compliance teams to slow product launches, narrow product offerings, or avoid features that could later be recast as self-preferencing, tying, or discriminatory access. It also rewards firms that excel at regulatory arbitrage rather than technical performance. Europe should want cloud customers to choose providers based on price, quality, latency, security, sustainability, and innovation—not on the unpredictable consequences of overlapping regulatory mandates.
Reading Draghi Backward
The proposed cloud designations are especially ill-timed in light of Mario Draghi’s report on EU competitiveness. Its central message is straightforward: Europe faces slowing productivity, intensifying global competition, and a widening innovation gap. The report argues that Europe has struggled to turn its scientific strengths into globally competitive technology companies and that fragmented and overly restrictive regulation has held back growth. The Commission has embraced Draghi’s agenda rhetorically. Expanding the DMA to cloud would undermine it in practice.
Cloud infrastructure is not a luxury for Europe’s innovation economy. It is the foundation on which businesses build artificial intelligence, cybersecurity, data analytics, biotechnology, advanced manufacturing, financial services, logistics, media, public-sector modernization, and startups. AWS and Azure are among the companies investing in that foundation. A regulatory regime that makes large-scale cloud investment less attractive, delays access to frontier AI services, or turns Europe into a compliance outlier will not create European champions. It will leave European businesses later to adopt new technologies, less productive, and less competitive.
A recent Amazon background paper makes the same point. It argues that European cloud customers enjoy broad choice, prices have fallen, and providers continue to invest billions in European data centers and related infrastructure. One need not accept every company claim at face value to recognize the underlying economics. Capital-intensive infrastructure markets depend on predictable returns. If Brussels signals that successful investment will trigger open-ended regulatory redesign, expected returns fall. Marginal projects get delayed, scaled back, or moved elsewhere. Compliance costs rise. Engineering talent shifts from building better products to building defensible ones. That is not a strategy for closing the technology gap with the United States or China.
The problem is not simply more regulation. It is regulation that treats scale as a liability. Draghi’s diagnosis points toward removing barriers to growth, investment, commercialization, and technological diffusion. Expanding the DMA points in the opposite direction. Europe cannot lament its shortage of globally competitive technology firms while treating scale, integration, and investment capacity as evidence of regulatory wrongdoing.
The Passport Test
A further concern is the DMA’s uneven application. The statute is formally nationality-neutral, but in practice, it falls overwhelmingly on large non-EU firms, particularly U.S. technology companies. The proposed cloud designations reinforce that perception. AWS and Azure are U.S.-based providers. Google Cloud, another U.S.-based hyperscaler, is the next-largest competitor. European providers, including OVHcloud, IONOS, Scaleway, telecommunications companies, and sovereign-cloud initiatives, remain outside comparable obligations.
There is no sound economic reason to treat a company’s nationality as a proxy for market failure. A U.S. cloud provider can benefit European customers, startups, public agencies, and exporters. A European intermediary can raise prices, reduce innovation, or lobby for protection. Competition policy should ask whether conduct harms the competitive process and consumers—not whether a successful company happens to be foreign.
The asymmetry also creates a political-economy problem. When regulation imposes costly obligations on a handful of foreign firms while redistributing traffic, access, data, or bargaining leverage to rivals and intermediaries, those beneficiaries gain a strong incentive to preserve and expand the regime. That is not competition policy. It is industrial policy dressed up in antitrust language.
Behind-the-Border Barriers, Brussels Edition
This is where Shanker Singham’s framework of anticompetitive market distortions (ACMDs), developed with colleagues including me, becomes useful. ACMDs are government-created or government-enabled measures that distort competition by raising rivals’ costs, restricting market entry, weakening property rights, or shifting economic rents to favored firms. They are “behind-the-border” barriers: not tariffs, but domestic rules that skew competition and reduce consumer welfare.
Viewed through that lens, the DMA—especially as applied to cloud—looks less like competition policy than an ACMD. It does more than prohibit proven exclusionary conduct. It imposes asymmetric obligations on designated firms because they are large and strategically important. It can require changes to product architecture, interoperability, data access, contracting, and business models. It shifts value from integrated platforms and their customers to rivals, intermediaries, and complainants who did not necessarily succeed through market competition. It also creates regulatory uncertainty that favors firms with the largest legal and compliance teams. In a cloud market that is already competitive and subject to sector-specific regulation, the DMA is more likely to distort competition than restore it.
The ACMD framework also exposes the limits of the Commission’s appeal to “fairness.” Fairness has no administrable limiting principle unless it is tied to consumer welfare, output, innovation, or a recognizable theory of competitive harm. A rule that makes life easier for competitors may make life worse for customers if it reduces quality, raises costs, delays new features, or discourages investment. A rule that fragments services may disadvantage users who value seamless integration. A rule that shifts rents from productive infrastructure providers to politically influential business users may harm the economy as a whole. Competition policy should not become a mechanism for politically allocating digital rents.
Calling It a Gatekeeper Doesn’t Make It One
The Commission should also pay close attention to the courts. In 2026, the General Court upheld parts of the Commission’s DMA designation of Meta but annulled the Facebook Marketplace designation because the Commission had not adequately justified its analysis. As I observed at the time, the ruling was narrow but significant. It underscored that labels, statutory categories, evidence, and reasoned decision-making still matter, even under an ex ante regulatory regime.
That lesson applies with even greater force to cloud. The Commission is not simply applying clear statutory thresholds to an obvious consumer platform. It is relying on a qualitative assessment to designate services that do not meet the DMA’s quantitative thresholds and that differ fundamentally from the services the statute was designed to regulate. If the courts insist on service-by-service analysis, current market evidence, and a genuine fit between legal categories and economic reality, the proposed cloud designations will face difficult questions.
The Commission could respond by issuing a longer decision. More pages would not fix the underlying problem. The real burden is not to produce a more elaborate explanation for forcing cloud into an ill-fitting category. It is to show that the designation will improve competition and consumer welfare after accounting for innovation costs, regulatory overlap, existing legal remedies, multi-cloud adoption, and the risk of deterring investment. On the current record, the Commission has not made that case.
Don’t Build the Gate
The Commission should change course before its final decision, expected in December 2026. It should not designate AWS as a DMA gatekeeper for cloud services. Nor should it designate Microsoft Azure. If a cloud provider engages in conduct that harms competition, the EU already has tools to address it: Article 102 TFEU, contract law, sector-specific rules, and the Data Act’s cloud-switching provisions. Those tools allow for a more tailored inquiry into conduct, effects, efficiencies, and remedies.
DMA designation would instead bring a blunt, overlapping, and innovation-hostile regime to a market that does not fit the gatekeeper model. It would compound the very weaknesses the Draghi report identifies: regulatory fragmentation, insufficient scale, weak investment incentives, and delayed adoption of frontier technologies. It would discriminate in practice against successful non-EU firms without establishing a sound economic reason to do so. And it would operate as an anticompetitive market distortion by burdening productive infrastructure providers and reallocating rents through regulation rather than competition.
Europe’s problem is not too much permissionless innovation from cloud providers. It is that too many promising technologies fail to scale, too much capital flows into compliance rather than growth, and too many regulatory initiatives treat market success as a public-policy defect. A stronger EU economy will not come from designating more gatekeepers. It will come from letting competition discover better products, lower prices, and more resilient infrastructure.
The Commission should withdraw its preliminary view, rely on the Data Act and ordinary competition law for specific cloud concerns, and resist turning the DMA into yet another anticompetitive market distortion. Cloud has no gate. Brussels should stop looking for one.
