Yann LeCun’s AMI Labs brings world-model bet to market

Yann LeCun left Meta in November 2025 after 12 years as chief AI scientist and is now executive chairman of AMI Labs, a Paris-based startup building world models as an alternative to LLMs. By March 2026, the company had raised $1.03 billion at a $3.5 billion pre-money valuation to pursue AI systems grounded in physical-world reasoning, persistent memory, planning, and action.

LeCun’s critique has been consistent since 2022: scaling LLMs is not, in his view, the path to human-level intelligence because next-token prediction does not create an internal model of how the physical world behaves. His preferred Joint Embedding Predictive Architectures, or JEPA, predict in abstract latent space, aiming to learn the semantic structure of scenes rather than reconstruct surface details.

The break with Meta followed a 2025 restructuring that shifted the company toward faster product delivery and competitive parity with OpenAI and Google. After Meta’s $14.5 billion deal with Scale AI, Alexandr Wang, the startup’s 28-year-old CEO, was elevated to lead Meta Superintelligence Labs, underscoring a move away from FAIR’s long-horizon research culture.

AMI Labs has operations in Paris, New York, Montreal, and Singapore. Its V-JEPA 2-AC model was trained on 62 hours of unlabelled robot observation video and completed a new physical pick-and-place task with an 80% success rate after forming a plan in 16 seconds. The company shipped three world-model research papers within 60 days of LeCun’s departure, but has no announced product or enterprise contract.