Yann LeCun Secures $1 Billion to Develop AI Capable of Grasping the Physical Universe

Advanced Machine Intelligence (AMI), a newly established startup in Paris co-founded by Yann LeCun, the former chief AI scientist at Meta, announced on Monday that it has secured over $1 billion to create AI world models.

LeCun contends that human reasoning predominantly originates from our experience with the physical world rather than through language, asserting that AI world models are essential for achieving genuine human-level intelligence. In an interview with WIRED, he stated, “The idea that you’re going to extend the capabilities of LLMs [large language models] to the point that they’re going to have human-level intelligence is complete nonsense.”

The funding round, which places a valuation of $3.5 billion on the startup, was led by several investors, including Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions. Notable supporters also include Mark Cuban, former Google CEO Eric Schmidt, and French billionaire and telecommunications mogul Xavier Niel.

AMI (pronounced like the French term for friend) strives to develop “a new breed of AI systems that comprehend the world, possess persistent memory, can reason and plan, and are safe and controllable,” as stated in a press release. The startup aims to be globally oriented from its inception, with locations in Paris, Montreal, Singapore, and New York, where LeCun will continue his role as a professor at New York University alongside heading the startup. This marks LeCun’s first commercial project since leaving Meta in November 2025.

LeCun’s startup serves as a counterpoint to some of the largest AI laboratories globally, including OpenAI, Anthropic, and his previous employer, Meta, all of which believe that enhancing LLMs will eventually result in AI systems with human-level intelligence or even superintelligence. While LLMs have fueled widely popular products such as ChatGPT and Claude Code, LeCun has been vocal about the limitations of these models. Known for his forthrightness, his skepticism carries significant influence, particularly as a recipient of the Turing Award in 2018 for his groundbreaking contributions to modern AI.

LeCun envisions AMI collaborating with companies in manufacturing, biomedical, robotics, and other sectors rich in data. For instance, he explains that AMI could create an accurate world model of an aircraft engine to assist manufacturers in optimizing efficiency, reducing emissions, or ensuring reliability.

AMI was co-founded by LeCun along with several colleagues from Meta, including Michael Rabbat, the former director of research science; Laurent Solly, the former vice president of Europe; and Pascale Fung, the former senior director of AI research. Additionally, Alexandre LeBrun, the former CEO of the AI healthcare startup Nabla, will take on the role of CEO at AMI, while Saining Xie, a former researcher at Google DeepMind, will serve as the startup’s chief science officer.

The Case for World Models

LeCun acknowledges the overall usefulness of LLMs but believes they are merely the latest in a series of trends within the tech industry, creating what he describes as a “kind of delusion” among their developers. “It’s true that [LLMs] are becoming really good at generating code, and it’s true that they are probably going to become even more useful in various applications where code generation can assist,” remarked LeCun. “That offers a multitude of applications, but it won’t lead to human-level intelligence at all.”

Having explored world models for years at Meta, where he established the Fundamental AI Research lab (FAIR), LeCun has concluded that his research would be better pursued outside the social media giant. He believes that the most viable applications for world models lie in selling them to enterprises, a direction that does not align neatly with Meta’s core consumer-focused business.

With the advancement of AI world models like Meta’s Joint-Embedding Predictive Architecture (JEPA), “there was a shift in Meta’s strategy towards racing to keep up with the industry on LLMs and mimic what other LLM firms are doing, which isn’t aligned with my interests,” LeCun stated. “So, in November, I spoke to Mark Zuckerberg, who has always been supportive of [world model research]. I expressed that I can achieve this more efficiently, at lower costs, and with greater effectiveness outside of Meta. I can also share development costs with other companies … His response was positive, indicating we could collaborate.”