Most enterprise AI deals are about access. Forge is about ownership.
Mistral AI introduced Forge at NVIDIA’s GTC 2026 conference, describing it as “a system that allows enterprises to build frontier-grade AI models grounded in their proprietary knowledge.” The platform is built for on-premises deployment, the model trains on a company’s own systems and stays there. No data leaves. TechCrunch confirmed the launch independently, framing it as Mistral’s bet that “build-your-own AI” is a viable competitive position against OpenAI and Anthropic.
Mistral says Forge supports the full model training lifecycle, including pre-training and post-training customization. The company is positioning it for industries with strict data privacy requirements, including finance and defense, Mistral’s own characterization, not an independently assessed market fit.
There’s a gap between announcement and adoption. CIO.com noted that analysts say adoption may be limited in the near term, a useful counterweight to the launch narrative worth carrying into any evaluation process.
Mistral CEO Arthur Mensch has projected the company will exceed $1 billion in annual revenue in 2026. Mistral was reportedly valued at €11.7 billion following a funding round in late 2025, though this figure wasn’t confirmed in the sources available for this brief. Treat both figures as executive statements and reported context, not verified financials.
For enterprise AI buyers evaluating Forge: the core question isn’t whether on-premises training is technically possible. It’s whether your organization has the infrastructure, data governance, and ML engineering capacity to run it. The deep-dive for this story addresses that question directly.