By 2030, the EU stands to unlock 1.2 trillion euros in annual GDP thanks to cloud computing and AI.
That assessment comes from the ECIPE (European Centre for International Political Economy). The neoliberal think tank concedes that this is an ambitious scenario. It implies that, across all economic sectors*, every member state would reach the adoption level of the most advanced country today.
When they say “AI,” they mean “cloud AI.”
The ECIPE generally argues that AI adoption depends on cloud adoption. It illustrates this with OECD data.
France singled out for its “underperformance”
The analysis also draws on MIT’s Global Cloud Ecosystem Index 2022 and Copenhagen Economics. The former shows cloud adoption is lower in the EU than in the United States, in Australia, and even in emerging markets like Colombia. The latter reflects uneven adoption within the public sector.
In this sector, Denmark and Estonia emerge as the most advanced. In the private sector, Finland leads the pack. Greece, Romania, and Bulgaria sit at the opposite end of the spectrum. As for France and Germany, they “underperform relative to their economic weight,” the ECIPE asserts, reinforced by Eurostat data (December 2023).
From the Data Act to the future CAIDA, regulation under the grill
The “ambitious” scenario would therefore have the EU generating an additional €1.2 trillion of GDP per year by 2030. The most pronounced gains would be observed in Romania (+32.4%), Bulgaria (+22.8%), and Poland (20.8%).
Conversely, growth would be limited to 1.6% in Denmark, 1.8% in Finland and 2% in Sweden. The ECIPE argues that these rates could be higher, but it has chosen “conservative” estimates for the more advanced countries.
Beyond the assessment, the think tank offers avenues focusing on three audiences: regulators, competition authorities, and standardization bodies. Its approach favors maximum consumer choice. It thus resonates—with and, in some sense, counters—a study recently highlighted by Cigref. The main conclusion: a non-negligible dependence on the United States for digital goods. In the U.S. there is about €260 billion of value generated each year through European companies’ purchases of American cloud services and software.
On the regulation front, the ECIPE weighs in on several texts:
- Data Act
- DMA (Digital Markets Act)
- DSA (Digital Services Act)
- DNA (Digital Networks Act; in draft)
- CAIDA (Cloud and AI Development Act; also in draft)
Regarding the Data Act
In the ECIPE’s view, there is a risk of over-regulating PaaS and SaaS, subjecting them to the same obligations as IaaS, even though switching providers is comparatively easier.
Another concern: the inappropriate application of switching-provider rules to services that are complementary rather than substitutable. The lack of clarity in the functional criteria by which the text defines services would feed this risk.
There is also a lack of precision on the unbundling rules. As it stands, the ECIPE believes they tend to apply to tightly integrated technical architectures and bespoke configurations, even though unbundling would be impractical in practice.
Regarding the DMA
What exactly constitutes an “undertaking using [the service]” is one of theDMA’s thresholds for application… and the ECIPE considers this definition insufficiently precise for cloud services. It notes that most customers use these services for internal needs, not to deliver consumer-facing products. Consequently, there is a risk that providers are subjected to the regulation despite the real market dynamics.
Another point flagged: the implicit assumption that the number of users equates to market power. The ECIPE argues this does not account for market contestability: despite concentration on the infrastructure side, the cost of changing providers can be mitigated by multicloud and containerization. Meanwhile, new entrants keep positioning themselves—often successfully—in niche or high-performance segments.
Regarding the DSA
A further broad definition here is that of the host platform. The same logic as with the DMA applies: it does not distinguish consumer-facing online platforms from enterprise cloud infrastructure. Obligations imposed on the former could apply to the latter, even when they merely serve internal needs.
Regarding the DNA
The concern here is primarily the prospect of cloud infrastructure services being classified on the same footing as electronic communications services. Risk arises of applying “disproportionate” requirements—for example, guarantees of availability and interoperability.
Regarding the CAIDA
While the ECIPE applauds the ambition to triple Europe’s available computing power within 5-7 years, it fears that this could be read as a directive to exclude non-European providers—or at least to impose restrictive localization requirements.
The notion of sovereign infrastructure does not need to clash with the idea of an open market for general-purpose computing services. The think tank reiterates the call not to treat cloud as on par with telecoms and not to be overly rigid about interoperability. Gaia-X, it argues, has shown how difficult it is to impose common architectures in a top-down manner.
Also to consult in addition:
Cloud sovereignty: CAIDA, a new battleground
Digital Networks Act: a few figures to contextualize this initiative
DMA: Apple and Meta fined €500m and €200m
How the CISPE proposes to implement the Data Act
* More precisely, the 12 sectors listed by the latest revision of the NACE (Statistical Classification of Economic Activities).
