Yann LeCun's AMI Labs Secures $1.03B for World Models

Yann LeCun’s AMI Labs Raises $1.03B to Advance World Models

Key Facts

What: AMI Labs, cofounded by Turing Award winner Yann LeCun after departing Meta, raised $1.03 billion in funding at a $3.5 billion pre-money valuation.
Focus: Developing “world models” — AI systems that learn directly from reality rather than language — based on LeCun’s Joint Embedding Predictive Architecture (JEPA) proposed in 2022.
Leadership: Yann LeCun serves as chairman; Alexandre LeBrun is CEO; key hires include Meta’s VP for Europe Laurent Solly as COO, Saining Xie as chief science officer, Pascale Fung as chief research & innovation officer, and Michael Rabbat as VP of world models.
Investors: Co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with participation from NVIDIA, Samsung, Sea, Temasek, Toyota Ventures, Tim and Rosemary Berners-Lee, Jim Breyer, Mark Cuban, Eric Schmidt, Xavier Niel and others.
Timeline: Launched in January 2026; first disclosed partner is digital health startup Nabla.

AMI Labs, the ambitious new AI research venture cofounded by Meta’s former chief AI scientist Yann LeCun, has secured $1.03 billion in funding at a $3.5 billion pre-money valuation to develop world models capable of understanding the physical world. The Paris-headquartered startup, which launched in January 2026, aims to move beyond the limitations of large language models (LLMs) by building systems that learn from reality rather than text. CEO Alexandre LeBrun told TechCrunch the company’s approach, grounded in LeCun’s JEPA framework, represents a fundamentally different path from today’s generative AI race.

World models seek to create AI that can predict, reason about, and interact with the physical environment in ways current language-only systems cannot. LeBrun predicted that “world models” will become the next industry buzzword, noting with a smile that “in six months, every company will call itself a world model to raise funding.” He emphasized that AMI Labs is distinct because its goal is genuine understanding of the real world rather than rebranding existing technology.

The Limitations of LLMs and the JEPA Approach

LeBrun and LeCun both reached similar conclusions about the shortcomings of LLMs while working in different domains. LeBrun, previously CEO of digital health startup Nabla, recognized that hallucinations in language models could have life-threatening consequences in healthcare settings. LeCun, who left Meta to pursue this vision, had long advocated for architectures that move beyond next-token prediction.

The foundation of AMI Labs’ work is the Joint Embedding Predictive Architecture (JEPA), a learning framework LeCun proposed in 2022. Unlike LLMs that predict the next word in a sequence, JEPA trains models to build internal representations of the world and predict future states in an abstract embedding space. This approach is designed to handle the inherent uncertainty and complexity of physical reality.

LeBrun was candid about the timeline. “AMI Labs is a very ambitious project, because it starts with fundamental research,” he said. “It’s not your typical applied AI startup that can release a product in three months, have revenue in six months and make $10 million in [annual recurring revenue] in 12 months.” He acknowledged it could take years for world models to transition from theory to commercial applications.

Competitive Landscape Heats Up

AMI Labs enters a rapidly growing field. Fei-Fei Li’s World Labs secured $1 billion last month, while SpAItial raised a $13 million seed round — considered large for a European startup. According to additional context from industry reports, World Labs has already launched its first product, Marble, which generates physically sound 3D worlds, and is reportedly in talks for fresh funding at a $5 billion valuation.

The space remains smaller than the generative AI sector but is attracting significant capital. AMI Labs’ raise exceeds the €500 million it was reportedly seeking in December 2025, likely due to the strength of its team and LeCun’s reputation. The company’s pedigree includes not only LeCun but also high-profile researchers and executives from Meta and academia.

LeBrun highlighted the company’s ability to be selective with investors due to high interest. The round included both venture firms aligned with the long-term research mission and strategic backers who could become future partners or customers.

Strategic Partnerships and Real-World Focus

AMI Labs’ first disclosed partner is Nabla, the digital health startup where LeBrun now serves as chairman. Healthcare represents a natural early focus area given the risks of LLM hallucinations in medical contexts. Nabla expects to access early world model technology from AMI Labs.

The company plans to engage with prospective customers early despite having no immediate revenue plans. “We are developing world models that seek to understand the world, and you can’t do that locked up in a lab,” LeBrun explained. “At some point, we need to put the model in a real-world situation with real data and real evaluations.”

This philosophy explains the presence of industrial players and potential partners among the investors, including NVIDIA, Samsung, Sea, Temasek, Toyota Ventures, and French organizations such as Association Familiale Mulliez, Groupe Industriel Marcel Dassault, and Publicis Groupe.

Global Talent Strategy and Compute Investment

The $1.03 billion will primarily fund two cost centers: compute and talent. LeBrun said the company will prioritize quality over quantity in hiring. AMI Labs plans to build teams in four key locations: Paris (headquarters), New York (where LeCun teaches at NYU), Montreal (where Rabbat is based), and Singapore (to recruit AI talent and be close to future clients in Asia).

The funding provides meaningful runway for this ambitious research agenda. While exact technical specifications and benchmarks were not disclosed in the announcement, the company’s focus remains on fundamental research into JEPA-based architectures capable of learning world dynamics, predicting future states, and reasoning about physical interactions.

Industry Context and Implications

This raise comes as the AI industry increasingly recognizes the limitations of purely language-based approaches. World models represent a potential next frontier, promising AI systems that can understand causality, physics, and real-world dynamics — capabilities essential for robotics, scientific discovery, autonomous systems, and reliable enterprise applications.

LeCun’s departure from Meta and launch of AMI Labs signals a significant shift. As a Turing Award winner and one of the godfathers of deep learning, his move to build an independent research lab focused on world models carries substantial weight. The $3.5 billion valuation for a company that launched in January 2026 reflects investor confidence in both the technical vision and the team assembled.

Impact on Developers, Users, and the Industry

For developers and researchers, AMI Labs’ work could open new avenues for AI architectures that move beyond scaling laws based on language data. The emphasis on JEPA and world models may influence academic research directions and provide alternatives to dominant LLM paradigms.

Healthcare stands out as an early target application where reliable world understanding could prove transformative. By reducing hallucinations through more grounded representations of reality, world models could enable safer AI assistance in medical diagnosis, treatment planning, and patient monitoring.

The broader industry may see increased competition and capital flow into world model development. LeBrun’s prediction about “world models” becoming the next buzzword appears to be materializing, with multiple well-funded players now pursuing similar goals using different technical approaches.

What’s Next

AMI Labs has no immediate plans to generate revenue, focusing instead on fundamental research and early partner engagements. The company will continue building its global team and investing in compute resources necessary for training world models at scale.

Future deployments will likely begin with controlled partner environments, starting with Nabla in healthcare. As the technology matures, AMI Labs expects to explore applications across multiple industries through strategic partnerships.

The company’s long-term success will depend on translating LeCun’s JEPA vision into practical systems that demonstrate clear advantages over existing approaches. While the timeline remains extended, the substantial funding and elite team position AMI Labs as a major player in what could become the next dominant paradigm in artificial intelligence.

Yann LeCun’s AMI Labs raises $1.03 billion to build world models

Sources

Original Source

Related Topics

Comments