Yann LeCun’s AMI Secures $1.03 Billion to Challenge LLM Dominance with "World Model" AI Systems

Ex-Meta AI chief Yann LeCun raises $1.03B for his new startup, AMI. The company focuses on "world models" and reasoning AI beyond current LLM capabilities.

By: AXL Media

Published: Mar 10, 2026, 4:20 AM EDT

Source: The information in this article was sourced from CNA

A Billion-Dollar Bet on "World Models"

Advanced Machine Intelligence (AMI) announced a massive $1.03 billion funding round on Tuesday, signaling a major shift in the AI landscape. Founded by Yann LeCun—the former Meta Chief AI Scientist and Turing Award winner—the startup is built on the premise that current Large Language Models (LLMs) are fundamentally limited. With a pre-money valuation of $3.50 billion, the round was co-led by heavyweights including Greycroft, Bezos Expeditions, and Cathay Innovation. The capital infusion establishes AMI as a formidable rival to "predictive" AI giants, focusing instead on LeCun’s vision of "world models" that can reason and plan autonomously.

Moving Beyond Word Prediction

LeCun, who left Meta in late 2025, has been a vocal critic of the industry’s reliance on autoregressive LLMs. In an interview with Reuters, he emphasized that simply predicting the "next word or pixel" is insufficient for true intelligence. AMI’s approach centers on building systems that understand the physical world and can navigate complex, multi-step tasks. This philosophy puts AMI in direct ideological competition with Meta’s newly formed "Superintelligence Labs," now led by former Scale AI CEO Alexandr Wang, which continues to double down on the scaling of LLMs.

Industrial Intelligence: The Primary Target

Unlike many AI startups focusing on chatbots or creative tools, AMI is targeting heavy industry as its primary market. Near-term customers include manufacturers, automakers, and aerospace firms—organizations that require AI capable of managing complex, physical-world variables. LeCun believes that for AI to be truly useful in these sectors, it must move beyond statistical probability and toward a functional understanding of cause and effect. "We want to become the main provider of intelligent systems, regardless of what the application is," LeCun stated.