World model instead of LLM: Yann LeCun's startup receives 890 million euros

The startup founded by the AI pioneer, Turing Award winner, and former Meta AI Chief Scientist Yann LeCun has raised 890 million euros in a seed funding round. The company, Advanced Machine Intelligence Labs (AMI Labs), founded in Paris, thus records the largest funding round of its kind in Europe to date.

Investors include numerous venture capitalists as well as renowned companies and private investors, including Nvidia, Samsung, Toyota, Amazon founder Jeff Bezos, investor Mark Cuban, and former Google CEO Eric Schmidt.

LeCun founded the startup together with former Meta executives and former Nabla CEO Alex LeBrun.

The startup is only a few months old and, according to the New York Times, employs only twelve people, which is likely to change soon thanks to the new funding. AMI Labs has an international focus and, in addition to its headquarters in Paris, also has locations in New York, Montreal, and Singapore.

AMI Labs focuses on AI with world understanding

LeCun previously worked at Meta for twelve years, where he founded the AI lab FAIR. He left Meta in November 2025 to dedicate himself to developing so-called world models with his own company. LeCun is convinced that this type of AI model, rather than generative language models like ChatGPT, will pave the way to truly intelligent AI systems.

"We share one belief: real intelligence does not start in language. It starts in the world," reads the programmatic statement on the startup's website. Accordingly, the company aims to develop a new generation of AI systems that can understand the world, have a permanent memory, and can reason and plan.

AMI Labs initially plans to develop its world models together with companies from data-intensive industries, such as manufacturing, biomedicine, or robotics. LeCun also sees potential applications in the field of consumer devices. Although Meta is not among the investors, the company is in talks about a collaboration. It is conceivable, for example, that AMI's world models could in the future power AI assistants in Meta's Smart Glasses, says LeCun to Wired.

World models: A lot of basic research still needed

The first official partner is the startup Nabla, which develops AI software for doctors that automatically documents patient conversations and creates medical notes, for example. Unlike large language models, which can tend to hallucinate, AMI Labs' world model is intended to better represent real processes and thus reduce risks in sensitive areas such as medicine.

However, a lot of basic research is still needed until then. LeBrun tells Wired that it could take years for world models to move from theory to commercial applications.

In addition to AMI Labs, the startup World Labs is also developing world models. It also received funding of around one billion US dollars in February. Meta, with its JEPA architecture developed under LeCun's leadership, and Google with Genie 3 are also experimenting with similar approaches.

(vbr)