The European Commission's first statutory review of the Digital Markets Act, mandated by Article 53(1) DMA and due by 3 May 2026, has now landed. Three years after the regulation's substantive obligations became applicable on 2 May 2023, Brussels is being asked a deceptively simple question: is it working? The honest answer is that it is too early to tell — and that itself is the most important finding the review can produce.
The DMA was sold as ex-ante competition rulebook for a small set of designated gatekeepers: Alphabet, Amazon, Apple, ByteDance, Meta, Microsoft, and Booking. Its premise was that traditional competition enforcement under Article 102 TFEU was too slow to discipline self-preferencing, anti-steering, and tying conduct on core platform services. Articles 5-7 substituted bright-line prohibitions for case-by-case analysis. Three years in, the evidence base is thin, the costs are real, and the temptation to expand scope before the existing framework has been properly measured is the central risk of this review.
What the first enforcement wave actually showed
The Commission's first non-compliance decisions, reported in April 2025, hit Apple with a roughly €500 million penalty over App Store anti-steering restrictions and Meta with a roughly €200 million fine over its 'consent or pay' advertising model. Both are now in litigation, and both expose genuine tensions in the DMA's design.
The Apple case turned on whether developers can freely communicate alternative purchasing options to users outside the App Store — a classic anti-steering question. Apple's response has been to restructure its EU developer terms multiple times, producing a notoriously complex 'Core Technology Fee' regime that developers, regulators, and even Apple's own users struggle to understand. Whatever one thinks of Apple's commercial choices, the spectacle of a flagship law producing a more confusing market — not a cleaner one — should give the review authors pause.
The Meta case raises a deeper structural problem: the DMA's prohibition on combining personal data without GDPR-grade consent collides with the economics of ad-funded services. The 'consent or pay' compromise was the Commission's own preferred remedy in Bundeskartellamt v. Meta (C-252/21, Court of Justice, 2023). When the same Commission later fines a gatekeeper for offering precisely that model, businesses and users are entitled to ask which voice represents the EU's settled position.
Contestability: measurable progress, not transformation
On the positive side, the DMA has produced real openings. Third-party app stores are now technically possible on iOS in the EU. Browser-choice screens have shifted some default share toward independent engines such as Vivaldi and Ecosia, though Google Chrome's position remains broadly intact. WhatsApp has published interoperability APIs for messaging under Article 7, with limited uptake so far. Booking has unwound certain parity clauses, and Amazon has adjusted Buy Box ranking logic.
These are meaningful but incremental wins. They are not, by themselves, evidence of structural transformation in digital market contestability — and the review should resist the urge to claim otherwise.
The expansion question: cloud and generative AI
The most consequential debate in the 2025-2026 stakeholder workshops has been whether to extend the DMA to cloud computing tiers and generative-AI assistants. Both proposals deserve a sceptical hearing.
Cloud markets are concentrated but actively contested. AWS, Microsoft Azure, and Google Cloud each face price competition, sovereign-cloud entrants (OVH, Aruba, IONOS), and the structural pressure of the EU Data Act's switching and portability rules, which become fully applicable in September 2025. Layering DMA obligations on top of the Data Act, the upcoming Cloud and AI Development Act, and sector-specific NIS2 duties risks regulatory accretion without identifying a specific market failure the existing instruments cannot address.
The AI case is even weaker. The generative-AI assistant market is barely 30 months old. OpenAI, Anthropic, Google, Mistral, xAI, Meta, and a long tail of open-weight providers compete on capability, price, and integration. Designating gatekeepers in a market this dynamic — under thresholds designed for mature platforms — would entrench incumbents the moment they crossed a revenue line, and chill the European AI investment the Commission's own competitiveness agenda depends on. The AI Act, fully applicable from August 2026, already imposes systemic-risk obligations on the largest models. Stacking DMA designation on top would be regulatory belt-and-braces with no evidentiary basis.
What a proportionate review looks like
A credible Article 53(1) review should do three things and resist a fourth.
- Measure before expanding. Commission a real ex-post evaluation of contestability and fairness outcomes, with quantitative metrics, before adding new core platform services.
- Simplify compliance. The Apple fee restructuring saga shows that vague obligations produce baroque compliance plans. Clearer guidance under Article 8 specification dialogues would help both sides.
- Resolve the consent-or-pay contradiction. The Commission should align Article 5(2) DMA enforcement with the CJEU's reasoning in Bundeskartellamt v. Meta, rather than fining gatekeepers for following EU case law.
- Avoid scope creep into cloud and AI. These markets are governed by newer, purpose-built instruments. Layering the DMA on top is regulation in search of a problem.
The DMA was never going to remake digital markets in three years. The honest finding of the first review is that it has produced modest openings, real compliance costs, and unresolved tensions with other EU laws. The right response is calibration, not enlargement.