Mistral launches European sovereign AI compute coalition, first hosts Z.ai GLM-5.2
On Aug 11 Mistral announced regional inference endpoints, Priority Tier and European Compute Units targeting 200MW by 2027 and 1GW by 2030; first third-party hosted model is Z.ai GLM-5.2.
On August 11, French AI company Mistral announced regional inference endpoints, third-party open-model hosting, and a European sovereign compute coalition targeting 1GW of European compute capacity by 2030. The coalition launches the European Compute Units (ECU) long-term compute commitment mechanism and regional priority-tier services, with expansion to third-party open models starting with Z.ai GLM-5.2.
Mistral currently operates under 200MW: a 44MW site near Paris (online Q2), a 23MW site in Sweden with EcoDataCenter, and a 10MW site in Les Ulis (online Q3). The first cohort of committed enterprises includes ASML, CMA CGM, Amadeus, and Caisse des Dépôts. The targets are 200MW of European infrastructure by end of 2027, scaling to 1GW by 2030.
Epoch AI estimates typical 1GW AI data center upfront capex at roughly $38 billion; Goldman Sachs estimates $15-20 million per megawatt before chips. Mistral CTO Timothée Lacroix stated that 1GW is a large investment requiring significant scaling and revenue behind it, framing the urgency as a supply crunch around 2027-28 especially in Europe.
Mistral simultaneously announced that customers can now choose whether AI inference workloads run in Europe or the US to meet cross-border data compliance requirements; the Mistral Priority Tier public preview provides committed service levels for mission-critical workloads. The first hosted third-party model, GLM-5.2, will run inference through Mistral's European data centers, ensuring data does not leave European jurisdiction. The move is seen as a key European strategic response to AI compute dominance by the US and China.