Turing Award pioneer Yann LeCun has secured $1 billion in funding for his new startup, AMI, aiming to transcend the limitations of current large language models (LLMs) by developing AI that masters physical world reasoning. The financing round, which values the venture at $3.5 billion, signals a massive industry shift toward “world models”—systems designed to understand reality through observation and logic rather than just text prediction.
Challenging the “Delusion” of Language-Centric AI
LeCun, a foundational figure in modern deep learning, argues that the current industry obsession with scaling LLMs will never result in true human-level intelligence. In his view, human cognition stems from an understanding of the physical environment, a dimension that text-based models like ChatGPT or Claude inherently lack. “The idea that you’re going to extend the capabilities of LLMs to the point that they’re going to have human-level intelligence is complete nonsense,” LeCun stated, characterizing the hype surrounding current generative AI as a “kind of delusion.”
The capital injection for AMI (pronounced like the French word for friend) saw participation from a high-profile syndicate. Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions co-led the round. The startup also attracted individual heavyweights, including Mark Cuban, former Google CEO Eric Schmidt, and telecommunications mogul Xavier Niel.
Strategic Autonomy: Why LeCun Left Meta
AMI marks LeCun’s first major commercial move since departing his role at Meta in November 2025. While he remains a professor at New York University, his decision to build outside the social media giant was driven by a divergence in strategic goals. As Meta pivoted to compete in the LLM arms race, LeCun realized that his research into world models—such as the Joint-Embedding Predictive Architecture (JEPA)—required an enterprise-focused environment rather than a consumer-centric one.
After discussing his vision with Mark Zuckerberg, LeCun opted for independence to iterate “faster, cheaper, and better.” Although Meta is not an official investor, the two entities maintain a collaborative relationship. LeCun indicated that AMI’s sophisticated world models could eventually power hardware applications, such as AI assistants within Meta’s smart glasses.
A Blueprint for Industrial Intelligence
Unlike general-purpose chatbots, AMI targets high-stakes industries including manufacturing, biomedicine, and robotics. The startup intends to build digital twins of complex systems, such as aircraft engines, to optimize efficiency and reliability through predictive modeling. By utilizing vast datasets from these sectors, AMI seeks to create AI that can plan, reason, and maintain persistent memory.
Global Operations and Leadership
The startup launches with a global footprint, establishing offices in Paris, Montreal, Singapore, and New York. The leadership team comprises several Meta veterans, including former director of research science Michael Rabbat, former VP of Europe Laurent Solly, and AI researcher Pascale Fung. Alexandre LeBrun, previously CEO of Nabla, takes the helm as CEO, while former Google DeepMind researcher Saining Xie joins as Chief Science Officer.
Open Source and the Geopolitics of AI Control
LeCun remains a staunch advocate for open-source AI, asserting that the technology is far too consequential to be monopolized by a handful of private corporations. He argues that democratic processes, rather than tech CEOs, should dictate the ethical boundaries of AI application. This stance aligns with growing concerns over AI sovereignty, particularly as governments grapple with the dual-use nature of autonomous systems.
Reflecting on the evolution of neural networks—including the convolutional nets he pioneered—LeCun acknowledges that technology can be exploited by authoritarian regimes. However, he points to the defensive use of autonomous drones in Ukraine as evidence that AI can also protect liberal democracies. “It is not for me to decide what society should do with technology,” LeCun noted, emphasizing the role of the democratic process in governing innovation.
The Road to a Universal World Model
AMI plans to release its initial models rapidly, starting with specialized partnerships involving industrial giants like Toyota and Samsung. These early collaborations will serve as the foundation for LeCun’s ultimate objective: a “universal world model.” This ambitious system would function as a generally intelligent framework capable of assisting companies across any sector, bridging the gap between digital processing and physical reality.
