Mistral’s Mensch on agentic AI, chips and adoption

I was in Paris in June 2023 when I heard news of a four-week-old startup called Mistral raising a funding round of over $100 million. No product. No marketing. And not much info about who they were. Over the coming months we learned more about Arthur Mensch, the CEO of Mistral, his team and ambition.

What started as a company developing AI models has evolved into a firm that is looking to position itself as Europe’s answer to the U.S.’s frontier labs OpenAI and Anthropic. Mistral is building out data centers and focusing on enterprise rollout of AI across various industries.

“Europe is starting to be looking at AI as a strategic asset,” Mensch told me on the latest episode of The Tech Download where we had a wide-ranging conversation spanning the future of agentic AI to chips and his views on the state of the AI market.

You’ll have probably heard the word “agent” or phrase “agentic AI” quite a bit lately. The basic idea is an AI system that can handle long and complex tasks on your behalf like an advanced digital assistant. Coding is an area where this has really taken off, with products such as Anthropic’s Claude Code and OpenAI’s Codex gaining traction. Mistral’s big agentic play is called Vibe, which combines its chatbot with its coding tool.

The push into Agentic AI could result in a change in the way organizations are structured. Mensch said that businesses will need to look at processes that can be automated and where the human should remain in the loop. Enterprises should be looking at “how to re-orchestrate all of the people that are involved in that process around an AI system,” Mensch said.

It’s not just about the software for Mistral. The company is also building data centers to power its models and AI. The rationale is that Mistral wants to own more of the “technology stack” from the models to the computing power it runs on.

Those data centers are heavily based on Nvidia chips. But there was a very interesting part of my conversation with Mensch, where he revealed for the first time, that Mistral is exploring designing its own chips. This is a play we’ve seen from big U.S. hyperscalers like Google and Amazon.

Ultimately, Mensch still sees this as an early stage of AI adoption.

“There’s still a lot of viscosity in adoption in enterprises, which means that there’s still … a lot of value creation to be had,” Mensch said.

I had fun recording this episode of The Tech Download in Paris and I hope you enjoy listening.

SpaceX CEO Elon Musk on Thursday spoke to employees of ASMLEurope’s most valuable company, underscoring the importance of chip manufacturing to his two main businesses.

OpenAI on Thursday announced it will acquire Onaa startup that provides secure, pre-configured cloud environments where artificial intelligence agents can access tools, systems and context.

Shares of several large freight carriers slumped Wednesday after Amazon announced it’s opening up its trucking services to companies beyond its own network.

A slew of U.S. Big Tech and AI companies are racing to expand in London as they look to take advantage of the city’s deep talent pools amid a push to develop and commercialize frontier tech.

Anthropic said it’s ready to release a “Mythos-class” AI model to the publictwo months on from limiting the rollout of its powerful Mythos model, citing concerns about the system’s cybersecurity abilities.