AI pioneer Yann LeCun launches new start-up, seeks $3.5B valuation
Meta's former chief AI scientist targets Paris headquarters for startup pursuing alternative to large language models
Yann LeCun confirmed the launch of Advanced Machine Intelligence (AMI) Labs on December 19, with the startup seeking approximately €500 million ($586 million) at a pre-launch valuation around €3 billion ($3.5 billion). The Turing Award winner and former Meta chief AI scientist will serve as executive chairman while Alex LeBrun, co-founder of medical transcription startup Nabla, assumes the CEO role.
World Models Vision
AMI Labs focuses on developing "world models"—AI systems that understand physics, maintain persistent memory, and plan complex actions rather than simply predicting the next word like current large language models. LeCun has publicly argued that LLMs will never achieve human-level reasoning and planning capabilities, positioning world models as a fundamentally different approach to artificial intelligence.
The technology aims to learn from video and spatial data to develop internal understanding of cause-and-effect relationships in the physical environment. LeCun previously stated such systems may require approximately a decade to mature, indicating a long-term research horizon rather than near-term commercial products.
Strategic Paris Location
AMI Labs will establish headquarters in Paris early in 2025, marking LeCun's return to his hometown and reflecting his criticism of Silicon Valley's current AI approach. "Silicon Valley is completely hypnotized by the current models of generative AI," LeCun explained at the AI-Pulse conference. "To pursue this kind of new research, you have to go outside the Valley—to Paris."
The location choice positions AMI Labs within Europe's growing AI ecosystem while avoiding direct competition with heavily funded Silicon Valley laboratories. Paris offers access to strong engineering talent from European universities, favorable startup policies, and government support for AI research initiatives.
Meta Relationship
LeCun departed Meta after 12 years—five years as founding director of Fundamental AI Research (FAIR) and seven as chief AI scientist. Meta will not provide financial backing for AMI Labs, avoiding potential conflicts of interest with the company's own AI initiatives focused on large language models and commercial products.
However, the companies plan to maintain a collaborative partnership allowing LeCun to leverage Meta's resources without direct investment. This arrangement potentially includes access to computational infrastructure, talent networks, or research collaboration while preserving AMI Labs' independence.
LeCun's exit coincides with Meta's strategic pivot toward more powerful LLM-based models under new chief AI officer Alexandr Wang, the founder of Scale AI. Multiple former employees told media outlets that FAIR has been "dying a slow death" as Meta prioritized commercially focused AI teams over long-term research, with more than half the authors of the original Llama research paper leaving within months of publication.
Funding Landscape
The €500 million fundraising target would represent one of the largest pre-launch raises in AI history, reflecting investor confidence in LeCun's vision and track record. For comparison, Fei-Fei Li's World Labs—also focused on world models—raised over $230 million in 2024 at approximately $1 billion valuation.
LeCun's credentials as a Turing Award winner who developed convolutional neural networks in the late 1980s provide significant credibility with investors. His LeNet architecture revolutionized computer vision, with systems eventually processing 10-20% of all bank checks in the United States during the mid-1990s.
Market Positioning
AMI Labs enters a competitive landscape including Google DeepMind's research efforts and specialized startups pursuing similar world model approaches. However, LeCun's reputation and Meta pedigree position the startup to attract top-tier AI talent in ways most ventures cannot match.
The company faces questions about commercial viability given the decade-long development timeline LeCun projects. While foundation model companies like OpenAI and Anthropic raised tens of billions pursuing near-term products, AMI Labs' research-first approach requires patient capital willing to fund extended R&D cycles without immediate revenue.
Technical Approach
World models aim to address fundamental limitations in current AI systems including hallucinations, lack of causal understanding, inability to plan multi-step actions, and absence of persistent world knowledge. By learning from visual and spatial data rather than text, these systems theoretically develop more robust understanding of physical reality.
The approach contrasts sharply with the scaling hypothesis driving current AI development, which assumes larger models trained on more data will achieve increasingly capable performance. LeCun's bet suggests architectural innovations matter more than pure scale for achieving human-level AI capabilities.
