Yann LeCun Leaves Meta, Plans AI Startup

Yann LeCun will leave Meta (META) at the end of 2025 to launch an AI startup focused on world models, highlighting a growing divide as Chief Executive Mark Zuckerberg accelerates commercialization of large language models.

LeCun Departure and Startup Focus

LeCun, Meta’s chief AI scientist and a foundational figure in artificial intelligence, will depart after more than a decade at the company. He plans to found a startup developing world models—AI systems designed to build internal representations of physical dynamics and cause-and-effect relationships. These models aim to enable machines to reason about the physical world rather than relying solely on pattern recognition from text.

Early fundraising discussions for the venture are underway, positioning it to compete with large language model (LLM)–focused approaches at major technology firms. LeCun has long criticized the limits of text-trained language models, saying they cannot grasp the physical world or plan complex actions. He has argued that even a house cat forms better causal judgments. LeCun has also said that creating world models capable of true physical reasoning could take a decade or more.

Meta AI Strategy and Industry Implications

Meta is shifting toward rapid commercialization of LLMs, including its Llama family, under CEO Mark Zuckerberg. This direction diverges from LeCun’s research priorities, which emphasize longer-term development of embodied, human-level intelligence. His departure underscores a broader split in the AI field between fast productization of LLMs and patient research into more advanced cognitive models.

LeCun’s stature and the early fundraising tied to his startup could redirect capital and talent toward long-horizon AI research, intensifying competition with companies focused on quick LLM rollouts. Meta continues to prioritize LLM commercialization without issuing updates to its financial or operational guidance related to LeCun’s exit.