Semiconductor designer AMD announced a landmark acquisition in artificial intelligence on September 28, 2026, agreeing to buy startup World Labs for approximately 8.2 billion US dollars in an all-stock transaction. The acquisition represents an aggressive strategic push into spatial intelligence and physical AI. AMD intends to link cutting-edge world models directly to its enterprise hardware portfolio. This move signals a broader effort to expand beyond pure language modeling into physical machine perception.
A major leadership transition lies at the heart of the transaction. Dr. Fei-Fei Li, renowned Stanford professor and co-creator of ImageNet, joins AMD as Executive Vice President and Chief Scientist. She will report directly to AMD Chief Executive Officer Lisa Su. Co-founders Justin Johnson and Ben Mildenhall are also joining AMD as part of the deal, securing one of the most prominent research teams in computer vision for the chipmaker.
World Labs focuses on building spatial intelligence and foundational world models designed to help machines comprehend physical geometry. In early September 2026, the company introduced Atlas, a foundational model built to solve the complex challenge of sparse reconstruction. Atlas can generate continuous, photorealistic three-dimensional environments from merely a handful of still images or brief video frames.
Under the hood, World Labs relies on 3D Gaussian Splatting and dense point clouds to synthesize persistent, navigable digital spaces. This capability is considered essential for real-to-sim pipelines, where autonomous robots train inside high-fidelity simulated worlds before deploying into physical environments. By bringing this technology in-house, AMD positions itself at the core of embodied robotics workflows.
The purchase directly targets Nvidia's dominance across high-performance computing and robotics development. AMD plans to natively integrate World Labs' model architectures into its open-source ROCm software stack and its Instinct accelerator hardware line. This tight vertical integration between algorithms and custom silicon could provide developers with an attractive alternative to Nvidia's proprietary graphics and simulation platforms.
The transaction reflects a decisive industry shift away from purely digital text generation toward embodied artificial intelligence. Lisa Su's multi-billion-dollar bet demonstrates that the next frontier of hardware demand will be shaped by how effectively chips run spatial simulations. By securing top-tier talent and proprietary 3D foundational models, AMD establishes a serious foundation for the next decade of physical computing.