Yann LeCun isn't just leaving Meta Platforms Inc. (META). He's building something that could fundamentally change how AI understands reality itself, and apparently investors are willing to bet big on that vision.

Building AI That Actually Understands the World

LeCun, who's wrapping up his tenure as Meta's AI chief at the end of this year, is in serious talks to raise €500 million (about $560 million) for his new startup, according to a Financial Times report. The company would launch with a valuation around €3 billion ($3.4 billion), which puts it in rarified air before it even opens its doors.

The venture, called Advanced Machine Intelligence Labs, is scheduled to officially launch in January. LeCun will serve as executive chair, and he's reportedly considering Alexandre LeBrun, founder of French health tech startup Nabla, to take the CEO role.

Here's what makes this interesting: LeCun isn't trying to build a bigger, faster version of ChatGPT. His startup aims to create "world models" that let AI systems actually comprehend the physical environment around them. Think robotics, transportation, and applications where understanding three-dimensional space and physics actually matters.

The core technology comes from work LeCun developed during his time at Meta. His new AI architecture is designed to learn from videos and spatial data, maintain long-term memory, and reason through complex sequences of actions. It's fundamentally different from current large language models that primarily learn from text.

The relationship with Meta is worth noting. Despite not investing directly in the new company, Meta will form a partnership with Advanced Machine Intelligence Labs that gives the tech giant access to commercialize whatever technology emerges. Everyone wins, apparently.

A Strategic Exit During Meta's AI Overhaul

LeCun's departure was first reported back in November, and the timing tells its own story. Meta CEO Mark Zuckerberg has been aggressively revamping the company's AI strategy to compete with heavy hitters like OpenAI and Google (GOOG) (GOOGL).

According to Hyperbolic CTO Yuchen Jin, LeCun's exit was somewhat inevitable. Jin claimed that Zuckerberg's support for Alexandr Wang and a broader shift in AI leadership effectively limited LeCun's role at the company.

LeCun's new venture joins a remarkable year for AI fundraising. Back in April, OpenAI co-founder Ilya Sutskever pulled in $2 billion for Safe Superintelligence, landing a $32 billion valuation before the company had even shipped a product. The numbers keep getting bigger.

It's worth remembering what LeCun said back in May: AI won't advance simply by throwing more data and computing power at the problem. His argument then was that systems need to learn to understand the physical world rather than just relying on scale. Now he's putting that philosophy into practice with his own company and someone else's half-billion dollars.