Finland's Google Halt Exposes Contradictions At The Heart Of Europe's AI Strategy

On September 9, 2026, Google announced it was planning to invest at least €13 billion ($15 billion) in AI infrastructure in Finland to fund three new data centres and expand an existing site, its largest single investment in Europe. On paper, it was a deal with no losers. Google secures clean power for its data centres from the Loviisa nuclear plant under a 22-year agreement. Finland keeps a plant alive that supplies around 10% of its electricity. And Brussels gets the build-out its AI strategy depends on – albeit built by an American organization.

Less than a month later, the project is in trouble. On October 6, the Finnish Supervisory Agency (LVV) halted work at the Muhos and Kajaani sites until mandatory environmental impact assessments are completed, after more than 300 hectares of forest had been cleared without one. It is a local dispute, but it exposes a much larger contradiction in Europe's AI strategy.

Europe wants to build like the US and China, but is bound by its own rules

In September's State of the Union Address, President of the European Commission Ursula von der Leyen called AI one of the tipping points of our time. Her Commission wants to move at the same speed as the US and China, tripling EU data centre capacity within five to seven years and fast-tracking sites in ‘acceleration zones’. Yet, the same speech praised the AI Act's guardrails and insisted AI should be done "in a European way". Finland exposes the tension of this position, as the assessments that stopped Google are required under EU law.

US infrastructure on European soil is not AI sovereignty

The halt also exposes a deeper problem for Europe's sovereignty agenda. Ursula von der Leyen declared that "the fate of Europe must remain in the hands of Europeans", yet Europe is not so much building sovereign AI as leasing its land and power to US firms. Keeping data on European soil does not resolve this issue. Sovereignty also depends on who controls the digital architecture, including data stores, ETL pipelines, systems of record and orchestration layers. Consider the UK, where MPs have urged the government to drop Palantir's NHS contract over fears that US law could force the firm to hand over patient data – even though that information resides in Britain. France is removing Microsoft's Windows operating system from government computers over similar concerns.

Enterprises adopting AI need to plan around Europe's contradictions

For enterprises, the implications are significant. In the Verdantix 2026 global corporate survey, data sovereignty was the most cited AI governance concern, rated most significant by 28% of respondents. Yet sovereignty is still too often treated as a question of where data are stored. If governments are already dropping US providers over who controls the software, regulation may follow – and a locally hosted model on American infrastructure, or a European tool built on an American platform, would offer little protection. Vendors, enterprises and investors should treat jurisdiction as a risk at every layer of the stack – from infrastructure to model to application – map their dependence on any single provider, and build exit options and permitting delays into contracts and capacity plans. While the EU may resolve its contradictions at some point, many firms cannot afford to wait.

To stay up-to-date on AI regulatory and governance news and its impact on enterprise AI adoption, explore Verdantix AI Applied research or book an inquiry with our analyst team.