Yann LeCun, the former chief AI scientist at Meta, has unveiled his new Paris-based startup AMI Labs with one of Europe’s biggest early-stage AI financings, raising €890 million to back a research agenda that breaks sharply from the large language model boom.
AMI, short for Advanced Machine Intelligence, is focused on so-called world models — AI systems designed to learn how the physical world works, retain memory, reason through situations and plan actions. LeCun has argued that text prediction alone will not be enough to reach human-level machine intelligence, and he is positioning the company as an alternative to the prevailing strategy pursued by much of Silicon Valley.
The startup was founded with former Meta executives and Alexandre LeBrun, the former head of health-tech company Nabla. AMI begins with a small international footprint spanning Paris, New York, Montreal and Singapore, but the size of the funding round gives it unusual room to hire researchers and pursue long-horizon work from the outset.
Investors in the round include major technology and financial backers such as Nvidia, Samsung, Toyota, Jeff Bezos, Mark Cuban and Eric Schmidt. The company plans to work first with enterprise and science-heavy sectors including industry, biomedicine and robotics, where LeCun believes AI needs a deeper understanding of real-world processes than current chatbots can offer.
The launch also sharpens Europe’s claim to a more prominent role in advanced AI. By building AMI in Paris and framing world models as the next frontier, LeCun is betting that the next decisive leap in AI may come not from scaling today’s language systems further, but from teaching machines how the world behaves.