Europe’s AI sovereignty runs into US cloud dependence

DeepL’s April 2026 partnership with Amazon Web Services shows how Europe’s AI ambitions are constrained by cloud concentration. The company had built its reputation on GDPR compliance and processing customer data on its own servers, but enterprise customers require global deployment, low latency and procurement through platforms they already use.

Europe’s AI industrial strategy since 2025 has focused on expanding capacity to train frontier models through facilities such as AI Factories and Gigafactories. The bigger commercial need, however, is inference: serving user requests quickly and reliably across distributed infrastructure. That market favors AWS, Microsoft Azure and Google Cloud, which already control global GPU fleets, customer relationships and managed AI services.

European cloud providers face a structural disadvantage because leading proprietary models from OpenAI, Anthropic and Google are largely distributed through US clouds. Providers that rely on open-weight models often offer smaller, older or compressed versions because frontier models are costly to serve at scale. Even new European inference capacity may depend on US AI firms as anchor customers, reinforcing the same dependence policymakers are trying to reduce.