AMD wants the researchers building 3D world models working alongside the teams designing its AI compute systems.
The chipmaker announced on September 28 that it has agreed to acquire Fei-Fei Li’s World Labs in an all-stock transaction valued at approximately $8.2 billion. AMD says the research team will help shape hardware, software, and systems for emerging workloads such as simulation and robotics. The deal is expected to close by the end of 2026, subject to regulatory approvals and other customary conditions.
The companies already worked together on model training and optimization for AMD GPUs. Bringing World Labs inside AMD could shape future AI systems for simulation and robotics. For enterprise buyers, though, neither company has announced a resulting chip design, deployment date, or measured performance gain.
A partnership that predates the deal
World Labs announced $1 billion in new funding in February, listing AMD among its investors. AMD Ventures discussed its investment in March and said it had first invested in an earlier World Labs Series B round. World Labs says the companies’ technical partnership began in 2025, with work on model training and inference optimization on AMD GPUs.
The acquisition would change that relationship from collaboration to shared ownership of the research. After the transaction closes, Li is set to become AMD’s executive vice president and chief scientist, reporting to CEO Lisa Su. World Labs says Justin Johnson and Ben Mildenhall will continue leading its team alongside Li.
The agreement also follows the World Labs fundraising talks eWeek covered earlier this year. The announced acquisition remains a pending transaction until it closes.
What World Labs brings to AMD
World Labs develops models that generate, reconstruct, and simulate interactive 3D environments from text, images, and video. Its Marble product creates persistent 3D worlds. In September, the company introduced Atlas, a model designed to work across text, images, video, and 3D inputs; World Labs says Atlas is in early access with select partners.
The company is also researching simulations for training and evaluating robots. These tasks may place different demands on compute systems than language models do, which helps explain AMD’s interest in having model researchers inform its infrastructure roadmap. AMD has not said when that research might appear in a commercial product.
What eWeek found: An existing GPU partnership, but no new performance data
AMD and World Labs were already working together on model training and inference optimization for AMD GPUs. AMD described that collaboration in March, and World Labs says the technical partnership began in 2025. The acquisition would bring World Labs’ researchers inside AMD, where the chipmaker says they will help shape future AI hardware, software, and systems.
The gap for buyers is measurable evidence. The acquisition announcement provides no new chip specification, comparative benchmark, customer deployment, or delivery date resulting from the partnership. That makes the deal a clearer signal of AMD’s research direction than of a capability organizations can evaluate today.
AMD has also expanded its AI infrastructure through the Helios rack-scale platform and model optimization work with Anthropic. For teams considering spatial AI or robotics, the useful next evidence will be product details and results from relevant workloads: how World Labs’ models run on AMD systems, what they cost to operate, and how they perform at scale. Those questions remain open while the transaction awaits closing.
Read more: AMD and Cerebras plan to split AI inference across two types of hardware, but pricing and performance in production remain open questions.