The Turing Award winner left Meta four months ago convinced that great language models are a dead end. Today it announced $1.03 billion in seed funding, the largest in European history, to build something different.
In November 2025, Yann LeCun walked into Mark Zuckerberg’s office and told his boss he was leaving. He had spent twelve years building Meta’s AI research operation into one of the most respected in the world, and had become one of the staunchest critics of the technology industry that dominated it.
He argued that large language models were a statistical illusion. Impressive, yes. Smart, no. He thought he could build something better and he thought he could do it faster outside of Meta than inside. On Tuesday, investors put $1.03 billion behind that conviction.
Advanced artificial intelligence laboratoriesAMI, which is pronounced like the French word for “friend,” announced its seed round on March 10, 2026, just four months after it was founded. The round values the company at $3.5 billion on a pre-money basis and is believed to be the largest seed round ever raised by a European startup.
Five companies co-led it: Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, the vehicle through which Amazon founder Jeff Bezos makes personal investments. Nvidia, Toyota, Samsung and Singapore’s Temasek also participated, along with French venture capital firm Daphni, South Korean investor SBVA and a long list of notables, including Tim and Rosemary Berners-Lee, venture capitalist Jim Breyer, entrepreneur Mark Cuban and former Google CEO Eric Schmidt.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
LeCun was initially seeking around €500 million, according to a leaked presentation reported by Sifted. Demand exceeded that figure significantly. It ended up with 890 million euros, about $1.03 billion, and told reporters this week that interest had been high enough that AMI could be selective about which investors it accepted.
The company’s headquarters are in Paris, with additional offices planned for New York, Montreal and Singapore. LeCun, who has dual French-American citizenship and remains a computer science professor at New York University, will be CEO. Day-to-day operations will be led by Alexandre LeBrun, a French entrepreneur who previously founded and led medical AI startup Nabla, and who now becomes CEO of AMI.
The rest of the founding team comes almost entirely from Meta’s AI research organization. Michael Rabbat, former director of scientific research at Meta, joins as vice president of global models. Laurent Solly, former vice president of Meta for Europe, becomes director of operations. Pascale Fung, formerly senior director of AI research at Meta, takes on the role of director of research and innovation. Saining Xie, formerly of Google DeepMind, becomes chief scientific officer.
What exactly is AMI building? The short answer is world models, a category of AI system that LeCun has been advocating for and working on for years. The longer answer requires understanding why you think the industry has taken the wrong turn.
The case against LLMs
Large language models learn by predicting which word comes next in a sequence. They have been trained with large amounts of human-generated text and the results have been remarkable: ChatGPT, Claude and Gemini have demonstrated their ability to generate fluent and plausible language on a huge variety of topics. But LeCun has been arguing for years, loudly and repeatedly, that this approach has fundamental limits.
Its alternative is JEPA: Joint Embedding Predictive Architecture, a framework it first proposed in 2022. Instead of predicting the future state of the world with pixel-by-word or word-by-word precision, the approach that makes generative AI powerful and prone to hallucinations, JEPA learns abstract representations of how the world works, ignoring unpredictable surface details. The idea is to build systems that understand physical reality as humans and animals do: not through language, but through bodily experience.
Within a year or two, LeCun told AFP, AMI plans to begin talks with corporate partners. Within three to five years, he said, the goal is to produce “fairly universal intelligent systems” capable of being deployed in almost any domain that requires artificial intelligence. He wants AMI, he added, to become “the leading provider of intelligent systems.”
The timing and geography of the announcement are not coincidental. LeCun has been explicit about AMI’s positioning as European, and specifically French, vis-à-vis American and Chinese AI giants. “We are one of the few frontier artificial intelligence laboratories that are neither Chinese nor American,” he said. The choice of Paris as the headquarters and the participation of French investors Cathay Innovation and Daphni reflect this framework.
Whether or not that ambition is achievable remains to be seen. AMI has no product, no revenue, and no short-term prospects for either. LeCun acknowledged to reporters this week that the company would spend its first year focused exclusively on research and development. World models, according to him, are a long-term scientific project, not the type of AI startup that launches a product in three months and generates revenue in six.
What the $1.03 billion seed round demonstrates, for now, is that the investors behind it are willing to wait. LeCun has one of the most credible research backgrounds in AI, sharing the Turing Award in 2018 for his work on convolutional neural networks that underpin most modern computer vision, and his argument that LLMs have fundamental architectural limits has been consistent enough, and for long enough, that dismissing it is no longer the safe assumption it once was. The question is whether being right about the problem is the same as being right about the solution.
