Yann LeCun exits Meta and raises $1.03B to build world models at AMI
The Turing Award winner left Meta four months ago convinced that large language models are a dead end. Today he announced $1.03 billion in seed funding, Europe’s largest ever, to build something different.
In November 2025, LeCun walked into Mark Zuckerberg’s office and told his boss he was leaving after twelve years building Meta’s AI research operation. He left as one of the loudest critics of the direction much of the industry has taken. That break set the stage for a bold, well-funded new lab.
LeCun has argued that large language models are a statistical trick: impressive in output but not genuinely intelligent. He believes the dominant approach is increasingly resistant to major new breakthroughs. Investors just bet that his alternative is worth backing with deep pockets.
The new venture is Advanced Machine Intelligence Labs, known as AMI, and it announced its seed round on 10 March 2026. The round values the company at $3.5 billion pre-money and raised €890 million, roughly $1.03 billion, making it one of the largest seed rounds in Europe. The name AMI is pronounced like the French word for friend.
Five firms co-led the round: Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions. Other strategic backers included Nvidia, Toyota, Samsung, and Temasek, alongside French VC Daphni and South Korean investor SBVA. A long list of individual investors also participated, including Tim and Rosemary Berners-Lee, Jim Breyer, Mark Cuban, and Eric Schmidt.
LeCun had initially sought around €500 million, according to reports, and demand far outstripped that target. He closed with €890 million and told journalists interest allowed AMI to be selective about investors. That selectivity and the size of the round underline how much faith the backers placed in the vision.
AMI will be headquartered in Paris, with planned offices in New York, Montreal, and Singapore. LeCun, who holds dual French-American citizenship and remains a professor of computer science at New York University, will serve as executive chairman. Day-to-day leadership goes to Alexandre LeBrun, the founder of the medical AI startup Nabla, who is now AMI’s CEO.
The founding team comes mostly from Meta’s AI research organization, bringing deep domain expertise. Michael Rabbat joins as vice president of world models, Laurent Solly becomes chief operating officer, and Pascale Fung takes the role of chief research and innovation officer. Saining Xie, who previously worked at Google DeepMind, signs on as chief science officer.
AMI’s core mission is to develop world models, a class of systems LeCun has promoted for years as an alternative to language-prediction architectures. World models try to capture how the world behaves at an abstract level, enabling systems to reason about objects and interactions rather than just sequence text. If an AI can visualize an object or an interaction, LeCun argues the right question to ask is, “What does it mean?” and that answering it yields correct responses.
LeCun advocates JEPA, the Joint Embedding Predictive Architecture, which he introduced in 2022 as a framework that learns abstract representations and ignores unpredictable surface detail. JEPA aims to avoid the pixel-perfect or word-by-word forecasting that makes generative systems both impressive and prone to hallucination. The result, he hopes, will be systems that understand physical reality more like humans and animals, through embodied experience.
On timing, LeCun told AFP AMI will start talks with corporate partners within one to two years, and within three to five years aims to produce “fairly universal intelligent systems” deployable across many domains. He said he wants AMI to become “the main provider of intelligent systems.” These are ambitious targets for a company with no product yet.
LeCun has framed AMI as a European counterweight to the major American and Chinese labs, noting “We are one of the few frontier AI labs that are neither Chinese nor American,” and locating the headquarters in Paris underscores that stance. The involvement of French investors supports the positioning. Whether the positioning helps with talent, partnerships, and regulation remains to be seen.
Right now AMI plans to spend its first year focused solely on research and development, with no immediate revenue in view. World models are presented as a long-term scientific endeavor rather than a quick product play. The core question for investors and the field is whether LeCun’s diagnosis of the problem will translate into a workable solution.