Artificial intelligence researcher Yann LeCun has launched a new startup focused on a different direction for machine intelligence. The company, Advanced Machine Intelligence Labs (AMI Labs), has raised $1.03 billion in a seed financing round at a $3.5 billion pre-money valuation. The venture is developing “world models,” an AI architecture designed to learn predictive representations of environments through observation of the real world.
LeCun is one of the most recognized figures in artificial intelligence research. He received the 2018 Turing Award alongside Geoffrey Hinton and Yoshua Bengio for contributions to deep learning. Before launching the new company, he served as Chief AI Scientist at Meta from 2013 until his departure earlier this year.
His decision to leave the company followed disagreements with Meta leadership, including CEO Mark Zuckerberg, about the future direction of AI research. LeCun has argued that continued scaling of large language models through increasing computing power will not produce human-level intelligence. His new company focuses on developing world models based on research he previously conducted.
AMI Labs is headquartered in Paris and led by Alexandre LeBrun, who previously served as CEO of the French digital health startup Nabla. LeCun holds the role of Executive Chair. Laurent Solly, formerly Vice President of Meta for Europe, has joined the startup as Chief Operating Officer. Additional members of the leadership team include Saining Xie as Chief Science Officer, Pascale Fung as Chief Research and Innovation Officer, and Michael Rabbat as Vice President of World Models.
The startup began with approximately twelve employees and researchers distributed across four locations: Paris, New York, Montreal, and Singapore. Paris serves as the company’s headquarters. New York connects the organization to LeCun’s academic position at New York University. Montreal is the base for Rabbat, while Singapore provides access to artificial intelligence talent and proximity to clients in Asia. The company describes itself as a global organization and intentionally operates outside Silicon Valley.
AMI Labs focuses on developing world models, a type of artificial intelligence system designed to understand environments through sensor data, visual information, and spatial input. Current large language models learn primarily from text and generate responses by predicting the next word in a sequence. World models instead learn representations of physical environments and predict how those environments evolve.
The research builds on LeCun’s Joint Embedding Predictive Architecture, or JEPA, introduced in 2022. JEPA trains systems to learn abstract representations of the world and to predict future states without generating raw sensory output. The goal is to develop systems capable of planning actions, maintaining persistent memory, reasoning about outcomes, and understanding cause-and-effect relationships in real environments.
LeCun summarized the underlying concept in a statement about the origins of intelligence: “We share one belief: real intelligence does not start in language. It starts in the world.”
LeBrun arrived at a similar position through his experience in healthcare technology. During his time leading Nabla, he encountered the limitations of large language models, particularly the problem of hallucinations in generated text. In medical contexts, incorrect outputs from AI systems can carry serious consequences.
“For anything that requires understanding the real world, we believe that Large Language Models and generative AI in general are not the right solution,” LeBrun said.
AMI Labs plans to apply world models in industries where reliability, controllability, and safety are critical. These fields include healthcare, robotics, industrial process control, automation, wearable devices, and transportation systems. Nabla will become the first partner of AMI Labs and will use the company’s early models in healthcare applications. LeBrun currently serves as chairman of Nabla.
Developing these systems will require long-term research. LeBrun described the company’s approach as different from startups that quickly release products and pursue rapid revenue growth.
“AMI Labs is a very ambitious project, because it starts with fundamental research. 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 also said the company intends to work with potential customers during development because world models require real-world testing with data and evaluation environments.
The seed financing of $1.03 billion, equivalent to about €890 million, is the largest seed round ever raised by a European startup and the second-largest worldwide. The only larger seed round was the $2 billion raised by the U.S. company Thinking Machines Lab in June 2025. The size of the investment places AMI Labs among the most heavily funded AI research startups.
The financing round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, the investment firm of Jeff Bezos. Strategic investors include Nvidia, Samsung, Sea, Temasek of Singapore, Toyota Ventures, SBVA from Seoul, and Alpha Intelligence Capital.
Additional institutional investors include Bpifrance Digital Venture, Aglaé Lab, Artémis, New Legacy Ventures, ZEBOX Ventures, Association Familiale Mulliez, Groupe Industriel Marcel Dassault, and Publicis Groupe. Individual investors include Tim Berners-Lee, Rosemary Berners-Lee, Jim Breyer, Mark Cuban, Mark Leslie, Xavier Niel, and Eric Schmidt.
LeBrun said strong interest from investors gave AMI Labs the ability to choose partners whose expectations matched the company’s research timeline and technical direction. The funding will primarily support computing infrastructure and the recruitment of researchers. According to LeBrun, the company will focus on hiring highly qualified talent instead of rapidly increasing staff numbers.
The investment arrives during growing industry interest in world models. Fei-Fei Li, another well-known AI researcher, recently raised $1 billion for her startup World Labs to develop spatial intelligence systems. A European startup named SpAItial raised a $13 million seed round for similar research.
LeBrun predicted that the concept will soon appear across the AI industry as companies seek funding for research in this area.
“My prediction is that ‘world models’ will be the next buzzword. In six months, every company will call itself a world model to raise funding.”
Although Meta is not listed as an investor in AMI Labs, the companies plan to form a partnership that will grant Meta access to AMI’s technology for commercialization once it is developed. Details of this collaboration are still under discussion.
AMI Labs intends to publish research papers throughout its work and release portions of its software as open source. LeBrun previously worked at Meta’s AI research laboratory, FAIR, where open research played a central role in development.
“We will also make a lot of code open source,” LeBrun said. “We think things move faster when they’re open, and it’s in our best interest to build a community and a research ecosystem around us.”
![how-to-avoid-fraud-when-planning-your-summer-travel-[q&a]](https://thetechbriefs.com/wp-content/uploads/2025/07/36964-how-to-avoid-fraud-when-planning-your-summer-travel-qa.jpg)