Three US companies control 70% of European cloud infrastructure. Mistral AI, Europe's most prominent homegrown AI lab, holds just 2% of the global large language model market. Only 1% of global chip design happens inside the EU.
Those numbers are the backdrop for a growing push among European policymakers to treat AI and tech independence as a security priority, not just an economic one. A new policy brief from the European Council on Foreign Relations lays out a "EuroStack" framework - an interconnected plan spanning four layers: space, chips, cloud computing, and AI.
The trigger is straightforward. The Trump administration has already threatened to weaponize digital dependencies, including restricting AI chip exports and, in the case of Ukraine, leveraging Starlink access during mineral negotiations. The brief warns Europe could face similar service shutdowns or import restrictions at any time.
The Numbers Behind the Plan
The proposals come with specific price tags. Space funding would jump from €12 billion to €60 billion annually. The private AI sector has pledged €150 billion through 2030. The EU's InvestAI initiative adds another €50 billion. For chips, a proposed "Chips Act 2.0" would focus on mainstream 65-90 nanometer chips for automotive and industrial use, rather than chasing bleeding-edge fabrication.
On cloud, the brief recommends designating AWS, Azure, and Google Cloud as "gatekeepers" under the Digital Markets Act, banning anti-competitive practices like egress fees (the charges companies pay to move their data off a cloud provider), and introducing "Buy European" procurement quotas for public bodies.
Open Source as Strategy
The AI proposals lean heavily on open-source models as a distinct European advantage. The logic: if you cannot outspend American labs on proprietary model training, you compete by funding open alternatives that lower the barrier for smaller European companies to build on top of. A major share of InvestAI funds would go to open-source AI development, with a medium-term goal of Europe controlling 5% of global AI compute (processing power available for training and running AI models), up from its current negligible share.
The brief is realistic about tradeoffs. It does not propose cutting off US tech entirely. Instead, it recommends defining sensitive domains like defense and public administration where European-only providers are required, while accepting continued dependencies in lower-risk sectors.
For AI tool users in Europe, the practical implications are still distant. But the policy direction is clear: expect more procurement restrictions, data residency requirements, and funding for European alternatives over the next 2-5 years. If you run a business that depends on US-hosted AI services, the regulatory ground is shifting under you.