AMD acquires Fei-Fei Li's World Labs

AMD announced on September 28 (U.S. Eastern Time) that it has entered into a definitive agreement to acquire World Labs, the AI model research lab led by Fei-Fei Li. The all-stock transaction is valued at about $8.2 billion. It is expected to close by the end of 2026, subject to regulatory approvals and other customary conditions. Once the deal closes, Fei-Fei Li will become AMD's executive vice president and chief scientist, reporting directly to CEO Lisa Su, and the World Labs team will continue its AI model research.

According to the AMD press release METAL reviewed, the aim of the acquisition is for the company that builds the chips to know in advance where models are heading. AMD said that as AI expands into reasoning, robotics, simulation and physical AI, the demands on computing infrastructure have become more diverse, and that World Labs' model development expertise will help it understand how workloads are evolving and shape its future technology roadmaps. "Building the compute platforms for the next generation of AI requires a deep understanding of how models are evolving," said CEO Lisa Su. "Together, we can use that insight to develop the hardware, software and systems that will power the next generation of AI and strengthen the open AI ecosystem."

Headquartered in San Francisco, World Labs builds spatial-intelligence models, commonly called world models, that generate, reconstruct and simulate interactive 3D environments from text, image and video inputs. It also develops technology for robot learning and simulation. Founded in 2024, the company's newest model, Atlas, builds a virtual replica of a factory or warehouse from a few photos and can even simulate moving objects such as conveyor belts. METAL previously reported on World Labs unveiling Atlas, which generates and reconstructs 3D worlds from a few photos.

The two companies were close well before the acquisition. According to reports, AMD was already an investor in World Labs and took part in a $1 billion funding round seven months ago that valued the startup at $5 billion. Rival Nvidia also joined that round. The $8.2 billion price is about 1.6 times the valuation set a little over half a year ago. Earlier this year, Li and Su appeared on stage together to demonstrate World Labs' model Marble creating a 3D scene from a few images. "Intelligent agents, whether it's robots or vehicles or even tools, can learn inside very rich physics-aware digital worlds before they even need to be deployed into the real one, making them much safer," Li said at the time.

Fei-Fei Li is the researcher who opened the door to deep learning by creating the large-scale image dataset ImageNet two decades ago, and she has been a Stanford professor and led AI research at Google. "Advancing the next generation of AI technology requires close collaboration across model research, systems and compute," she said in the press release. "Joining AMD will give our team the resources and engineering depth to accelerate our research and help define the infrastructure needed for the next era of AI." According to reports, in a lengthy post on X, Li explained that the two companies have had a deep technical partnership since last year focused on model training and inference optimization on AMD GPUs, and wrote that without a focused hardware effort, AI is hobbled in efficiency and scale and remains trapped in the digital world. Li called Su a great friend and an early investor in World Labs.

The size of the deal stands out even in AMD's history. According to reports, this is AMD's second-largest acquisition on record, after Xilinx, which it bought for about $50 billion in 2022. If Xilinx brought AMD the FPGA chips widely used in robots, World Labs brings in models that build the virtual worlds those robots learn in. AMD also recently launched the Ryzen AI Embedded X100 series, robot-oriented chips that combine a CPU, GPU and AI accelerators in a single package. AMD plans to keep World Labs separate from its chip business until the deal closes, and its shares were little changed in after-hours trading following the announcement.

Big Tech's race to secure research talent is also part of the backdrop. According to reports, OpenAI bought Jony Ive's AI devices startup io for about $6.4 billion last year, and Nvidia spent a combined $33 billion on assets of AI chip company Groq and on the open-source platform Hugging Face. METAL has reported on Nvidia's acquisition of Hugging Face. Meta spent $14 billion in 2025 on a stake in Scale AI, bringing in founder Alexandr Wang. In world models, Yann LeCun, Meta's former chief AI scientist, founded AMI Labs in March, and Nvidia is building its own world models while partnering with startups such as self-driving company Wayve.

Read through a journalist's lens, what the chip company bought is not a product but a question about the years ahead. Where large language models learn from text, world models learn from data such as video that captures the physical, three-dimensional world, aiming for AI that understands how objects move, such as the way a ball bounces. According to reports, World Labs' research on next-generation models will let AMD plan years in advance what its AI chips need to be capable of. For AMD, which is chasing Nvidia, it means having a lab in-house that reads demand beyond language models first. "We built the company guided by the foundational belief that language alone isn't sufficient for model development," Li wrote in her post on X. "The universe isn't made up of words, it's made of real things."

The remaining hurdles are regulatory approval and integration. The companies did not disclose how many people work at World Labs, saying only that the whole team will join AMD. With the researcher who taught computers to see through ImageNet now designing, as a chip company's chief scientist, the computation machines need to understand space, the deal signals that world models are moving out of the lab and into the industrial mainstream.