Mistral AI is asking Europe’s largest companies to sign five-year compute contracts with no exit clause, and it is using those pledges to fund a buildout it says will reach one gigawatt of capacity by 2030.
The French lab announced a three-part infrastructure push: regional inference endpoints that let customers pin workloads to Europe or the US, a Priority Tier with uptime guarantees for mission-critical deployments, and an anchor group of enterprises whose long-term commitments convert into European Compute Units, a claim on Mistral-built capacity.
The math is daunting. Mistral currently operates under 200 megawatts, and independent estimates put the cost of a single gigawatt of AI data center capacity near $38 billion. Chief technology officer Timothée Lacroix told VentureBeat that commitments of around five years are the point, and there is no early exit.
Anchor customers already include Amadeus and ASML, whose chief executive led Mistral’s $13.4 billion Series C last year. Microsoft, which struck a multibillion-dollar capacity deal with Mistral in July, effectively becomes an anchor tenant that de-risks the buildout.
In a twist for sovereignty purists, Mistral will also host third-party open models, starting with GLM-5.2 from Chinese lab Z.ai, under the same regional controls and service levels as its own models. The pitch: any open model, consumed safely in Europe.
Mistral has raised roughly $4 billion to date, a fraction of what American rivals command, which is why it is turning customers into financiers. Whether enterprises accept five-year, no-exit commitments will decide if the strategy works.
