The archive · AI & Models · Strategic decision · 2025–2026
AMI Labs bets world models replace LLMs; $1.03B seed, Europe's largest
Yann LeCun's Paris lab raises $1.03B at $3.5B pre-money to build AI that learns from the physical world, not just text.
AMI Labs (Advanced Machine Intelligence)
What the business is
Paris AI lab developing 'world models' — AI that learns from and interacts with the physical world via JEPA (Joint Embedding Predictive Architecture) — with healthcare startup Nabla as its first deployment partner.
Starting capital：$1.03B (about €890M) seed round at a $3.5B pre-money valuation, announced March 2026
How it started
After leaving Meta in late 2025, Turing Award winner Yann LeCun co-founded AMI Labs in Paris with Alexandre LeBrun, then CEO of healthcare AI startup Nabla. Both had concluded that LLMs can't be trusted where errors carry real-world consequences, so they set out to build AI that learns from reality rather than text.
What happened
AMI assembled a senior team — Laurent Solly (COO, ex-Meta VP Europe), Saining Xie (chief science officer), Pascale Fung and Michael Rabbat — and in March 2026 raised $1.03B at a $3.5B pre-money valuation, reportedly more than the €500M it first sought. The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with Nvidia, Samsung, Sea, Temasek, Toyota Ventures and French groups Publicis, Dassault and Mulliez among backers; plans include offices in Paris, New York, Montreal and Singapore.
How it ended up
No revenue or product yet: AMI says it will publish papers and open-source code while it develops world models, with Nabla as the first partner to access early models.
Background
AMI Labs (Advanced Machine Intelligence) is a Paris-based AI startup co-founded by Turing Award winner Yann LeCun after he left Meta, with Alexandre LeBrun as CEO. It is building 'world models': AI designed to learn from and interact with three-dimensional reality, rather than predict text tokens.
The bet rests on JEPA, the Joint Embedding Predictive Architecture LeCun proposed in 2022. AMI argues that generative LLMs 'mimic intelligence' but don't understand the world, and that factories, hospitals and robots need AI that grasps continuous, noisy, high-dimensional reality.
In March 2026 AMI raised $1.03B at a $3.5B pre-money valuation — the largest seed round ever for a European startup, per Crunchbase — co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with Nvidia, Samsung, Sea, Temasek, Toyota Ventures and French industrial groups among backers. It reportedly sought about €500M initially.
The lab plans offices in Paris, New York, Montreal and Singapore, says it will publish papers and open-source code, and has named Nabla — LeBrun's healthcare AI company, where he remains chairman — as the first partner to access early models, starting with applications where hallucination is unacceptable.
What has to be true
- LeCun's authority gives the world-model thesis credibility that a no-name lab couldn't command, attracting Bezos Expeditions, Nvidia and sovereign-linked backers.
- LLM hallucinations are a hard ceiling in safety-critical fields like healthcare, where AMI's first partner Nabla operates.
- JEPA is a concrete technical alternative to next-token prediction, not just a criticism of LLMs.
- The 'coconut round' wave — after World Labs' $1B and similar mega-seeds — made investors comfortable funding a research-stage lab at a $3.5B valuation.
What can be applied
A famous researcher's contrarian thesis can raise a billion-dollar seed before any product exists; the commercial test comes later, when world models must justify their compute and talent runway.
Aftermath
As of Sept 2026 AMI Labs remains in research mode with no revenue plans; it intends to publish papers and open-source code, keep teams in Paris, New York, Montreal and Singapore, and let Nabla test early models in healthcare, where errors are life-threatening. Whether world models become the next buzzword or the next breakthrough is still unproven.
Sources
- Yann LeCun's AMI Labs raises $1.03B to build world models
- Turing Winner LeCun's New 'World Model' AI Lab Raises $1B In Europe's Largest Seed Round Ever
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card