AMI Labs raises $1B in Europe's largest seed round to rethink AI from scratch
What's the deal? Yann LeCunDealroom has a profile for this one. Try Dealroom →'s new AI venture, AMI Labs, has raised $1.03 billion at a $4.5 billion post-money valuation to develop so-called "world models" — AI systems designed to learn from real-world interactions rather than just text data. The company is led by chief executive officer Alexandre LeBrun and is headquartered in Paris.
The round was co-led by Cathay InnovationDealroom has a profile for this one. Try Dealroom →, Greycroft, Hiro CapitalDealroom has a profile for this one. Try Dealroom →, HV Capital, and Bezos Expeditions, with participation from Nvidia, SamsungDealroom has a profile for this one. Try Dealroom →, Temasek, and Toyota VenturesDealroom has a profile for this one. Try Dealroom →.
The final raise is significantly larger than early expectations. In December 2025, reports indicated the startup was seeking about €500 million at a roughly €3 billion valuation before launch — underscoring how investor demand grew as the project took shape.
Why now? Interest in world models is rising as researchers question the limits of large language models (LLMs). Advocates argue that systems trained mainly on text struggle to reason about the physical world and can produce unreliable outputs.
AMI Labs joins a small but growing group exploring the concept, including World Labs, founded by Fei-Fei LiDealroom has a profile for this one. Try Dealroom →, which has also raised a large sum to pursue similar research. The surge in funding reflects broader investor appetite for foundational AI breakthroughs — particularly when led by well-known researchers like LeCun.
What could go wrong? AMI Labs is pursuing fundamental research rather than near-term products, meaning revenue could take years to materialise. High computing costs, scarce AI talent, and the need to sustain investor patience over a long horizon all add risk.
There is also scientific uncertainty. World models remain largely experimental, and it is still unclear whether architectures such as LeCun's Joint Embedding Predictive Architecture will scale into reliable systems.
The signal: The deal highlights a shift in AI investment toward deeper research into new architectures beyond today's generative models. It also shows how the reputations of leading scientists can unlock large funding rounds before a startup ships a single product — a dynamic that rewards long-term bets but raises the stakes if progress stalls.
Sources:
TechCrunch
Reuters
WSJ
A.M.
