AMD to acquire World Labs in $8.2 billion all-stock deal

AMD announced on 28 September 2026 a definitive agreement to acquire World Labs, the San Francisco-based spatial-intelligence AI lab led by Fei-Fei Li, in an all-stock transaction valued at approximately $8.2 billion. The deal is expected to close by the end of 2026, subject to regulatory approvals and customary closing conditions.techcrunch.com+1

The acquisition matters because it gives AMD direct access to research on models that understand and simulate the physical world, a category the industry is moving toward as it develops beyond chatbots. World Labs builds spatial-intelligence models that generate, reconstruct and simulate interactive three-dimensional environments from text, image and video inputs, with applications in robotics, simulation and design.channelnewsasia.com

What the deal includes

Fei-Fei Li will join AMD as executive vice president and chief scientist after the transaction closes, reporting to CEO Lisa Su. Justin Johnson and Ben Mildenhall, World Labs' co-founders, will continue leading the team as it joins AMD. Li founded World Labs at the beginning of 2024 with Johnson and Mildenhall, guided by the belief that language alone is not sufficient for model development.

Lisa Su, AMD's chair and chief executive, said building compute platforms for the next generation of AI requires a deep understanding of how models are evolving, and that Fei-Fei Li and the World Labs team bring exceptional research leadership and model expertise. Li said joining AMD will give her team the resources and engineering depth to accelerate research.

AMD expects World Labs' research to provide deeper insight into emerging AI workloads and help guide its technology roadmaps across hardware, software and systems. The deal also helps AMD compete with Nvidia, which already offers open-weight world models like Cosmos, while AMD has only offered text- and video-based models to the public.

Why AMD wants world models

World models are seen as vital for deploying generative AI on robotic platforms, from autonomous vehicles to industrial robots and general-purpose humanoids. Because useful real-world data to train general-purpose robots is scarce, synthetic data from world models is expected to be key for companies like Tesla and Figure pursuing that vision.

World Labs raised $1 billion in a funding round earlier in 2026. On 1 September 2026, the lab introduced Atlas, an omni world model pretrained from scratch as a multimodal autoregressive diffusion transformer operating natively on text, images, video and 3D. Atlas handles camera-controlled generation of up to one minute of video at 1440p resolution and spatial reconstruction from one to dozens of input images.

In its own evaluations, World Labs reported that Atlas outperformed recent video models at camera-controlled generation and specialized open-source models at 3D reconstruction. Atlas is entering early access with select partners and will power future versions of Marble, World Labs' first product, pitched as a tool for creating entertainment experiences and simulated environments for robot training.

What Atlas adds to the mix

AMD and World Labs began a deep technical partnership in 2025, starting with model training and inference optimization on AMD GPUs. Li wrote that Su was an early investor in World Labs and that the companies had partnered since the lab's earliest days. Li also wrote that, with the acquisition of SceniX, World Labs is building toward robotics simulation capability.

Li, a Stanford computer science professor, is known for building the ImageNet database and the AI competitions it inspired, and for serving as chief scientist of Google Cloud's AI unit and founding director of Stanford's Human-Centered AI Institute. In a post titled "To Seek a Newer World," she wrote that scaling the lab's efforts requires widening its reach and getting closer to the hardware.

The term "world model" remains loosely defined, covering everything from language models trained to understand visual inputs to models capable of generating and sustaining a high-fidelity simulation of reality. World Labs' research into spatial intelligence differs from AI systems focused on language or two-dimensional images by seeking to model how objects, spaces and actions work in three dimensions. The transaction is expected to close by the end of 2026, subject to regulatory approvals and other customary closing conditions. AMD datelined its announcement from Santa Clara, California, and World Labs announced the same day that it signed a definitive agreement to join AMD.