Yann LeCun just pulled back the curtain on AMI Labs, and the AI legend's new venture is making waves before it's even raised a dollar. The startup confirmed this week it's building world models to create AI systems that understand the real world - putting it on a collision course with rival World Labs, which just hit a reported $5 billion valuation. With LeCun as executive chairman and a leadership team poached from Meta, AMI is already in funding talks at a $3.5 billion valuation, according to Bloomberg.
AMI Labs just became the latest entrant in AI's hottest race - and it's got one hell of a pedigree. Yann LeCun's new venture confirmed this week that it's building world models, the cutting-edge AI architecture that promises to bridge the gap between artificial intelligence and physical reality. The startup's mission is deceptively simple: "build intelligent systems that understand the real world."
The announcement puts AMI squarely in competition with World Labs, founded by AI pioneer Fei-Fei Li, which became a unicorn shortly after emerging from stealth. World Labs recently launched Marble, a product that generates physically sound 3D worlds, and is now reportedly in talks to raise fresh funding at a $5 billion valuation according to Bloomberg. Building foundational models that understand physics, space, and causality has become one of AI's most exciting pursuits - attracting top scientists and deep-pocketed investors alike, product or no product.
VCs are already circling. AMI Labs might be raising funding at a $3.5 billion valuation, with investors including Cathay Innovation, Greycroft, and Hiro Capital (to which LeCun is an advisor) reportedly in talks, according to Bloomberg. Other potential backers include 20VC, Bpifrance, Daphni, and HV Capital. That's a massive number for a startup that just launched its website, but LeCun's track record speaks for itself - the Turing Prize winner pioneered convolutional neural networks and led Meta's AI research lab FAIR for years.
Here's the twist: LeCun isn't CEO. He's executive chairman. The CEO role belongs to Alex LeBrun, previously co-founder and CEO at Nabla, a health AI startup with offices in Paris and New York. LeBrun knows a thing or two about building AI companies - Facebook acquired his previous startup, Wit.ai, back in 2015, after which he worked under LeCun's leadership at FAIR.
LeBrun's transition from Nabla to AMI is part of a strategic partnership announced last December. In exchange for privileged access to AMI's world models, Nabla's board supported LeBrun's shift from CEO to chief AI scientist and chairman, clearing the way for his new role. The deal gives AMI a direct path into healthcare applications - a sector where LeBrun has deep expertise and where the limitations of current AI systems are painfully obvious.
The leadership team keeps getting more interesting. Laurent Solly, who stepped down as Meta's vice president for Europe last December, is joining AMI according to French media reports. The talent overlap between AMI and Meta likely won't stop there. LeCun told the MIT Technology Review that his former employer could well be AMI's first client.
But AMI Labs represents more than just a Meta spinout - it's a contrarian bet against the entire large language model paradigm that's dominated AI for the past few years. LeCun has been publicly critical of LLMs, pointing out fundamental limitations including hallucinations, which are a serious concern in contexts like medicine. "Real intelligence does not start in language," AMI's mission statement declares. "It starts in the world."
Unlike generative approaches, which LeCun and his team see as poorly suited for unpredictable data such as sensor input, AMI promises AI systems with persistent memory, the ability to reason and plan, and controllability that actually works. The startup is targeting high-stakes applications "where reliability, controllability, and safety really matter, especially for industrial process control, automation, wearable devices, robotics, healthcare, and beyond."
LeBrun told Forbes that a big reason he took the CEO role was the prospect of applying world models to healthcare. It's a natural fit given his Nabla background, and it positions AMI to tackle problems where today's AI systems fail spectacularly - like understanding physical consequences, maintaining context over time, or operating safely in unpredictable environments.
The business model is taking shape too. AMI plans to license its technology to industry partners for real-life applications, while also contributing to AI research "with the global academic research community via open publications and open source." LeCun said he plans to keep his professor position at NYU, where he teaches one class per year and supervises PhD and postdoctoral students.
Geography matters here. While LeCun will remain based in New York, he told MIT Technology Review that AMI Labs "is going to be a global company [that's] headquartered in Paris." French President Emmanuel Macron expressed his pride that LeCun chose Paris, saying "we will do everything we can to ensure his success from France." The startup will also have offices in Montreal, New York, and Singapore.
The Paris headquarters decision consolidates the city's reputation as an AI hub, where AMI will join H, Mistral AI, and several international labs including Meta's own FAIR. It's fitting that AMI is pronounced "a-mee" - like "ami" in French, which means friend, LeCun has pointed out. Whether that friendly positioning extends to Meta, where LeCun spent years building FAIR before departing to launch AMI, remains to be seen. But with a potential $3.5 billion valuation and a team of AI heavy hitters, AMI Labs is already reshaping the world model race.
AMI Labs' emergence marks a critical inflection point in AI development - a bet that the future belongs to systems that understand physics and causality, not just language patterns. With LeCun's scientific credibility, LeBrun's startup execution chops, and a reported $3.5 billion valuation on the table, AMI has the ingredients to challenge World Labs and reshape how AI interacts with the physical world. The real test comes next: can world models deliver on their promise in high-stakes applications like healthcare and robotics where LLMs have fallen short? If AMI's first client really is Meta, we'll get our answer faster than expected.
