Hyperscalers Capture Sovereign AI Spending
$100B sovereign AI boom
Sovereign AI spending is not bypassing US cloud giants, it is being routed through localized versions of their stack. Governments still need the same compute, storage, security controls, and model tooling, so AWS, Microsoft, and Oracle are repackaging their platforms with EU residency, local operators, and disconnected deployment options. That keeps the software layer, service roadmap, and much of the operations model in American hands even when procurement is framed as digital independence.
-
AWSs European Sovereign Cloud went live in January 2026 as a physically and logically separate EU cloud with EU resident operations, EU metadata residency, and independent governance. In practice, that means a ministry can keep workloads inside EU borders without switching away from AWS primitives like IAM, EC2, and managed networking.
-
Microsoft is attacking the hardest government use case, fully disconnected environments. Azure Local disconnected operations and Foundry Local let agencies run infrastructure and large models on their own hardware inside an air gapped perimeter, so the sovereign option is still built around Microsoft software, update paths, and developer interfaces.
-
Oracle is selling a similar compromise, EU Sovereign Cloud for customers that want Oracle cloud services with EU hosting and operations. The broader pattern is that sovereignty demand is creating a premium cloud segment for hyperscalers, while also pulling more sovereign projects onto Nvidia based infrastructure and standard US cloud tooling.
The next phase is a split market. Hyperscalers will keep winning the infrastructure layer because they can localize faster than countries can rebuild the cloud stack from scratch. The biggest openings for local champions are higher up the stack, where companies like Mistral can pair competitive models with on premise deployment and implementation work that feels genuinely national rather than merely regionally hosted.