AMD acquires World Labs AI from Fei-Fei Li for $8B

AMD acquires World Labs AI from Fei-Fei Li for $8B

The chipmaker is betting billions on spatial intelligence to compete in the AI arms race beyond traditional GPUs

AMD just wrote an $8.2 billion check for World Labs, the AI startup founded by Stanford professor and AI pioneer Fei-Fei Li. The deal, announced on September 28, represents one of AMD’s largest acquisitions ever and a clear signal that the company sees its future well beyond selling graphics cards to gamers.

World Labs specializes in what the industry calls “spatial intelligence,” building AI models that can generate, perceive, and reason about 3D environments.

What AMD is actually buying

World Labs was founded in 2024 by Li alongside co-founders Justin Johnson, Christoph Lassner, and Ben Mildenhall, all of whom bring deep expertise in computer vision and 3D reconstruction.

The startup raised a $230 million seed round in September 2024, then followed that up with a massive $1 billion equity round in February 2026. AMD actually participated in that round alongside Nvidia and Autodesk, meaning it got a close look at World Labs’ technology before deciding to buy the whole thing.

Advertisement

Total funding raised before the acquisition came to roughly $1.23 billion. AMD is now paying between $8 billion and $8.2 billion to take full ownership. Specific deal terms beyond the headline number remain undisclosed.

On the product side, World Labs has two key offerings. The Marble model launched in November 2025, followed by the Atlas model in September 2026. Both target applications in gaming, robotics, and virtual reality.

World Labs also picked up SceniX, a robotics-simulation startup, in July 2026. That acquisition added capabilities in synthetic training environments for robots.

AMD’s bigger AI play

This deal doesn’t exist in isolation. AMD acquired Taalas, an inference technology company, in August 2026. Combined with World Labs, the two acquisitions paint a picture of a chipmaker trying to build an end-to-end AI platform rather than simply selling silicon.

Why spatial intelligence matters now

Fei-Fei Li’s academic work on computer vision, including the ImageNet dataset that helped spark the deep learning revolution, positioned her uniquely to understand that AI’s next frontier wasn’t just reading text but seeing and understanding the world.

World Labs’ Marble and Atlas models represent some of the most advanced publicly known work in this space. Marble handles 3D scene generation and understanding, while Atlas focuses on interactive reasoning within those environments.

What investors should watch

The competitive landscape includes Nvidia’s Omniverse platform for 3D simulation, Google DeepMind’s research on world models, and Meta’s investments in 3D understanding.

With the World Labs and Taalas deals, AMD has spent well over $8 billion on AI software and model companies in the span of two months.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
AMD acquires World Labs AI from Fei-Fei Li for $8B
AMD acquires World Labs AI from Fei-Fei Li for $8B

The chipmaker is betting billions on spatial intelligence to compete in the AI arms race beyond traditional GPUs

AMD just wrote an $8.2 billion check for World Labs, the AI startup founded by Stanford professor and AI pioneer Fei-Fei Li. The deal, announced on September 28, represents one of AMD’s largest acquisitions ever and a clear signal that the company sees its future well beyond selling graphics cards to gamers.

World Labs specializes in what the industry calls “spatial intelligence,” building AI models that can generate, perceive, and reason about 3D environments.

What AMD is actually buying

World Labs was founded in 2024 by Li alongside co-founders Justin Johnson, Christoph Lassner, and Ben Mildenhall, all of whom bring deep expertise in computer vision and 3D reconstruction.

The startup raised a $230 million seed round in September 2024, then followed that up with a massive $1 billion equity round in February 2026. AMD actually participated in that round alongside Nvidia and Autodesk, meaning it got a close look at World Labs’ technology before deciding to buy the whole thing.

Advertisement

Total funding raised before the acquisition came to roughly $1.23 billion. AMD is now paying between $8 billion and $8.2 billion to take full ownership. Specific deal terms beyond the headline number remain undisclosed.

On the product side, World Labs has two key offerings. The Marble model launched in November 2025, followed by the Atlas model in September 2026. Both target applications in gaming, robotics, and virtual reality.

World Labs also picked up SceniX, a robotics-simulation startup, in July 2026. That acquisition added capabilities in synthetic training environments for robots.

AMD’s bigger AI play

This deal doesn’t exist in isolation. AMD acquired Taalas, an inference technology company, in August 2026. Combined with World Labs, the two acquisitions paint a picture of a chipmaker trying to build an end-to-end AI platform rather than simply selling silicon.

Why spatial intelligence matters now

Fei-Fei Li’s academic work on computer vision, including the ImageNet dataset that helped spark the deep learning revolution, positioned her uniquely to understand that AI’s next frontier wasn’t just reading text but seeing and understanding the world.

World Labs’ Marble and Atlas models represent some of the most advanced publicly known work in this space. Marble handles 3D scene generation and understanding, while Atlas focuses on interactive reasoning within those environments.

What investors should watch

The competitive landscape includes Nvidia’s Omniverse platform for 3D simulation, Google DeepMind’s research on world models, and Meta’s investments in 3D understanding.

With the World Labs and Taalas deals, AMD has spent well over $8 billion on AI software and model companies in the span of two months.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.