AMD has agreed to buy World Labs for $8.2 billion.
The two firms announced the deal today. World Labs builds deep learning systems designed to interpret physical reality. AMD stated that future chip designs must account for the heavy demands of such frontier workloads.
Fei-Fei Li, the founder of World Labs, will become executive vice president and chief scientist at AMD. The companies formed an inference optimisation partnership last year. Li appeared as a guest at AMD’s CES presentation earlier this year.
Li is a professor at Stanford University. She is widely known for creating the ImageNet database and the competitions that followed. In 2024 she launched World Labs. Her goal was to build models that understand the physical world, not just text. She argued that true general intelligence requires grounding in physics.
Li explained that the acquisition allows them to move technical breakthroughs beyond the laboratory. “Now that we have tangible proof of the possibilities, we want to do everything we can to accelerate the future,” she wrote. “To do this requires scaling our efforts, widening our reach, and getting closer to the hardware.”
The term “world model” covers various approaches. These range from language models processing visual inputs to systems generating high-fidelity simulations of reality. World Labs’ first product, Marble, targets entertainment experiences and simulated environments for robot training.
This move helps AMD challenge Nvidia in the AI chip market. Nvidia already offers open-weight world models like Cosmos. AMD has previously released only text- and video-based models to the public.
World models are essential for deploying generative AI on robotic platforms. This includes autonomous vehicles, industrial robots, and general-purpose humanoids. A lack of useful real-world data for training general robots means synthetic data from world models will be key. This supports the visions of companies like Tesla and Figure.
The acquisition is expected to close before the end of the year, pending regulatory approval.
What it means
Developers building robots or simulating environments now have access to AMD’s hardware alongside World Labs’ simulation tools. The combination aims to provide the synthetic data necessary to train machines that can operate safely in the real world.
