On March 9, 2026, TechCrunch reported that AMI Labs, the startup Yann LeCun co-founded after leaving Meta, has closed a $1.03 billion seed round (roughly €890 million) at a $3.5 billion pre-money valuation — the largest seed round in European history. The money has exactly one purpose: to prove that world models are a better path to machine intelligence than large language models.
For developers and product teams, the signal is not the number but the roster. Nvidia, Samsung, Toyota Ventures, and Temasek all came in as strategic investors, which means chips, manufacturing, and capital markets simultaneously placed a bet on a non-LLM route. A Turing Award winner leaving Big Tech with a fundamental-research agenda normally gets a polite small check; this time he got ten digits.
$1.03 Billion: Europe’s Largest Seed Round Ever
Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions co-led the round. Nvidia, Samsung, Sea, Temasek, and Toyota Ventures participated as strategic backers. The French contingent is long: Association Familiale Mulliez, Groupe Industriel Marcel Dassault, and Publicis Groupe, alongside funds such as Bpifrance Digital Venture, Artémis, Aglaé Lab, and SBVA. Individual angels include Eric Schmidt, Mark Cuban, Jim Breyer, Xavier Niel, and Tim and Rosemary Berners-Lee.
The pace of the expansion matters. In December 2025, rumors had AMI Labs targeting just €500 million; the final figure nearly doubled that. Compare Fei-Fei Li’s World Labs, which raised $1 billion in February, and SpAItial, which closed a $13 million seed — capital concentration in the world-model race is rising fast. European AI fundraising is hot overall — see Nscale’s $2 billion Series C — and the phrase “European record” has been rewritten several times in a matter of weeks.
World Models and JEPA: The Other Path
The technical core of AMI Labs is JEPA, the Joint Embedding Predictive Architecture LeCun proposed in 2022. One sentence: AI that learns from reality, not just from language. The model understands the physical world by predicting abstract representations of what it observes, rather than grounding intelligence in next-token prediction.
LeCun and CEO Alexandre LeBrun have both argued publicly that LLMs carry a limitation that refuses to die: hallucination. In healthcare, a single hallucination can be lethal. Their judgment is that this is not a patchable application-layer problem but an assumption-level difference — language data cannot cover the common sense of the physical world.
The cold water deserves equal billing. AMI Labs positions itself as fundamental research first: no revenue plans for now, and commercialization could take years. LeBrun’s own forecast is blunt — “world models” is about to become a buzzword, and “in six months, every company will call itself a world model to raise funding.” The only honest test of who is real: a verifiable architecture and reproducible experiments.
The Team and Four Locations
The team sheet reads like a manifesto. LeCun serves as chairman and still teaches at NYU. CEO Alexandre LeBrun is chairman of digital-health company Nabla and previously worked at Meta’s FAIR. COO Laurent Solly was Meta’s VP for Europe. Research is led by chief science officer Saining Xie, chief research and innovation officer Pascale Fung, and VP of world models Michael Rabbat.
The company operates from four cities: Paris headquarters, New York, Montreal, and Singapore — the last chosen for proximity to Asian clients and talent. LeBrun’s description of costs is equally direct: two main cost centers, compute and talent, and hiring favors quality over quantity.
Starting in Healthcare: Nabla and Open Research
AMI Labs’ first partner is Nabla, the digital-health company LeBrun founded. Starting with medicine is consistent with the technical thesis: hallucinations carry the highest cost in clinical settings, which makes healthcare the sharpest test of whether world models are genuinely more reliable than LLMs. Although there are no near-term revenue plans, the team intends to engage prospective customers early for real-world data and evaluation feedback.
Openness is the other commitment. AMI Labs says it will keep publishing papers and open-source a lot of code; in LeBrun’s words, “we think things move faster when they’re open.” For the research community, that is the most concrete externality of this round: whatever happens to the route, the code and papers will be public.
What It Means for Developers
Three things to watch. First, don’t rush to write “world model” into your pitch deck — calibrate against the original JEPA papers and the code AMI releases later, and use that to filter the imitators. Second, if you build for healthcare or robotics, follow Nabla’s integration progress; it is the first public litmus test of world-model reliability. Third, the open-source commitment means engineering assets will appear incrementally — for teams working on evaluation, simulation, and data pipelines, this is a low-cost way to participate in a new route. LLMs are not about to be replaced, but the “learn from reality” hypothesis finally has a decade’s worth of ammunition.
Sources
- Yann LeCun’s AMI Labs raises $1.03 billion to build world models — TechCrunch
- World model AI lab AMI raises Europe’s largest seed round — Crunchbase News
- Yann LeCun’s AMI Secures $1B Seed to Develop AI World Models — HPCwire
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.