LeCun just raised $1B to prove ChatGPT is built wrong

Yann LeCun, the Turing Award-winning scientist behind much of modern AI, just raised $1.03 billion in seed funding for his new company, AMI Labs. It's the largest seed round in European history — and it's built on one audacious claim: ChatGPT, Claude, and Gemini are fundamentally wrong about how to build intelligence.

AMI Labs isn't building another chatbot. It's building what LeCun calls "world models" — AI systems that understand how reality works, the way humans and animals do, rather than just guessing the next word in a sentence.

Why LeCun thinks every major AI company got it wrong

After 12 years leading AI research at Meta, LeCun left in November 2025 with a blunt message: "Large language models are a statistical illusion. Impressive, yes. Intelligent, no."

His argument is straightforward. ChatGPT, Claude, and Gemini all work the same way — they predict the next word in a sentence, billions of times over. That's why they can write essays and answer questions. But it's also why they hallucinate (confidently make things up) and struggle with basic reasoning about the physical world.

LeCun believes real intelligence doesn't start with language. It starts with understanding how the world actually works — physics, cause and effect, what happens when you push a cup off a table. A baby learns this before it speaks a single word. Current AI can't.

JEPA: the technology behind LeCun's billion-dollar bet

AMI Labs is built on a framework LeCun proposed in 2022 called JEPA (Joint Embedding Predictive Architecture). Here's the simplest way to understand it:

Instead of processing text tokens one by one, JEPA processes cameras, video, audio, and sensor data. It learns abstract representations — compressed mental models of how reality works — while ignoring unpredictable surface details like exact pixel colors or background noise.

The result, in theory: AI that can plan actions, predict consequences, and reason about cause and effect — without hallucinating, because it actually understands the world rather than just predicting text patterns.

Who's backing this — and who's building it

LeCun initially sought €500 million. Investor demand was so overwhelming he raised nearly double that, becoming selective about who could participate.

The founding team reads like a who's-who of AI research, drawn almost entirely from Meta's AI division:

Investor names include Jeff Bezos (through Bezos Expeditions), Nvidia, Samsung, Toyota, Singapore's Temasek, and individuals like Mark Cuban, Eric Schmidt (former Google CEO), and Tim Berners-Lee (inventor of the World Wide Web). French President Emmanuel Macron celebrated the launch as a "new page for AI."

Not a chatbot factory — a 5-year scientific mission

AMI Labs is refreshingly honest about what it isn't: a company with products, customers, or revenue.

The roadmap LeCun shared publicly:

• Year 1-2: Pure research and development. Begin discussions with corporate partners.
• Year 3-5: Deploy "fairly universal intelligent systems" across multiple industries.
• Long-term goal: Become "the main provider of intelligent systems" worldwide.

Target applications include robotics (robots that understand physics), healthcare (AI with fewer hallucinations for medical decisions), industrial automation (factories that can predict equipment failures), and autonomous vehicles (cars that truly understand road physics, not just pattern-match from training data).

The European AI independence play

LeCun explicitly positioned AMI as "one of the few frontier AI labs that are neither Chinese nor American." With headquarters in Paris and backing from France's national investment bank Bpifrance, the company represents Europe's most serious attempt to compete with OpenAI, Anthropic, Google DeepMind, and China's growing AI ecosystem.

While Silicon Valley dominates the chatbot race, LeCun is betting that the next phase of AI — one built on understanding reality rather than generating text — doesn't have to be built in San Francisco.

The billion-dollar question

The AI community is split. Supporters call AMI an "overdue counterweight" to the industry's obsession with making chatbots bigger. Critics ask: can JEPA-style world models actually ship products before LLM-based agents run away with the market?

LeCun's diagnosis — that text prediction alone can't produce truly intelligent systems — has gone from fringe opinion to mainstream concern as hallucination problems persist across ChatGPT, Claude, and Gemini. But as one researcher noted: "Being right about the problem is not the same as being right about the solution."

With $1 billion, a world-class team, and the backing of some of tech's biggest names, LeCun has the resources to find out. The rest of the AI industry will be watching closely.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News