The promise of sovereign AI is that a country can keep its models, data, and compute power at home. But there’s one critical component that has to come from outside: the chips.
“Today, there is no equivalent in Europe,” said Timothée Lacroix, cofounder and chief technology officer of French AI company Mistral AI, referring to the GPUs, CPUs, TPUs, and other types of advanced chips that are for the most part made by U.S. companies.
“I’m sure there are many alternatives that are going to be built, and we’ll help them out if that’s the case, but today it’s really about getting control where we can,” Lacroix said at the Fortune Brainstorm Tech conference in Aspen on Monday.
Mistral has spent the past year extending beyond models and into infrastructure, with 50 megawatts of its own compute due this summer. The sovereign rationale is control at every layer of the technology.
“To me, it’s really about giving our customers the choice on all layers of the stack for where they want to run what part of their workflow,” Lacroix said. That includes private, agent-driven work on a company’s own premises, and heavier inference hosted inside a customer’s own country “for strategic autonomy.”
But silicon remains a missing piece of the sovereign AI puzzle.
Lacroix said it’s a bit too early for three-year-old Mistral to start thinking about designing its own Europe-based chip. However, “if someone designs a great chip for transformers,” Mistral would likely use it.
“There are European companies that are up and coming, and in that case, we like to talk to them and help them out,” said Lacroix. “It’s always a longer game.”
More from the 25th annual Fortune Brainstorm Tech conference:
Anthropic’s Boris Cherny, creator of Claude Code, says there are days he manages tens of thousands of AI agents at once
Anduril CEO Brian Schimpf says economic warfare is the ‘new normal’ for military conflicts—and the U.S. needs to get serious
Twitch CEO: Social media has become ‘antisocial’ and can’t match the shared, human connection of livestreaming
Daniela Braga, CEO and founder of data marketplace Defined.ai, who spoke alongside Lacroix on the panel, said that political instability is pushing nations to rely on “their own sovereign models, their own sovereign data.”
Braga said Europe and the Middle East serve as necessary counterweights so that frontier AI isn’t controlled by the U.S. and China alone.
The topic came up during several panels at Brainstorm Tech. “Data sovereignty is real, and it’s here to stay,” Snowflake CEO Sridhar Ramaswamy said during a separate panel.
For Mistral, being viewed as centrist is a benefit.
“It’s not really about not being American,” Lacroix said. “The goal is to be a global company.” Coming from Europe, he said, the firm is “seen as more neutral,” which can appeal to customers who “want to be strategically autonomous.”
As for whether the Trump administration will take equity stakes in U.S. AI companies and what that might look like, Lacroix demurred.
“I honestly don’t know what the future looks like,” said Lacroix. But he added, “Who owns that power is very important.”
