Yann LeCun Raises $1 Billion to Build AI That Understands the Physical World

Advanced Machine Intelligence (AMI), a new Paris-based startup co-founded by Meta's former chief AI scientist Yann LeCun, has secured over $1 billion in funding to pursue a fundamentally different path to artificial intelligence. The round, co-led by investors including Cathay Innovation and Bezos Expeditions, values the company at $3.5 billion. LeCun, a 2018 Turing Award winner, argues that the industry's focus on scaling large language models (LLMs) like GPT-4 and Claude is misguided, stating that "the idea that you're going to extend the capabilities of LLMs to the point that they're going to have human-level intelligence is complete nonsense." Instead, AMI will develop "world models"—AI systems that understand the physical world, possess persistent memory, and can reason and plan.

AMI represents a direct challenge to the strategies of OpenAI, Anthropic, and even LeCun's former employer, Meta. The startup, with offices in Paris, Montreal, Singapore, and New York, will commercialize research on architectures like Meta's Joint-Embedding Predictive Architecture (JEPA). LeCun left Meta in November 2025, convinced that selling enterprise-focused world models did not fit the social media giant's core consumer business. AMI's first applications will be in industries with rich physical data, such as manufacturing and biomedicine. For example, the company could build a detailed model of an aircraft engine to help manufacturers optimize for efficiency and reliability. The founding team includes several former Meta AI leaders and Alexandre LeBrun, ex-CEO of health AI startup Nabla, who will serve as AMI's CEO.