AMI Labs company profile: valuation, funding & revenue

This AMI Labs company profile covers the company's valuation, revenue, funding rounds, investors and history. AMI Labs works in AI, with a focus on world models research. It was founded in 2025 and is headquartered in Paris, France. Its latest valuation on record is US$3.5 billion pre-money (March 2026). Every figure links to a public source, and anything the company hasn't disclosed is marked as such.

Revenue

Funding history

AMI Labs timeline

Key dates from the sources on this page, oldest first. Metric snapshots are listed after the funding events.

• 2025 — AMI Labs is founded in Paris by Yann LeCun (executive chairman), with Alexandre LeBrun (CEO).
• Mar 2026 — Seed: US$1.03B, valuing the company at US$3.5B pre-money. Investors: Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions (co-leads).

Investors

Overview

AMI Labs, short for Advanced Machine Intelligence Labs, is an artificial intelligence research company founded by Yann LeCun, one of the pioneers of deep learning and a winner of the Turing Award, computing's highest honour. LeCun left Meta, where he was chief AI scientist, in late 2025 to start AMI Labs. The company is headquartered in Paris, with offices reported in New York, Montreal and Singapore. LeCun serves as executive chairman, and the chief executive is Alexandre LeBrun, a French AI entrepreneur who previously co-founded and led the medical AI company Nabla.

AMI Labs is building "world models": AI systems that learn how the physical world works from video and other sensory data, can remember, reason and plan, and do not rely only on predicting the next word in a text, as large language models do. In March 2026 it raised a US$1.03 billion seed round at a US$3.5 billion pre-money valuation, co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, according to Bloomberg Law. It is one of the largest seed rounds ever and one of the largest financings for a European AI company.

What AMI Labs is building

• World models: AI that builds internal representations of how the world behaves, so it can predict the consequences of actions.
• Joint-embedding architectures: research building on LeCun's JEPA (Joint Embedding Predictive Architecture) ideas, which learn abstract representations rather than predicting every pixel or word.
• Planning and reasoning: systems that can plan sequences of actions to achieve goals.
• Applications: potential uses in robotics, wearable devices, industrial automation, healthcare and assistants that understand the physical world.

Funding

LeCun's plans attracted investor interest before the company officially launched, with reports in late 2025 of talks at a valuation of several billion dollars. In March 2026, Bloomberg Law reported that AMI Labs had raised US$1.03 billion in seed funding co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, at a pre-money valuation of US$3.5 billion. A seed round of this size reflects LeCun's reputation and the high cost of AI research at scale.

Business model

AMI Labs has said it will focus on research first. Over time it could license world models to companies building robots, devices and industrial systems, offer models through APIs, or partner with companies in specific sectors. LeCun has been a strong advocate of open research, and AMI Labs is expected to publish research and possibly release open models, as Meta did with its Llama models.

Competition

AMI Labs competes with other world-model efforts, including World Labs, founded by Fei-Fei Li, Google DeepMind's Genie models, and robotics AI companies such as Physical Intelligence and Skild AI. It also competes for talent with large AI labs and newer labs such as Humans& and Thinking Machines Lab. In Europe, it joins Mistral AI as one of the continent's most prominent AI companies.

AMI Labs at a glance

AMI Labs is a Paris AI research company founded in 2025 by Yann LeCun after he left Meta. Its chief executive is Alexandre LeBrun. It builds world models that learn how the physical world works. It raised a US$1.03 billion seed round at a US$3.5 billion pre-money valuation in March 2026, co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions.

Timeline

• Late 2025: Yann LeCun announces his departure from Meta and plans for a new company.
• December 2025: AMI Labs is formally established, with Paris as headquarters.
• March 2026: raises US$1.03 billion in seed funding at US$3.5 billion pre-money.

Why LeCun favours world models

LeCun has argued for years that large language models, while impressive, are limited because they learn mainly from text. Humans and animals, by contrast, learn largely by observing the world and interacting with it: a child learns that objects fall when dropped long before learning language. LeCun believes that AI capable of human-level intelligence will need to learn in a similar way, building models of how the world works from sensory data, and using those models to reason and plan. This view puts him at odds with many in the industry who expect scaling language models to keep delivering progress. AMI Labs is his attempt to prove his approach.

What JEPA means

Many generative AI systems learn by predicting the next word, pixel or frame. LeCun's JEPA approach instead learns to predict abstract representations: rather than trying to predict every detail of the next video frame, the system predicts a compressed description of what will happen. This lets it ignore unpredictable details, such as the exact movement of leaves in the wind, and focus on what matters, such as where a moving car will be. Meta published several JEPA research models while LeCun was there. AMI Labs aims to scale and extend this work.

Paris as an AI hub

LeCun is French, and AMI Labs' choice of Paris as headquarters strengthens the city's position as Europe's leading AI hub, alongside companies such as Mistral AI and research labs run by Meta, Google and others. France has encouraged AI investment through government programmes and has a strong supply of mathematicians and engineers. Our guide to the largest French startups covers the wider ecosystem.

Who funds AMI Labs

AMI Labs' seed round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions.

Valuation context

A US$3.5 billion pre-money valuation for a company that had just been formed reflects investors' confidence in its founder and the scale of the opportunity. Our guides to new AI labs and the most valuable AI startups compare AMI Labs with peers.

Risks

• Research uncertainty: world models may take years to deliver commercial value.
• Competition: large labs are also investing heavily in world models and robotics AI.
• Commercial model: the path to revenue is not yet defined.
• Capital needs: training on video data at scale is expensive.

Questions to ask as it develops

• Which applications will AMI Labs target first?
• Will its models be released openly?
• How will it measure progress compared with language models?
• What partnerships will it form with robotics and device makers?

Where world models could be used

A robot in a warehouse needs to understand how boxes stack, how objects slide and what happens if it pushes something. A self-driving car needs to predict how pedestrians and other vehicles will move. Smart glasses that assist users need to understand what they are looking at and what they are trying to do. In each case, an AI that has learned a model of how the physical world behaves could plan better and make fewer mistakes than one trained mainly on text. World models could also help in industrial settings, such as predicting equipment behaviour, and in healthcare, where understanding sequences of events matters. These applications are why investors see world models as a potential next frontier after language models.

What to watch

Watch for research releases, open models, first commercial partnerships and further funding.

Takeaways

Researchers will watch whether world models can rival language models. Founders can see how a renowned scientist can raise large capital at launch. Investors should weigh the founder's track record against long research timelines.

Sources

Every figure on this page comes from the public sources below. Where sources disagreed, we used the company's own announcement or a primary news report. Private-company data changes; check the date beside each figure.

1. Bloomberg Law