Story
September 30, 2026
AMD’s $8.2 Billion Fei-Fei Li Deal Is a Direct Shot at Nvidia’s Physical-AI Lead
AMD sees World Labs as the research bridge between its compute hardware and the next wave of AI applications, while Fei-Fei Li argues that genuinely useful intelligence must understand the physical world. The wager is also a competitive one: AMD is buying talent and models to challenge Nvidia where robots and simulation are becoming the new battleground.
World Labs began in 2024 with a bold premise: language alone would not be enough to create the next generation of AI. Co-founded by computer-vision pioneer Fei-Fei Li, the startup focused on world models—systems trained to interpret and simulate three-dimensional reality rather than simply predict text.1
Last year, that ambition started moving closer to the hardware. Li said World Labs and AMD formed a “deep technical partnership” around model training and inference optimisation on AMD GPUs, and concluded that joining their software, hardware, foundation-model and application ecosystems was “a natural fit.”2
Now, AMD has agreed to acquire the two-year-old company in an all-stock transaction valued at about $8.2 billion. Li will become AMD’s executive vice president and chief scientist, reporting to CEO Lisa Su, while the World Labs team continues its research.3 The deal is subject to regulatory approval and is expected to close by the end of the year, according to reports.3
AMD’s pitch is not merely to own another AI lab. Su says building next-generation compute platforms requires “a deep understanding of how models are evolving,” and argues World Labs will help shape the hardware, software and systems needed for emerging workloads.4 On X, the company’s welcome message framed the combination as World Labs’ expertise in AI and world models paired with AMD’s compute leadership.
5
The competitive subtext is unmistakable. World Labs’ Marble tool generates persistent 3D spaces from images, video, text and layouts; such models can feed content creation, simulation and synthetic training data for robots. Nvidia already holds a strong position in that physical-AI stack through Cosmos and related robotics efforts, while AMD has lacked a comparable integrated offering.1
Li’s rationale is blunt: “Without having a focused hardware effort, AI is hobbled in efficiency. And scale.”2 AMD is betting $8.2 billion that this is how it stops playing catch-up—and starts defining the machines that understand the real world.