AMD Acquires World Labs for $8.2 Billion to Expand Its AI Model Efforts

AMD turns to World Labs after Nvidia acquires Hugging Face

Hugging Face, the open-weight AI model distribution hub, hosts more than 3 million AI models, over 1 million datasets, and more than 1.4 million interactive spaces. AMD had been focused on the platform as it worked to increase adoption of its Instinct GPU accelerators. However, Nvidia acquired Hugging Face for $12.9 billion earlier this month.

AMD is now taking a more direct role in model development. The company has agreed to acquire World Labs, a startup that develops so-called world models. These systems are intended to understand the physical world rather than generate text and images primarily from the corpus of human knowledge.

A $8.2 billion stock deal

AMD will pay $8.2 billion in stock for World Labs. The startup was founded in January 2024 by Fei-Fei Li, Justin Johnson, Ben Mildenhall, and Christoph Lassner, and was formally established in April 2024. It has raised $1.23 billion across two funding rounds, with angel investors including Jeff Dean of Google, Geoff Hinton of Google, Marc Benioff of Salesforce, Reed Hoffman of LinkedIn, and Eric Schmidt of Sun Microsystems and Google.

Nvidia and AMD participated in World Labs' $1 billion Series B round in February of this year. That round valued the company at approximately $5 billion, meaning AMD is paying a premium of about 75 percent for the team and its world-model technology.

AMD, which recently surpassed a $1 trillion market capitalization, is using stock rather than cash for the acquisition. This preserves cash for investments in chip capacity, materials, and the advance costs associated with manufacturing CPUs and GPUs.

World Labs' products and competitors

World Labs launched its Marble world model publicly in November 2025. In January of this year, it introduced World API, a cloud-access version of Marble. The company also launched early access to Atlas in September. Atlas is described as an “omni world model” and is technically a multimodal autoregressive diffusion transformer. The technology is expected to be incorporated into future versions of Marble.

World Labs operates in an increasingly competitive field. Tencent has HunyanWorld, Google has Genie 3, and AI companies including Runway, Odyssey, and Decart are developing world models. Nvidia is also pursuing the area with its Cosmos world models. Yann LeCun, formerly head of AI research at Meta Platforms, is working on related technology through his AMI Labs startup.

AMD's ambitions may extend beyond World Labs' current focus. The acquisition could support efforts to move AMD compute engines into robotics, an area where Nvidia is already active.

An acquisition centered on the team

The transaction is also a talent acquisition, similar in that respect to AMD's $40 billion purchase of Xilinx. Fei-Fei Li, who was World Labs' chief executive officer, is becoming AMD's chief scientist.

Li holds a bachelor's degree in physics from Princeton University and a PhD in electrical and electronics engineering from Caltech. She served as Google's chief scientist for AI and machine learning beginning in January 2017 and left the company in October 2018 after internal controversy over Google's Project Maven collaboration with the US Department of Defense.

Since 2014, Li has directed the Stanford AI Lab and has served as co-director of the Stanford Institute for Human-Centered Artificial Intelligence. She has also been a professor of computer science at Stanford University for nearly two decades, following earlier positions at Princeton and the University of Illinois.

Justin Johnson earned a bachelor's degree in computer science from Caltech and a PhD from Stanford, where he met Li. He worked as an AI research scientist at Meta Platforms from 2016 through December 2023 and is an adjunct professor of computer science and engineering at the University of Michigan.

Ben Mildenhall earned a bachelor's degree in computer science from Stanford and a PhD in computer science from the University of California at Berkeley. He spent several years as a research scientist at Google, working on computer vision and generative 3D models.

Christoph Lassner studied informatics at the University of Tübingen in Germany. He became an applied scientist at Amazon in September 2017, later moved to Meta Platforms as a research scientist and manager, and spent a year at Epic Games working on 3D worlds for video games.

According to AMD's announcement, Johnson and Mildenhall will continue running World Labs within AMD. Li will report directly to AMD chief executive officer Lisa Su and is expected to take on a broader role than developing world models for robotics. Lassner was not mentioned in the announcement.

AMD has not disclosed World Labs' headcount, revenue, or customer count. The acquisition is expected to bring the company's team into AMD, although the status of every co-founder is not clear from the announcement.

AMD's model strategy

The acquisition gives AMD a path to compete with Nvidia in AI model development. Nvidia has developed open-weight Nemotron generative AI models and the free, open-source Cosmos world model. Cosmos was trained using 20 million hours of video, but that training data is not open.

A central question is whether AMD and Li will develop an open-source generative AI model with publicly available source code, weights, and training data. Such a project would give AMD an opportunity to emphasize openness while expanding its work in world models and other AI systems.