AMD is acquiring World Labs, a developer of deep learning models aimed at understanding physical reality, in an $8.2 billion deal announced today. World Labs says AI development now requires close collaboration across model research, systems, and compute; AMD says understanding frontier workloads like World Labs‘ will shape its chip-making roadmap. Fei-Fei Li, World Labs‘ founder, will join AMD as executive vice president and chief scientist. The companies had already formed an inference optimization-and-training partnership last year, and Li was a guest at AMD‘s CES presentation earlier this year.

Li, a Stanford computer science professor and AI pioneer known for building ImageNet and the AI competitions it inspired, founded World Labs in 2024 to develop models with a more robust understanding of the physical world. Her argument was that true general intelligence requires grounding in physics and the ability to reason about data beyond text. In announcing the deal, Li said the partnership came from wanting to scale World Labs‘ technical breakthroughs beyond the lab, and that doing so requires scaling efforts, widening reach, and getting closer to the hardware.

‘World model’ is still loosely defined, covering language models trained to understand visual inputs as well as models that can generate and sustain high-fidelity simulations of reality. World Labs‘ first product, Marble, is for creating entertainment experiences and simulated environments for robot training. The acquisition is likely to help AMD compete with Nvidia in building an ecosystem for AI-specific chips. Nvidia has open-weight world models like Cosmos, while AMD has so far only offered text- and video-based models publicly.

World models are seen as essential for deploying generative AI on robotic platforms, from autonomous vehicles and industrial robots to humanoids. With real-world data for general-purpose robot training scarce, synthetic data from world models is considered key to realizing the approach pursued by companies like Tesla and Figure. The acquisition is expected to close before the end of the year, subject to regulatory approval.