Mistral AI is a Paris-based company that builds large language models and sells them as a full-stack AI platform — covering frontier models, a developer API, a cloud compute layer (Mistral Cloud/Compute), and an AI assistant (Le Chat). Enterprises buy access via a pay-per-token API, on-premises deployments, or bespoke "sovereign" contracts where the model runs entirely inside a customer's or government's own infrastructure. The company's edge is European origin and open-weight model releases, which appeal to regulated industries and governments that won't hand sensitive data to US-controlled AI providers. Target buyers include banks, manufacturers, defense agencies, public-sector bodies, and scientific institutions. Beyond API access, Mistral increasingly co-builds custom AI systems with customers — think Airbus for engineering simulation or Tesco for retail operations — embedding engineers directly at client sites. It also sells access to its own GPU data center capacity for inference workloads.