AI news ·
AMD acquires World Labs in $8.2 billion deal to strengthen AI chip ecosystem
AMD buys Fei-Fei Li's AI startup World Labs for $8.2 billion, making her its chief scientist. The deal aims to shape future chip designs around AI models that understand physical reality.
AMD is acquiring World Labs, the AI startup led by Stanford professor Fei-Fei Li, in an $8.2 billion deal that places one of the field's most prominent researchers at the chipmaker's helm. The acquisition, announced today, signals a direct bid to shape the hardware that will power the next generation of AI models designed to understand physical reality.
World Labs said the deal reflects a conviction that advancing AI requires "close collaboration across model research, systems and compute." AMD said that understanding frontier workloads like those developed at World Labs will directly influence its future chip designs. The two companies already had a working relationship, having formed an inference optimization-and-training partnership last year. Li was also a guest at AMD's CES presentation earlier this year.
The scientist behind ImageNet joins AMD
Li, a professor of computer science at Stanford, is widely recognized for her work building the ImageNet database and the AI competitions it sparked. That work helped catalyze the modern era of computer vision. In 2024, she founded World Labs to push beyond text-based AI, arguing that true general intelligence requires models grounded in physics and capable of reasoning about the physical world.
Li will join AMD as executive vice president and chief scientist. In a post announcing the deal, she framed the move as a practical step toward scaling the company's research. "Now that we have tangible proof of the possibilities, we want to do everything we can to accelerate the future," Li wrote. "To do this requires scaling our efforts, widening our reach, and getting closer to the hardware."
World models and the race for synthetic data
The term "world model" covers a broad range of AI systems, from language models trained on visual inputs to simulations that maintain a high-fidelity representation of reality. World Labs' first product, Marble, is designed for entertainment experiences and for generating simulated environments to train robots.
World models are increasingly seen as critical for deploying generative AI in physical settings - autonomous vehicles, industrial robots, and general-purpose humanoids. Real-world training data for these platforms remains scarce. Synthetic data generated by world models offers a path around that bottleneck, a point that companies like Tesla and Figure have publicly emphasized.
The acquisition also sharpens AMD's competitive position against Nvidia. Nvidia already offers a suite of open-weight world models under the Cosmos brand. AMD, by contrast, has released only text- and video-based models to the public. Bringing World Labs in-house fills a gap in AMD's AI software ecosystem.
Why this matters for executives and strategy leaders
The deal underscores a structural shift in AI: hardware roadmaps are now being shaped directly by the demands of frontier model research. For executives overseeing technology strategy, the message is that silicon decisions and AI model development are no longer separate planning tracks. Firms building on physical AI - robotics, autonomous systems, spatial computing - will need to watch how AMD's chip designs evolve under Li's influence. The acquisition is expected to close before year-end, pending regulatory approval.