Yann LeCun’s AMI Labs raises $1.03 billion to build world models
The startup, co-founded by the Turing Prize winner, secured massive funding to build AI that learns from reality, not just text.
AMI Labs, the ambitious new AI venture co-founded by Turing Prize winner Yann LeCun, has secured a massive $1.03 billion in funding at a $3.5 billion pre-money valuation. The startup is dedicated to building 'world models'—a next-generation AI paradigm that learns from and understands the physical world, rather than just processing language like today's large language models (LLMs). CEO Alexandre LeBrun predicts this category will become the industry's next major buzzword, but argues AMI is fundamentally different by focusing on true world understanding, starting with a partnership in healthcare with digital health company Nabla.
The funding round, co-led by Cathay Innovation, Greycroft, and Bezos Expeditions among others, far exceeded initial targets and provides a long runway for fundamental research. Unlike typical applied AI startups, AMI Labs acknowledges it could take years to move from theory—based on LeCun's Joint Embedding Predictive Architecture (JEPA)—to commercial applications. The capital will fuel two main cost centers: massive compute requirements and top-tier talent recruitment across hubs in Paris, New York, Montreal, and Singapore. While revenue generation is not an immediate goal, the company plans early engagement with partners like Nabla to test models in real-world scenarios with real data.
- Raised $1.03B at a $3.5B valuation, co-led by Cathay Innovation and Bezos Expeditions, exceeding its initial €500M target.
- Building 'world models' based on Yann LeCun's JEPA architecture to understand reality, aiming to solve LLM hallucinations critical for fields like healthcare.
- First commercial partner is digital health startup Nabla; will recruit talent across Paris, New York, Montreal, and Singapore with no near-term revenue plans.
Why It Matters
This massive bet on fundamental AI research could lead to more reliable, reality-grounded systems, especially for critical applications like healthcare where current AI fails.