Yann LeCun's AMI Labs Raises $1.03 Billion in Europe's Largest-Ever Seed Round to Build 'World Models'

Yann LeCun spent more than a decade as the public face of Meta's AI ambitions and one of the intellectual architects of the deep-learning era. Now he is betting roughly a billion dollars that the industry he helped build is chasing the wrong idea.

Advanced Machine Intelligence (AMI) Labs, the Paris-based startup the Turing Award winner co-founded after departing his role as Meta's chief AI scientist, has closed $1.03 billion (about 890 million euros) in seed funding at a $3.5 billion pre-money valuation. The raise is described as the largest seed round in European startup history, per Crunchbase data, and ranks among the region's biggest fundings for an AI company of any stage. The capital lands as AMI Labs presses ahead with a contrarian thesis: that the large language models powering today's AI boom are a dead end on the road to genuine machine intelligence.

The company is building what researchers call "world models" — AI systems designed to learn how physical reality behaves through sensory data such as video and cameras, rather than by predicting the next word in a stream of text. It is the architecture LeCun has championed for years, even as the rest of the field, including his former employer, poured tens of billions of dollars into ever-larger language models.

A billion-dollar bet against the consensus

The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Jeff Bezos's Bezos Expeditions, with a strikingly industrial roster of additional backers. Nvidia, Samsung, Sea, Temasek and Toyota Ventures all participated, alongside French heavyweights including Bpifrance, Groupe Industriel Marcel Dassault, Publicis Groupe and the Mulliez family's Association Familiale Mulliez. The angel list reads like a who's-who of the tech and finance worlds: Tim and Rosemary Berners-Lee, Jim Breyer, Mark Cuban, Xavier Niel and Eric Schmidt.

That the round drew so many strategic and corporate investors is no accident. AMI Labs is led day-to-day by CEO Alexandre LeBrun — a serial entrepreneur and chairman of digital-health startup Nabla — with LeCun serving as chairman. The founding team also includes former Meta Europe VP Laurent Solly as COO, Saining Xie as chief science officer, Pascale Fung as chief research and innovation officer, and Michael Rabbat as VP of world models.

LeBrun has been candid that the company is a long-horizon research bet, not a product shop. "AMI Labs is a very ambitious project, because it starts with fundamental research. It's not your typical applied AI startup that can release a product in three months, have revenue in six months, and make $10 million in [annual recurring revenue] in 12 months," he told TechCrunch. The work is grounded in JEPA, or Joint Embedding Predictive Architecture, an approach LeCun first proposed in 2022.

He is also bracing for a wave of imitators chasing the same funding. "My prediction is that 'world models' will be the next buzzword," LeBrun said. "In six months, every company will call itself a world model to raise funding." He argues AMI is fundamentally different because its goal is to understand the real world, not merely to generate plausible-sounding output.

Why world models matter — and why LLMs may not be enough

The pitch cuts to the heart of a debate splitting the AI field. Large language models are trained to predict tokens — fragments of text — which makes them formidable at tasks like summarization, coding and information retrieval. But LeCun and LeBrun contend that token prediction, however powerful, cannot capture the messy, continuous nature of the physical world.

"Generative architecture trained by self-supervised learning mimic intelligence; they don't genuinely understand the world," LeBrun wrote on LinkedIn. "However, factories, hospitals, and robots operating in open environments demand AI that grasps reality. And reality is not tokenized: it's continuous, noisy and high-dimensional. Despite their immense power, I do not believe that generative architectures are the path to achieving this true understanding."

That framing points AMI Labs squarely at robotics, manufacturing and healthcare — domains where a hallucinating chatbot is not a quirk but a liability. Nabla, where LeBrun is chairman, is AMI's first disclosed partner and is expected to get early access to the models; the company says the strong presence of industrial investors in the round reflects similar appetite among potential partners. AMI plans to engage prospective customers early, even without near-term revenue. "We are developing world models that seek to understand the world, and you can't do that locked up in a lab," LeBrun said. "At some point, we need to put the model in a real-world situation with real data and real evaluations."

The raise also lands as world models attract serious money elsewhere. Fei-Fei Li's San Francisco-based World Labs recently pulled in $1 billion of its own, signaling that some of AI's most decorated researchers see the physical world, not just text, as the next frontier.

A statement for European AI

The deal is as much a milestone for Europe as it is for LeCun. The continent has lagged the U.S. in megadeals — global venture funding hit a monthly record in February when OpenAI announced a $110 billion round — with only a handful of billion-dollar-plus AI deals, such as Mistral AI's $2 billion raise. AMI Labs, headquartered in Paris with planned hubs in New York, Montreal and Singapore, plants a flag for the idea that frontier AI research can be funded and built on European soil. LeBrun has said the team intends to publish papers and open-source much of its code, a deliberate echo of the open-research culture LeCun cultivated at Meta's FAIR lab.

What to watch next

The hard part starts now. AMI Labs has the money, the talent and the brand, but world models remain largely unproven at scale, and LeBrun himself concedes commercial applications could be years away. Watch for the first technical papers and open-source releases as early proof points, for how the Nabla healthcare partnership performs in real clinical settings, and for whether the flood of "world model" startups LeBrun predicts materializes — and dilutes the term before AMI can define it. The biggest question of all: whether one of the founders of the deep-learning revolution can convince the industry that its current path, the one he helped pave, leads somewhere short of true intelligence.