Analyst(s): Nick Patience
Publication Date: March 13, 2026
Advanced Machine Intelligence Labs (AMI Labs), founded by Turing Award winner Yann LeCun following his departure from Meta, has closed a $1.03 billion seed round at a $3.5 billion pre-money valuation – Europe’s largest seed round on record. The company is building world models based on LeCun’s Joint Embedding Predictive Architecture (JEPA), targeting industrial, robotic, and healthcare applications where the limitations of large language models are most consequential.
What is Covered in This Article:
- AMI Labs has raised $1.03 billion in seed funding at a $3.5 billion pre-money valuation – Europe’s largest seed round – co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions.
- The company is building world models using LeCun’s JEPA framework, designed to learn abstract representations of real-world sensor data rather than predicting outputs token by token.
- The founding team draws heavily from Meta’s AI research organization, with Alexandre LeBrun as CEO, Saining Xie as Chief Science Officer, and Pascale Fung as Chief Research and Innovation Officer.
- AMI’s investor base spans strategic and financial backers across three continents, including NVIDIA, Temasek, Samsung, Toyota Ventures, and individual investors such as Jeff Bezos, Mark Cuban, and Eric Schmidt.
- The company’s near-term commercial horizon is deliberately long – LeCun has indicated the first year will be focused on research, with meaningful product timelines measured in years rather than quarters.
The News: Advanced Machine Intelligence Labs (AMI Labs) announced a $1.03 billion seed round at a $3.5 billion pre-money valuation, the largest seed round ever raised by a European company. Founded in late 2025 by Yann LeCun following his departure from Meta, where he spent more than a decade leading the Facebook AI Research (FAIR) group, AMI Labs is focused on developing world models – AI systems that learn to understand physical reality through sensors and cameras rather than through next-token prediction over text. The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, with additional backing from NVIDIA, Temasek, Samsung, Toyota Ventures, Bpifrance, and individual investors including Jeff Bezos, Mark Cuban, Eric Schmidt, and Tim Berners-Lee. AMI Labs operates across hubs in Paris, New York, Montreal, and Singapore.
Yann LeCun’s AMI Raises $1BN Seed Round – Is the World Model Era Finally Here?
Analyst Take: The AMI Labs seed round is notable on several dimensions simultaneously: the scale of the raise, the stature of the founding team, the breadth of the investor syndicate, and the deliberate positioning against the dominant AI paradigm. It warrants careful analysis on each of those dimensions separately, rather than treating the funding total as a proxy for technical credibility.
The Funding
At $1.03 billion in seed funding, AMI Labs sits in rarefied company. For comparison, Fei-Fei Li’s World Labs raised $230 million at a $1 billion valuation at launch in August 2024, itself considered a large pre-product raise at the time. Former OpenAI CTO Mira Murati’s Thinking Machines Lab was valued at $12 billion in its seed round. The AMI raise is not the largest in absolute terms, but it represents a structurally different profile: a company with roughly a dozen employees, no product, and a research agenda measured in years, not quarters. LeBrun was explicit about this in his public comments, noting that AMI is not a typical applied AI startup with a six-month revenue runway. The valuation of $3.5 billion pre-money implies investors are paying primarily for scientific credibility and long-term option value – a reasonable framing, but one that creates real execution pressure when the next financing round requires evidence of progress.
The investor composition is also worth parsing. The co-lead group – Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions – spans US and European venture capital, with Cathay and Daphni providing French institutional backing consistent with AMI’s positioning as a European counterweight to American and Chinese AI labs. The strategic investors are arguably more significant: NVIDIA’s participation signals alignment with infrastructure; Toyota Ventures and Samsung point toward industrial and device applications; and Temasek’s involvement reflects the Singapore hub and Southeast Asian sovereign capital’s interest in AI infrastructure. The individual backers – Bezos, Cuban, Schmidt, Berners-Lee – add credibility without contributing materially to AMI’s research capacity. This is a carefully constructed syndicate, not a momentum-driven pile-on.
What AMI Is Actually Building
AMI’s technical program centers on the Joint Embedding Predictive Architecture (JEPA) that LeCun has been developing and advocating since 2022. The core insight behind JEPA is that generative models, which attempt to predict the future state of the world in high-dimensional detail (whether pixel-by-pixel or token-by-token), are necessarily imprecise, because much of what happens in the real world is inherently unpredictable at that level of granularity. JEPA instead trains models to make predictions in abstract representation space – learning what matters about how the world changes, rather than attempting to reconstruct its surface appearance. This approach, LeCun argues, is better suited to the demands of robotics, industrial process control, wearables, and healthcare, where decisions must be reliable, and hallucinations carry real costs.
The framing is coherent, and the underlying critique of LLMs is well grounded in empirical evidence. Language models do hallucinate, and the failure modes of autoregressive generation are well-documented. However, the claim that world models represent a categorical solution to those failure modes – rather than a different set of tradeoffs – is more contested. World models have their own generalization challenges, particularly in novel environments, and the path from laboratory JEPA demonstrations to reliable real-world deployment in healthcare or robotics is non-trivial. AMI’s leadership acknowledges this, but the gap between technical promise and commercial application is where many well-funded research labs have historically struggled.
Sovereign AI and the European Positioning
LeCun has been explicit that AMI’s Paris headquarters and European identity are intentional. His framing – that AMI is one of the few frontier AI labs that is neither American nor Chinese – speaks directly to the sovereign AI theme that has become a significant structural force in global AI investment. European governments, sovereign wealth funds, and enterprise buyers are actively looking for AI infrastructure that does not route through US hyperscaler supply chains or expose sensitive data to US cloud jurisdiction. AMI’s multi-hub structure (Paris, New York, Montreal, Singapore) and its investor mix are consistent with a company designed to serve sovereign-sensitive markets.
That said, the sovereign AI framing carries its own complexity. As Futurum has argued in recent research on digital sovereignty, true AI independence is largely aspirational given the depth of supply chain interdependencies: compute, memory, networking, and foundational model components are concentrated in a small number of vendors, most of them American or Taiwanese. NVIDIA’s participation in AMI’s round is a case in point: a European AI lab building world models on NVIDIA silicon is not independent of the US technology ecosystem in any meaningful sense. The more accurate framing is strategic autonomy – the ability to develop and deploy AI systems without being wholly dependent on any single foreign provider – and on that metric, AMI has more credibility than most.
The Team and the Execution Risk
The founding team is a material part of AMI’s investment thesis. LeCun’s scientific credibility is not in question – his work on convolutional neural networks and his long tenure at FAIR represent a genuine track record. The rest of the leadership: LeBrun as CEO (with operational experience from Nabla, a French medical AI startup focused on clinical documentation), Saining Xie as CSO (formerly Google DeepMind), Pascale Fung as Chief Research and Innovation Officer and Mike Rabbat as VP of World Models, represents a concentration of relevant expertise that is genuinely unusual for a company of this age and size.
The risk is not team quality – it is the structural tension between a research-first mandate and investor expectations calibrated to a billion-dollar raise. LeBrun’s public acknowledgment that commercial products may take years to materialize is honest, but it also creates a difficult dynamic as AMI enters subsequent financing rounds. Research labs that have raised at comparable scales – DeepMind before its Google acquisition, for instance – have generally required either a deep-pocketed patron or a defined path to revenue. AMI’s partnership with Nabla for healthcare applications is a sensible early anchor, but healthcare AI development, particularly any pathway to FDA certification, is a long and uncertain process. The company’s stated focus on industrial process control, robotics, and wearables as additional verticals provides diversification, but none of these markets can absorb a world-model-as-a-service offering on a short timeline.
What to Watch:
- Progress on JEPA benchmarks and open publications: AMI has committed to open research and open-source contributions. Watch for whether early publications substantiate the performance claims being made for world models over generative approaches in physical-world tasks.
- The Nabla healthcare partnership: Any movement toward FDA-certifiable AI systems will be a meaningful signal of near-term commercial traction – and a test of whether AMI’s research timelines align with healthcare product development realities.
- Google DeepMind, Meta FAIR (now under Meta Superintelligence Labs), and robotics-focused startups such as Physical Intelligence and 1X are all working on adjacent problems. If the major LLM providers – Google DeepMind, OpenAI, Anthropic, etc. – accelerate their own work on world models and self-supervised spatial learning, AMI’s window of architectural distinctiveness narrows significantly. Watch for whether world model terminology proliferates, as LeBrun predicted, and how AMI differentiates when it does.
- Follow-on financing structure: The $1.03 billion will fund an extended research phase, but Series A terms (when they come) will be the first real market test of whether AMI’s research output translates to commercial credibility.
- European regulatory and policy alignment: AMI’s sovereign AI positioning creates opportunities to work closely with EU institutions and national AI strategies. Watch for government partnerships or public funding relationships that could extend AMI’s runway and validate its European identity.
See the full announcement from AMI Labs on the AMI Labs website.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other Insights from Futurum:
Sovereign AI: What Nations Want (And What They’ll Actually Get) – Report Summary
S3NS & Sovereignty: Can Thales-Google Venture Make AI Sovereignty Work at Scale?
NVIDIA’s European AI Sovereignty Push: Infrastructure, Partnerships, and Policy – Report Summary
Author Information
Nick Patience is VP and Practice Lead for AI Platforms at The Futurum Group. Nick is a thought leader on AI development, deployment, and adoption - an area he has researched for 25 years. Before Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, responsible for 451 Research’s coverage of Data, AI, Analytics, Information Security, and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm that Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
