AMD has signed a definitive agreement to acquire World Labs, a relatively new AI research firm, for $8.2 billion, with the entire payment made in shares. The deal is expected to be completed by the end of 2026, pending regulatory approvals and other standard conditions. World Labs was founded in 2024 and is led by Fei-Fei Li, a Stanford University computer science professor known for her role in developing the ImageNet database, a large-scale image recognition project. Li will become AMD's executive vice president and chief scientific officer, reporting directly to CEO Lisa Su. World Labs specializes in developing "spatial intelligence" models, which generate, reconstruct, and simulate interactive 3D environments using textual, image, and video data. Its first commercial product, Marble, allows users to create modifiable 3D environments from text, images, or 360-degree panoramas, with applications in entertainment and robot training. The company’s latest model, Atlas, can generate virtual replicas of factories or warehouses from just a few photos and simulate moving objects like conveyors. This acquisition follows an earlier partnership between AMD and World Labs, focused on optimizing AI model training and performance on AMD's graphics processing units (GPUs). In February 2026, World Labs raised $1 billion in new funding, giving it a valuation of $5 billion, with AMD and NVIDIA among the investors. This is the second-largest acquisition in AMD’s history, after its $50 billion purchase of Xilinx in 2022. AMD said the acquisition is part of its broader strategy to build computing platforms for the next generation of AI, which requires a deep understanding of how AI models evolve. Fei-Fei Li emphasized the importance of integrating model research with systems and computing to advance AI infrastructure. She noted that joining AMD would provide the necessary resources and engineering expertise to accelerate research and define the infrastructure for future AI development. The acquisition aims to help AMD anticipate future needs for its chips, especially in areas like robotics and autonomous driving, where synthetic data—computer-generated data used to train AI systems—is essential. NVIDIA, AMD’s main competitor in the AI chip market, has its own world models, such as the Cosmos family, designed for physical AI applications. By acquiring World Labs, AMD hopes to gain insights into the development of such models and provide tools for developers to use them on AMD platforms. World Labs will remain an independent entity until the acquisition is finalized, after which the team will continue its research into AI models. The move is also seen as a way for AMD to strengthen its position in the AI ecosystem, complementing recent efforts in AI infrastructure, including the Helios server racks and MI400 series GPUs, supported by major tech companies like OpenAI, Meta, and Anthropic.