Artificial intelligence (AI) is beginning to move beyond systems that generate text, images and software into machines that must understand and operate in the physical world. AMD now wants a bigger role in providing the technology that makes that possible.
The chipmaker has agreed to acquire World Labs, the spatial intelligence startup led by AI researcher Fei-Fei Li, in an all-stock deal valued at about $8.2 billion, subject to regulatory approvals.
For AMD, the deal provides direct access to AI research focused on understanding three-dimensional environments and physical interactions. The company says that knowledge can help it design future chips, software, and computing systems around the changing demands of AI models.
What AMD Is Buying
World Labs was founded in 2024 by Fei-Fei Li, Justin Johnson, Ben Mildenhall, and Christoph Lassner. Its work focuses on spatial intelligence, which aims to help AI understand objects, places, and interactions across space and time.
Its first product, Marble, can generate persistent 3D environments from text, images, video, and 3D layouts. According to World Labs, this technology is being developed for uses across robotics, simulation, design, scientific discovery, and other applications where AI needs to understand how a physical or virtual environment works.
The chipmaker was already closely involved with the startup, when World Labs raised $1 billion earlier this year from investors including AMD and Nvidia, while AMD and World Labs had also been working together on model training and inference optimisation using AMD GPUs.
And once the acquisition closes, Li will become AMD’s executive vice president and chief scientist and report directly to CEO Lisa Su.
Why Physical AI Matters to AMD
Physical AI requires systems to understand how the world around them behaves before they can make useful decisions inside it. A robot, for example, needs to recognise objects, understand distance and movement, predict what may happen after an action, and adjust when its surroundings change.
World models also create simulated environments where robots and autonomous systems are trained and tested before being deployed in real settings, and Nvidia is already building heavily in this area through its Cosmos world models and Omniverse simulation technology.
As such, bringing World Labs inside AMD gives the chipmaker researchers who are building the models that could eventually require new kinds of computing power. AMD says this will give it deeper insight into emerging AI workloads and help shape its future hardware, software, and systems.
Additionally, World Labs could help AMD combine chips, software, and AI models for robotics more closely, an area where Nvidia has already established its own platform.
The World Labs deal also fits into AMD’s wider effort to build more of the technology surrounding its processors. AMD has previously acquired AI software company Silo AI in 2024 and data centre systems company ZT Systems in 2025. In August, it also agreed to acquire Taalas, whose technology is designed to improve AI inference performance.
