The European Union has scaled back its artificial intelligence data centre procurement plan. With a core infrastructure plan adjusted not long after it announced a large investment blueprint under the banner of AI sovereignty, criticism is emerging that the gap between Europe’s AI strategy and its ability to execute has again been exposed.

According to Cryptopolitan, a cryptocurrency outlet, on June 18 local time, the European Commission recently revised its AI data centre procurement plan to a smaller form than originally planned.

The commission on June 3 announced a “technology sovereignty package” combining cloud, semiconductors and AI development. It set a goal of investing about 422 billion euros over the next 10 years in data centres, chips, cloud, AI and open-source software.

But in the latest data centre tender, the procurement scale was reduced from the initial concept. The EU had reviewed a plan to deploy 100,000 advanced chips across 5 data centres, but the new plan was adjusted to seek bids to build 7 data centres. Under a Phase 1 standard, 4 centres would each have at least 25,000 graphics processing units, while 3 centres would each have at least 40,000 processors.

The financing structure has also shifted toward the private sector. The EU’s policy is to speed up data centre construction by attracting private investment rather than expanding infrastructure led by the public sector. But assessments say it is uncertain whether Europe can catch up in a short period with AI computing power that already lags the United States and China.

Europe’s share of global AI computing capacity currently stands at about 5 percent. The United States accounts for about 80 percent. U.S. big tech companies invested more than $400 billion in expanding AI infrastructure in 2025 alone. That is larger than the 200 billion euros AI investment plan the EU said it would deploy over several years.

Dependence in the cloud market is also cited as a structural weakness for Europe. Amazon Web Services, Microsoft Azure and Google Cloud account for about 70 percent of Europe’s cloud market. A report released in late May by Allianz said U.S. companies dominate about 80 percent of Europe’s cloud computing market and about 60 percent of enterprise software sales.

The U.S. CLOUD Act also allows U.S. authorities to require U.S. companies to submit data regardless of where it is stored, fuelling controversy over data sovereignty in Europe.

Europe’s sense of crisis has grown further after recent U.S. measures restricting access to AI models. The U.S. government on June 13 limited overseas access to Anthropic’s top-tier AI models Mythos and Fable. In response, European policy authorities and industry said the possibility of U.S. technology blocking has become real. French Prime Minister Sebastien Lecornu (세바스티앙 르코르뉘) said, "We cannot depend on tools made by foreign powers," and added, "France must have its own tools." French telecoms operator Orange also said after the restrictions on access to Anthropic models that access in Europe to "AI that does not capriciously switch off" has become a strategic necessity.

The incident is also adding momentum to the push for a “Cloud and AI Development Act” that aims to triple EU data centre capacity within the next 5 to 7 years. The bill includes applying strong European sovereignty standards to sensitive public-sector work such as in healthcare, finance and the judicial sector.

Still, there is criticism that current procurement standards are not strong enough. Cristina Caffarra (크리스티나 카파라), a competition economist leading the EuroStack industry initiative, called the commission’s measures "very weak." She argued that stronger procurement mechanisms were weakened under U.S. pressure.

Under the currently proposed rules, only about 10 percent of all cloud contracts would have strong European sovereignty standards applied, and the remaining 90 percent would be open to all providers, including U.S. hyperscalers.

Industry also raises a clash between AI self-reliance and the pace of technology adoption. Siemens Chief Executive Roland Busch (롤란트 부시) warned that technology adoption could slow if Europe prioritises building its own infrastructure over using existing AI tools. He criticised the EU’s AI regulatory approach as an "out-of-sync way," likening the spread of AI in the United States to a "fast-flowing river" and Europe to "stagnant water."

Costs are also a variable. Capgemini Chief Operating Officer Karin Brune (카린 브루네) pointed out that European cloud alternatives could be up to 40 percent more expensive than U.S. products. That could become a burden for companies choosing European infrastructure.

Against this backdrop, French AI company Mistral is cited as a representative AI company Europe can put forward. Mistral is reportedly in talks to raise $3.5 billion based on a $23.2 billion valuation. But even that leaves a large gap compared with OpenAI’s valuation, which exceeds $500 billion.

The commission plans to begin a call for proposals in July to build AI gigafactories and to consult member states and the European Investment Bank on financing options.

With the scaling back of the data centre tender and the risks of dependence on foreign AI highlighted at the same time, Europe’s AI sovereignty strategy has now come under test not as a declaration but in actual investment scale, procurement standards and execution speed.