Mistral goes 'build‑your‑own' enterprise
Mistral launched a 'build‑your‑own AI' enterprise platform at NVIDIA GTC, positioning itself against OpenAI and Anthropic with customizable, agentic models aimed at large corporate clients. The product push intensifies the enterprise AI platform arms race for on‑prem and hybrid deployments. (techcrunch.com)
Mistral named specific industrial partners for Forge development, citing ASML, DSO National Laboratories Singapore, Ericsson, the European Space Agency, HTX Singapore and Reply as early collaborators to train models on proprietary datasets. (mistral.ai) Forge’s documentation says the platform supports pre‑training on “large volumes of internal documentation, codebases, structured data and operational records,” plus post‑training refinements and reinforcement learning to align agent behavior with internal policies. (mistral.ai) Product messaging and press coverage describe Forge as an end‑to‑end pipeline that covers data curation and tokenization through distributed training on GPU clusters and final deployment to customer environments. (i10x.ai) Mistral emphasizes control and strategic autonomy, stating Forge enables models to be trained and governed inside an organization’s own infrastructure and compliance frameworks to meet regulated‑environment requirements. (mistral.ai) Mistral’s commercial muscle is notable: the company completed a roughly €2 billion funding round that valued it near €12–13.8 billion (~$14 billion) in September 2025, and CEO Arthur Mensch told TechCrunch the firm is on track to exceed $1 billion in annual recurring revenue this year. (bloomberg.com) (techcrunch.com) Reports and vendor materials link Forge to NVIDIA’s stack—Mistral’s Mistral‑3 models are described as optimized for NVIDIA supercomputing and edge platforms—and analysts say Forge’s on‑prem/hybrid focus will increase demand for enterprise NVIDIA GPU capacity for private training. (blogs.nvidia.com) (economictimes.indiatimes.com)
