Yann LeCun, one of the most influential figures in modern artificial intelligence, is preparing a return to the startup arena as discussions begin around a new AI venture that could rank among Europe’s most highly valued technology companies at launch. According to people familiar with the matter, LeCun is in early-stage talks to raise approximately €500 million for Advanced Machine Intelligence Labs, a new research-focused company expected to be unveiled in January.
The fundraising would imply a valuation close to €3 billion before the company formally begins operations, though sources caution that terms remain subject to change.
LeCun, a French-American scientist widely regarded as a pioneer of deep learning and a recipient of the Turing Award, recently confirmed plans to step down from his role as Meta’s chief AI scientist at the end of the year. In the new venture, he is expected to serve as executive chair, shaping long-term scientific direction rather than managing daily operations. Alexandre LeBrun, founder of French health-tech company Nabla, has been appointed chief executive officer, according to people briefed on the plans. Neither LeCun nor LeBrun has publicly commented on the funding discussions.
Advanced Machine Intelligence Labs is expected to focus on so-called “world models,” an emerging class of AI systems designed to reason about the physical and digital world through simulation, causality, and continuous learning. The approach contrasts with today’s dominant large language models, which excel at text generation but can struggle with determinism, multimodal reasoning, and reliability in high-stakes environments. LeCun has long argued that progress toward more general and trustworthy AI will require architectures that go beyond probabilistic language prediction.
Nabla
The launch of Advanced Machine Intelligence Labs coincides with a deepening relationship with Nabla, Alexandre LeBrun’s Paris-based company building AI assistants for clinical workflows. Nabla has announced an exclusive strategic partnership with LeCun’s new venture, granting it early access to the company’s developing world model technologies. The partnership is positioned as a step toward enabling more autonomous and regulator-ready AI systems in healthcare, a sector where safety, auditability, and predictability are essential.
Nabla’s leadership argues that while large language models have already transformed clinical documentation and administrative workflows, their inherent limitations make them insufficient on their own for more advanced decision-support or autonomous use cases. Hallucinations, non-deterministic outputs, and difficulty handling continuous data streams such as vital signs or medical imaging present obstacles to regulatory approval and clinician trust. World models, by contrast, are designed to support simulation-based reasoning and “what-if” analysis, potentially enabling AI systems to anticipate outcomes rather than simply generate responses.
LeCun has been involved with Nabla since its early days as an investor and advisor, and the partnership builds on that longstanding relationship. In a statement released by Nabla, LeCun described the company as having recognized early both the promise and the limits of language models in medicine, emphasizing its focus on real-world deployment and safety. Nabla, founded in 2018, began exploring advanced machine learning approaches in healthcare well before the current generative AI boom.
Nabla CEO to Lead AI Research Startup
As part of the broader collaboration, Nabla also announced a planned leadership transition. LeBrun will move from his role as Nabla’s CEO to become CEO of Advanced Machine Intelligence Labs, while remaining closely involved with Nabla as chief AI scientist and chairman. Day-to-day operations at Nabla will continue under co-founder and chief operating officer Delphine Groll, with the board initiating a search for a permanent chief executive. The company says this structure is intended to preserve continuity while strengthening its long-term technical roadmap.
The strategic alignment highlights how foundational AI research and applied industry use cases are becoming increasingly intertwined. By securing early access to next-generation AI architectures, Nabla aims to position itself at the forefront of what it describes as “agentic” healthcare AI – systems capable of acting autonomously within defined constraints while remaining certifiable under regulatory frameworks such as those enforced by the U.S. Food and Drug Administration.
Nabla’s ambitions are backed by significant investor support. The company has raised $120 million from backers including HV Capital, Highland Europe, and Cathay Innovation, and counts figures such as Tony Fadell among its investors. Its leadership team includes medical and product executives with experience in large healthcare organizations, reflecting its focus on enterprise-scale deployment rather than experimental tools.
For LeCun, the new venture represents both a continuation and an evolution of his long-standing critique of current AI paradigms. While large language models have driven rapid commercialization, he has consistently argued that achieving more general, robust intelligence will require new architectural foundations. With Advanced Machine Intelligence Labs, he appears to be betting that world models could provide that foundation – and that healthcare, with its stringent demands, may become one of the first domains where their advantages are fully realized.
Executive Insights FAQ
What is Advanced Machine Intelligence Labs aiming to build?
The company is expected to focus on world model AI systems that emphasize simulation, causality, and deterministic reasoning beyond current language models.
Why is Yann LeCun leaving Meta to start a new venture?
LeCun has indicated a desire to pursue foundational AI research independently, particularly architectures he believes are necessary for the next phase of AI development.
How does the partnership benefit Nabla?
Nabla gains early access to emerging world model technologies, supporting its goal of developing regulator-ready, agentic AI systems for healthcare.
What are world models, and why do they matter?
World models aim to represent and reason about real-world dynamics, enabling more reliable decision-making and simulation than text-focused models alone.
Why is healthcare a key target for this collaboration?
Healthcare demands high reliability, auditability, and safety, making it a proving ground for advanced AI systems that must meet strict regulatory standards.
