AMD to Buy Fei-Fei Li's World Labs for $8.2 Billion in an All-Stock Bet on Spatial Intelligence

AMD said on Monday it has signed a definitive agreement to acquire World Labs, the spatial-intelligence startup founded by AI pioneer Fei-Fei Li, in an all-stock deal valued at approximately \$8.2 billion. The transaction is expected to close by the end of 2026, subject to regulatory approval and customary closing conditions.

When it closes, Li will join AMD as executive vice president and chief scientist, reporting directly to CEO Lisa Su. World Labs co-founders Justin Johnson and Ben Mildenhall will continue leading the team's model research. Li, who created ImageNet and previously served as chief scientist of Google Cloud AI, quoted Tennyson's "Ulysses" in her announcement post: "Come, my friends, 'tis not too late to seek a newer world."

That newer world is literal. Founded in early 2024, World Labs builds "world models" — AI systems that generate, reconstruct and simulate interactive 3D environments from text, image and video inputs. "The universe is not made of words; it is made of real things," Li wrote, arguing that science, entertainment and robotics all demand models that understand where objects sit, how space is structured, and how a scene changes as the camera — or an agent — moves through it. The company's Atlas research model, trained from scratch on a combination of generative modeling and multi-view geometry, predicts novel camera viewpoints from limited input images and has made progress on sparse reconstruction, a longstanding computer vision problem.

The company is small but formidable: roughly 70 employees, and it raised \$1 billion earlier this year in a round that included an AMD investment. In July it acquired SceniX, a startup whose software trains robots in photorealistic virtual worlds, sharpening the path from 3D scene generation to robot simulation. Late last year it shipped Marble, its first commercial spatial-intelligence product, which generates explorable 3D worlds from images or text prompts.

The strategic logic is workload-shaped rather than talent-shaped alone. Li said Su was an early investor in World Labs and a longtime friend, and that the two teams spent the past year on model training and inference optimization for AMD GPUs. "We began a deep technical partnership with AMD last year, starting with model training and inference optimization on AMD GPUs," Li wrote on LinkedIn. "As our teams worked together, we realized it would be a natural fit to bring together our AI ecosystem of software and hardware, foundation models, and applications."

Su's explanation was blunter: "The more you understand about the end-to-end [stack], the better systems you can build. That's why we're acquiring World Labs." Spatial intelligence, robotics and simulation place very different demands on silicon than chat models do — and AMD wants its hardware architects in the room while those models are still being designed. AMD, which recently touched a \$1 trillion market valuation, is making the same move its larger rival made earlier this month, when Nvidia closed its roughly \$13 billion acquisition of Hugging Face to lock in the text-and-LLM ecosystem. Su is planting AMD's flag in the physical world instead.

Li listed drug discovery, creative production, learning and next-generation robotics as long-term targets, and said World Labs will keep operating as a model research team inside AMD. The deal still has to clear regulators, and spatial intelligence is years from mass deployment. But the message from Austin and San Francisco is the same: chip companies no longer want to wait for workloads to mature before finding out what the hardware needs to be.