The largest seed round in Europe: $1.03 billion for a "rebellion against LLMs"

Advanced Machine Intelligence Labs (AMI Labs), which LeCun co-founded after leaving his position as head of AI research at Meta, announced on June 8, 2026, the closing of an investment round of $1.03 billion at a valuation of $3.5 billion (pre-money valuation). This is the largest seed round in European history and one of the largest early-stage AI investments ever. For comparison: France's Mistral AI raised €600 million in its last round in 2024 — AMI Labs surpasses that by nearly double.

Among the investors are names that hint at where the startup is headed. Nvidia will supply computing infrastructure, Toyota sees potential in autonomous driving and robotics, Samsung and Singapore's sovereign fund Temasek represent the industrial and strategic link. Other backers include Bezos Expeditions (Jeff Bezos' investment fund), Cathay Innovation, Greycroft, Hiro Capital, and HV Capital.

What world models are and why LeCun pits them against ChatGPT

LeCun's criticism of large language models (LLMs) is nothing new. Already during his time at Meta, where he contributed to the development of the Llama models, he repeatedly argued that predicting the next token will never result in genuine understanding of the world. "Intelligence really lies in learning," he said in an interview with the Financial Times. According to him, language models lack three key capabilities: planning, reasoning about the physical world, and common-sense memory.

The answer to these shortcomings is precisely world models. While ChatGPT or Claude learn exclusively from text data and operate on the principle of predicting the next word, a world model tries to understand how things work in the physical world — how objects move, how forces interact, what happens when you drop something. This is an approach much closer to how living beings learn.

"A child does not learn to understand the world just from spoken words. They have to see, touch, bump into things. Similarly, AI needs interaction with the physical environment," LeCun explains his philosophy. AMI Labs is therefore developing an architecture that combines learning from visual and sensory data with mathematical models of reality.

Toyota, robots, and industry: why this is not just an academic debate

Toyota's participation in the investment round is no coincidence. Automakers, manufacturing companies, and logistics firms urgently need AI that understands space, motion, and physics — something language models fundamentally cannot do. World models could become the foundation for:

• Autonomous vehicles — AI that understands traffic dynamics, not just recognizes signs
• Industrial robots — machines that adapt to new tasks without complex reprogramming
• Healthcare — systems understanding anatomy, movement, and patient interactions
• Smart transport infrastructure

This is especially relevant for European and Czech industry. Countries like the Czech Republic, Germany, and Poland are strong in manufacturing and the automotive sector. While AI chatbots will find applications more in administration and customer support, physical AI changes the rules exactly where Europe traditionally excels — in factories, on assembly lines, and in logistics.

Who is Yann LeCun: from Bell Labs to a billion-dollar startup

Yann LeCun is not among the fringe critics of the current AI boom. In 2018, together with Geoffrey Hinton and Yoshua Bengio, he received the Turing Award — the equivalent of the "Nobel Prize of computing" — for pioneering work on deep learning. He was at the birth of convolutional neural networks (CNNs), the architecture behind image recognition in your phone, autonomous vehicles, and computer vision systems.

From 2013 to 2025, he led the FAIR (Fundamental AI Research) lab at Meta, where he helped develop the open Llama models. His departure from a company that is investing billions in LLMs underscores the ideological rift within the AI community. LeCun literally stated that "people at Meta probably wouldn't be thrilled if I went around telling the world that LLMs are a dead end on the path to superintelligence."

He leads AMI Labs as Executive Chair, with CEO Alex LeBrun handling day-to-day operations of the company. The duo is betting on an approach LeCun has been theoretically developing for years — and now has a billion dollars to turn theory into practice.

Context: a market at a crossroads

AMI Labs is entering an environment where the AI industry is fragmenting. On one side stand text-oriented models (OpenAI GPT-5.5, Google Gemini, Anthropic Claude Opus 4.8), on the other side, demand for physical AI is growing — systems that understand 3D space. Nvidia is investing massively in this direction through the Omniverse platform and Isaac for robotics. Tesla is building the humanoid robot Optimus. And Chinese companies like AgiBot are already physically deploying robots in manufacturing.

LeCun himself sees AMI Labs as a direct alternative to the "text mainstream." In an interview with Wired, he said that "the next big leap in AI won't come from making chatbots bigger, but from teaching machines to understand the physical world."

For the Czech Republic, which is an industrially oriented economy with a strong automotive sector, this direction brings interesting implications. While no one is currently developing Czech language models (and probably won't — training GPT-4 cost over $100 million), applications of physical AI in robotics and industry are feasible even for smaller tech companies. It's not about training your own GPT, but about deploying specialized systems into specific manufacturing processes.

What's next: a billion-dollar bet

AMI Labs has not yet proven its technology in the market — the company has only existed since December 2025 and has yet to release a product. The billion dollars gives the startup the luxury of several years of development without pressure for immediate returns, which is rare in the AI industry. At the same time, it means enormous expectations: investors expect LeCun's team to achieve what no one has managed so far — to build AI that learns from the world similarly to humans.

The success or failure of AMI Labs will answer a fundamental question: Are language models just a transitional phase on the path to artificial general intelligence, or the definitive architecture? LeCun has bet a billion dollars on the former. And investors believe in him enough to give him the largest seed round in European history.