AMD to buy Fei-Fei Li's World Labs for $8.2 billion

Chipmaker AMD is acquiring World Labs — one of the leading developers of deep learning models built to understand physical reality — in a deal worth $8.2 billion, the two companies announced.

World Labs framed the move as a natural step, saying that advancing AI requires close collaboration across model research, systems and compute. AMD, for its part, argues that understanding frontier workloads like those created at World Labs will help shape its chip-making roadmap.

Li joins AMD

World Labs founder Fei-Fei Li will take on the role of executive vice president and chief scientist at AMD. The two firms were already closely tied: they struck an inference-optimization and training partnership a year ago, and Li appeared as a guest at AMD's CES presentation earlier this year.

Li, a Stanford computer science professor, is regarded as a pioneer of AI, especially in computer vision. She is known for building the ImageNet database and the competitions it inspired. In 2024 she founded World Labs to develop models with a more robust grasp of the physical world, arguing that true general intelligence requires a grounding in physics and the ability to reason about data beyond text.

In a post announcing the deal, Li described the partnership as the result of a desire to scale the company's breakthroughs beyond the lab. "Now that we have tangible proof of the possibilities, we want to do everything we can to accelerate the future," she wrote, adding that this means scaling the effort, widening its reach and getting closer to the hardware.

The race for world models

The term "world model" remains loosely defined, covering everything from language models trained to interpret visual inputs to systems that can generate and sustain a high-fidelity simulation of reality. World Labs' first product, Marble, is pitched as a tool for entertainment experiences but also for creating simulated environments to train robots.

The acquisition is likely to help AMD compete with its long-standing rival Nvidia in building an ecosystem for AI-specific chips. Nvidia already offers a suite of open-weight world models such as Cosmos, whereas AMD has so far released only text- and video-based models to the public.

World models are seen as vital to deploying generative AI on robotic platforms — from autonomous vehicles to industrial robots and general-purpose humanoids. With a shortage of useful real-world data for training such machines, synthetic data from world models is expected to be key to the visions pushed by companies like Tesla and Figure.

The deal is expected to close before the end of the year, subject to regulatory approval.

Source: TechCrunch