Governments Invest $100B In Sovereign AI
$100B sovereign AI boom
This spending wave is creating a new buyer in AI, the nation state, and that shifts competition from selling chatbots to selling sovereign stacks. Winning vendors are not just offering a model API. They are offering local data centers, on premises deployments, implementation support, and political comfort that sensitive government data can stay inside national borders while still using frontier grade models.
-
Europe has moved furthest from rhetoric to procurement. Mistral reached $400M ARR by February 2026, selling open weight models, private deployments, and support for industrials and governments. Cohere, at $240M ARR in 2025, also pushed into Europe, showing this budget is already turning into software revenue for non US labs.
-
The money is not going into one template. The current map has at least three models, Europe backing local labs and compute, Gulf states funding regional infrastructure hubs, and countries like India and Indonesia building national systems on top of US cloud and model suppliers. That means sovereign AI often still routes dollars to AWS, Microsoft, Oracle, and Nvidia.
-
The practical product is usually not a public chatbot. It is an air gapped or private deployment that sits inside a ministry or state contractor, runs on approved hardware, connects to local documents and workflows, and comes with hands on integration work. That setup favors vendors like Mistral and Cohere over labs optimized mainly for consumer distribution.
The next phase is a shakeout where sovereign AI budgets concentrate on a small set of vendors that can match frontier performance while meeting local control requirements. The winners are likely to become cross border suppliers for allied countries, while weaker national champions slide into consulting, integration, or niche government contracts.