Europe’s AI Plan B takes shape as sovereignty concerns grow
Europe is quietly assembling what insiders are calling an AI Plan B — a fallback strategy for the continent’s artificial intelligence ambitions that doesn’t rely on American hyperscalers or Chinese hardware. And with geopolitical tensions reshaping global tech supply chains, the pitch is getting a serious hearing in Brussels.
Why Europe thinks it needs a backup
The core argument is straightforward. Europe’s current AI infrastructure is deeply dependent on U.S. cloud providers like Microsoft, Google, and Amazon. That’s fine in stable times. But policymakers are increasingly nervous about what happens if political relationships sour, or if Washington decides to restrict technology exports the way it has with advanced semiconductors. The EU’s experience scrambling for energy alternatives after 2022 is, for many officials, a cautionary tale they don’t want to repeat in digital infrastructure.
A senior EU official familiar with the discussions put it bluntly: “We cannot build a digital single market on foundations we don’t control.”
What the plan actually involves
The proposal centers on three pillars: sovereign compute capacity, open-source AI model development, and coordinated public procurement. On compute, the EU is looking at expanding its EuroHPC Joint Undertaking, which already operates seven supercomputers across the continent, to include dedicated AI training infrastructure. The target, according to documents circulating among member state delegations, is to have at least 20% of Europe’s AI compute needs met by publicly accessible, EU-governed infrastructure by 2030.
Open-source development is the second pillar. Europe has genuine assets here. Projects like Mistral in France and Aleph Alpha in Germany have produced competitive large language models. Still, they’re outgunned on resources. The plan would channel roughly €1.5 billion toward open-source AI research through existing instruments, including Horizon Europe and the newly proposed AI Factories initiative.
Public procurement is trickier. Getting 27 member states to coordinate purchasing decisions is notoriously difficult.
The skeptics aren’t quiet
Not everyone is sold. Tech industry representatives have pushed back, arguing that Europe risks fragmenting global AI development at exactly the wrong moment. Some economists warn that favoring European providers could raise costs for businesses and slow AI adoption in sectors like healthcare and logistics, where the productivity gains are potentially enormous.
But supporters counter that the plan isn’t about building walls. It’s about having options. The distinction matters. Europe isn’t proposing to ban American AI — it’s proposing to ensure there’s a credible alternative if one is ever needed.
What comes next
The European Commission is expected to formally present elements of the strategy as part of its broader AI Continent Action Plan later this year. Member states will then need to agree on funding commitments, which is where political momentum often goes to die.
The window for action is real but narrow. AI capabilities are consolidating fast, and the infrastructure choices made in the next two to three years will likely define Europe’s position in the technology for a decade. Whether the continent seizes that moment or defers it again is the question now hanging over Brussels.
