French AI Startup Mistral AI Raises $830 Mn in Debt for its Sovereign AI Infra

French AI startup Mistral AI has raised $830 million in debt financing to deploy NVIDIA Grace Blackwell infrastructure with 13,800 NVIDIA GB300 GPUs and bring its powered capacity to 44MW, accelerating its push into large-scale sovereign AI infrastructure. 

“We are building 200 MW of capacity across Europe by the end of 2027 to support the demand from governments and enterprises that seek to build and control their own AI,” the startup announced in a LinkedIn post. 

The funds will be used to acquire advanced hardware, including thousands of high-performance chips, for a new facility in Bruyères-le-Châtel near Paris.

The financing was backed by a consortium of seven banks, including BNP Paribas, Crédit Agricole CIB, HSBC, and MUFG, underscoring growing institutional confidence in Europe’s AI ecosystem. 

The move signals a strategic pivot for Mistral, which has so far relied heavily on equity funding. By tapping debt markets, the company is aligning itself with capital-intensive infrastructure players as AI competition increasingly hinges on access to compute.

Media reports also highlighted that the funding reflects rising investor appetite for European AI firms seeking to rival dominant US players such as Microsoft, Google, and Amazon.

Mistral CEO, Arthur Mensch, emphasised the broader vision behind the expansion, noting that scaling infrastructure across Europe is critical to ensuring “AI innovation and autonomy” for the region, according to a statement shared with Reuters. 

The latest fundraise builds on Mistral’s earlier plans to expand its data centre footprint across Europe, including a proposed facility in Sweden and a target of securing 200 megawatts of computing capacity by 2027. 

At the beginning of this month, Mistral introduced ‘Small 4’, the latest addition to its completely open-source ‘Small’ series, designed to enhance the capabilities of its leading models. Moreover, they launched ‘Forge’, a platform that enables companies to create custom models using their own data for training.