AMD Acquires World Labs for $8.2 Billion to Compete in AI Models

Advanced Micro Devices (AMD) announced on September 28, 2026 that it has entered into a definitive agreement to acquire World Labs, an AI research startup co-founded by computer vision pioneer Fei-Fei Li, for approximately $8.2 billion in an all-stock transaction. The acquisition is expected to close by the end of 2026, subject to regulatory approvals and customary closing conditions.

World Labs, founded in early 2024, specializes in developing spatial-intelligence models that generate, reconstruct, and simulate interactive 3D environments from text, image, and video inputs. The company also develops technology for robotic learning and simulation. Li, a Stanford computer science professor and pioneer in AI research, co-founded World Labs with Ben Mildenhall and Justin Johnson. Following the acquisition, Li will join AMD as executive vice president and chief scientist, reporting directly to AMD CEO Lisa Su, while Mildenhall and Johnson will continue leading the World Labs team.

The companies characterized the acquisition as a natural extension of their existing partnership. AMD and World Labs began a technical collaboration in 2025, focusing on model training and inference optimization on AMD GPUs. Both companies emphasized that advancing next-generation AI requires close integration of model research, hardware systems, and compute infrastructure. AMD said the acquisition will provide deeper insight into how AI workloads are evolving—particularly as AI expands into reasoning, robotics, simulation, and physical systems—and will inform AMD's future hardware and software roadmaps.

World Labs has developed several products aligned with this vision. The company released Atlas, described as an omni-model architecture designed to predict new camera views from 2D images, and previously acquired SceniX to build robotic simulation capabilities. World Labs also released Marble, positioned as a tool for creating entertainment experiences and simulated training environments for robots.

The acquisition positions AMD to compete more directly with Nvidia in the emerging world models space. Nvidia already offers a suite of open-weight world models including Cosmos, while AMD has previously offered primarily text- and video-based models. World models are increasingly viewed as critical infrastructure for deploying AI systems on robotic platforms, including autonomous vehicles and general-purpose humanoids, as they can generate synthetic training data when real-world data is scarce.