Arthur Mensch, co-founder and chairman of Mistral AI, on November 2, 2023 in Bletchley, central England (POOL / TOBY MELVILLE)
Mistral AI, the French artificial intelligence start-up created in May by industry leaders, announced on Sunday that it had raised 385 million euros, mainly from American groups, becoming one of the two AI champions in Europe.
The start-up had already raised 105 million in June, and among European AI companies, only the German Aleph Alpha has so many funds, after having raised almost 500 million euros at the beginning of November.
“Since the creation of Mistral AI in May, we have followed a clear ambition: to create a European champion with a global vocation in artificial intelligence (…), declared its boss, Arthur Mensch, quoted in the group’s press release.
This second round of financing, led by the Californian fund Andreessen Horowitz, values Mistral AI at 2 billion euros, according to financial sources, which makes it a French unicorn (tech company valued at more than a billion).
The emergence of European AI players comes as the EU has just agreed on Friday on future regulation of the sector, without restricting European innovation. France and Germany feared that excessive regulation would nip their emerging champions in the bud.
Among the backers of Mistral AI are several American tech giants, such as the software publisher Salesforce, but also, according to industry sources, the Nvidia group, a global specialist in chips for supercomputers.
Mobilizing the biggest players in Silicon Valley shows the enthusiasm generated by Mistral AI in just eight months of existence. The American press already cites it as a potential rival to Open AI, the origin of ChatGPT.
Like many of its competitors, Mistral AI, which now has 22 employees, offers companies open source language models powered by public data.
Its main asset is to have been co-founded by three French AI experts, trained at X or ENS, hired by the American giants but who decided to return to Paris.
The CEO, Arthur Mensch, 31, a polytechnician and normalien, spent almost three years at DeepMind, Google’s AI laboratory. His associates come from Meta (Facebook): Guillaume Lample is one of the creators of the LLaMA language model unveiled by Meta in February, and Timothée Lacroix was also a researcher at Meta.