Yann LeCun Launches AMI Labs With $1.03 Billion Seed Round to Build World Models

Yann LeCun, Turing Award laureate and former chief AI scientist at Meta, has announced the launch of Advanced Machine Intelligence (AMI) Labs alongside a historic $1.03 billion seed round (approximately €890 million). The funding values the company at $3.5 billion pre-money, making it the largest seed round ever raised by a European startup.

Key Highlights

$1.03 billion raised in a seed round, an all-time European record
$3.5 billion pre-money valuation just four months after founding
JEPA architecture (Joint Embedding Predictive Architecture) at the core
Top-tier investors: NVIDIA, Bezos Expeditions, Samsung, Temasek, Eric Schmidt, and Mark Cuban
Four offices: Paris, New York, Montreal, and Singapore

A World-Class Team

AMI Labs brings together some of the most respected names in AI research:

Yann LeCun — Executive Chairman, 2018 Turing Award winner
Alexandre LeBrun — CEO, former founder of Nabla (medical AI)
Mike Rabbat — VP of World Models, former Meta research director
Saining Xie — Chief Science Officer, former Google DeepMind researcher
Pascale Fung — Chief Research & Innovation Officer, former Meta senior director

World Models: Beyond LLMs

The project is built on LeCun's long-held conviction that large language models (LLMs) are not the path to AI that truly understands the real world. While LLMs predict the next token in a text sequence, world models learn to understand the physical dynamics of reality — gravity, causality, and object interactions.

At the heart of this approach is JEPA (Joint Embedding Predictive Architecture), a framework LeCun first proposed in 2022. Unlike generative models that predict raw pixels, JEPA works in a compressed, abstract representation space. The model learns to predict high-level representations rather than raw data, making it more efficient and robust.

The first product under development, called "AMI Video," is a world model trained on video to understand physical environments. The V-JEPA 2 prototype has already demonstrated zero-shot robot planning capabilities with just 62 hours of training data.

A Global Investor Consortium

The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions. Other investors include NVIDIA, Toyota, Samsung, Singapore sovereign fund Temasek, French VC firm Daphni, and South Korean investor SBVA.

Notable individual backers joined as well: Tim and Rosemary Berners-Lee (inventor of the Web), Jim Breyer (venture capitalist), Mark Cuban (entrepreneur), and Eric Schmidt (former Google CEO).

A Two-Horse Race in World Models

AMI Labs enters an emerging competitive field. Fei-Fei Li, another leading figure in computer vision, raised $1 billion for World Labs, which is also developing world models but with a more product-oriented approach. While World Labs has already shipped its Marble product with Autodesk integration, AMI Labs is taking a more research-first approach.

"My prediction is that 'world models' will be the next buzzword. In six months, every company will call itself a world model to raise funding," said Alexandre LeBrun, CEO of AMI Labs.

Industry Implications

This massive fundraise sends a strong market signal: the "everything is an LLM" era may be drawing to a close. With over $2 billion collectively invested in world models by AMI Labs and World Labs, this alternative approach is gaining institutional credibility.

For robotics, manufacturing, autonomous vehicles, and augmented reality, models capable of understanding real-world physics represent a potentially transformative breakthrough. If JEPA delivers on its promise, AI systems could finally move beyond text prediction to achieve functional understanding of their environment.

LeCun summarizes his vision: current LLMs are a "dead end" for achieving artificial general intelligence. With AMI Labs, he is betting $1 billion on an alternative that could redefine the very foundations of AI.

Source: TechCrunch