Mistral wants open-weight AI to compete at the frontier. It just raised $3.5 billion to do it.

Mistral has just closed a €3 billion Series D, a round the company says lifts its post-money valuation above €21 billion. The money, which works out to $3.5 billion, is earmarked for frontier research, more compute for training, and a bigger infrastructure buildout.

That spending plan is the real story. Mistral is arguing that open-weight models alone don’t change who has power in AI if the chips, servers, and training capacity stay concentrated in a few hands. You can open the model, but if the machines underneath are still scarce and controlled elsewhere, the leverage doesn’t move very far.

The company says it is building the “full stack” needed to answer that problem. In Mistral’s telling, that means the models themselves, the compute they run on, the infrastructure around them, and the products that turn all of that into something enterprises can actually use in production.

And the new funding should help that stack grow. Samsung Electronics led the round, with Scaleup Europe Fund, managed by EQT, and existing investor PSG Equity as co-leads. Mistral says the aim is to free customers from dependence on a single vendor’s roadmap, pricing, and availability, while keeping their data, workflows, and institutional knowledge inside their own walls.

The company has been moving in this direction for a while. It launched three years ago with open-weight models, then last month said it would begin hosting third-party open models too. Arthur Mensch, Mistral’s co-founder and CEO, has also been publicly pushing back on closed-model dependency, warning that data retention rules can give those providers a lot of leverage.

That’s why this round matters beyond the headline number. Mistral isn’t just trying to build a better model; it’s trying to make open AI less dependent on someone else’s infrastructure. That may be the smarter fight.