This is a news item from our media watch, potentially of interest to part of the university community. The content relayed does not reflect a position of UNIL.
On 11 August 2026, Mistral AI announced an expansion of its infrastructure business: regional endpoints that let customers choose whether inference runs in Europe or the United States, and multi-year compute commitments signed by large European companies. The company currently operates less than 200 megawatts and is targeting 1 gigawatt in Europe by the end of 2030. Its documentation mentions limited, safeguarded transfers outside the selected region for certain features such as web search.
The same announcement opens its platform to open models built by third parties, starting with GLM-5.2 from the Chinese lab Z.ai. Choosing a European provider therefore no longer determines on its own where the model comes from.
https://mistral.ai/news/regional-inference-open-models-new-compute
In an opinion column, Pierre-Louis Biojout, founder of the AI company NanoCorp, reads this as an abdication by Europe’s AI champion and describes Mistral as an installer of Chinese models. As of 19 August 2026, however, Mistral’s catalogue listed a single third-party model, the company had released a model of its own a week before the announcement, and Amazon Bedrock, Google Vertex AI and Microsoft Foundry already distribute Chinese open models. Its largest model, on the other hand, remains Mistral Large 3, released in December 2025.