Yann LeCun spent twelve years building Meta's AI research operation, won the Turing Award in 2018 for foundational work in convolutional neural networks, and has spent the last several years arguing, loudly and repeatedly, that the entire generative AI industry is on the wrong path. Last November he walked into Mark Zuckerberg's office and resigned. On Tuesday, investors gave his new company $1.03 billion to prove it.
AMI Labs — Advanced Machine Intelligence, pronounced like the French word for "friend" — announced its seed round today, just four months after founding. The round values the Paris-based startup at $3.5 billion on a pre-money basis and is reportedly the largest seed round ever raised by a European company. Backers include Nvidia, Bezos Expeditions, Temasek, Mark Cuban, and several European venture firms.
The company has no product, no revenue, and no near-term prospect of either. LeCun acknowledged this week that AMI's first year will be entirely focused on research and development. That's a significant departure from the current AI startup playbook, where speed to market and revenue timelines have compressed to almost absurd levels. AMI is explicitly not that kind of company.
What LeCun is building — and what the billion dollars is funding — is a bet against large language models at the architecture level. His critique is precise: LLMs learn by predicting the next word in a sequence, which produces impressive fluency and systematic hallucination in equal measure. His alternative is JEPA, the Joint Embedding Predictive Architecture, which he proposed in 2022. Rather than generating future states in high detail, JEPA learns abstract representations of how the world works — making predictions in representation space rather than pixel or token space. The goal is AI that can plan, reason about physical environments, and generalize in ways that LLMs structurally cannot.
The operational structure is worth noting. LeCun is executive chairman, not CEO. Day-to-day operations are run by Alexandre LeBrun, the former CEO of Nabla, a medical AI company where the limitations of LLMs had direct clinical consequences. The founding team is drawn almost entirely from Meta's AI research division (FAIR). LeCun has suggested Meta could be one of AMI's first clients — a striking detail that says something about how he left and on what terms.
AMI is building out offices in Paris, New York, Montreal, and Singapore. Its first disclosed development partner is Nabla, with whom it plans to build FDA-certifiable agentic AI systems for healthcare.
Comments