Yann LeCun, Meta's former chief AI scientist and 2018 Turing Award winner, has launched Advanced Machine Intelligence (AMI), a Paris-based startup that announced a funding round of over $1 billion at a $3.5 billion valuation. The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, with individual backers including former Google CEO Eric Schmidt, Mark Cuban, and French telecom billionaire Xavier Niel. LeCun departed Meta in November 2025 after the company reoriented its AI strategy toward LLM scaling — a direction that diverged sharply from his conviction that world models, not language scaling, are the path to human-level AI.
AMI's core thesis is a direct challenge to the dominant approach pursued by OpenAI, Anthropic, and Meta alike. LeCun argues that large language models are structurally incapable of reaching human-level intelligence because they learn from static text rather than grounded physical experience. "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 told WIRED. Instead, AMI plans to build AI systems with persistent memory, reasoning, and planning capabilities rooted in physical world understanding, building on the <a href="/news/2026-03-14-jepa-v0-pinch-research-self-supervised-audio-encoder-speech-translation">Joint-Embedding Predictive Architecture (JEPA) research</a> LeCun led at Meta's FAIR lab. The startup will release its technology as open source — LeCun's argument being that AI is too consequential to be controlled by any single private entity.
The founding team draws heavily from Meta's AI leadership. Core cofounders include Michael Rabbat (former director of research science at Meta), Pascale Fung (former senior director of AI research), and Laurent Solly (former VP of Europe). Running the company are Alexandre LeBrun, former CEO of AI healthcare startup Nabla, as CEO, and Saining Xie, formerly of Google DeepMind, as chief science officer — the two tasked with translating JEPA research into commercial products. AMI has secured Toyota and Samsung as early enterprise partners, targeting manufacturing, biomedical, and robotics applications. LeCun says he envisions a "universal world model" as the company's long-term goal, though he acknowledges the ambition will take time to demonstrate at scale.
Reaction in the AI research and developer community has been cautious. Some observers note that LeCun had substantial resources at Meta's FAIR lab for years without producing landmark applied results at scale, and question whether a startup structure changes that dynamic. Others point to commercial video generation models like Sora as evidence that video-based world modeling is already advancing without AMI's particular approach. What AMI does have is a genuinely distinct thesis, $1 billion, and a founding team that collectively built much of the AI infrastructure it now aims to supersede. Whether that's enough to crack a problem LeCun couldn't solve with Meta's full resources behind him is the question his new company will spend years answering.