Physical AI & Robotics
AMD Acquires World Labs for $8.2 Billion, Pushing Into Physical AI and Intelligent Robotics
AMD has announced it is acquiring World Labs, the world models and spatial intelligence company founded by Fei-Fei Li, in an all-stock deal worth approximately $8.2 billion. The deal reflects AMD's shift from being purely an AI chipmaker toward building infrastructure for next-generation AI, including robotics, simulation, and Physical AI.
AMD Acquires World Labs for $8.2 Billion
On September 28, 2026, AMD announced an agreement to acquire World Labs, an AI company focused on spatial intelligence and world models, in a deal worth approximately $8.2 billion, paid entirely in AMD stock.
AMD says it expects the acquisition to close by the end of 2026, subject to regulatory approval and customary closing conditions.
The acquisition matters for more than just adding another AI company to AMD's portfolio. World Labs specializes in AI models that can understand and simulate three-dimensional environments — technology directly tied to the growing trend of Physical AI, or AI that can perceive and interact with the real world.
AMD says that as AI expands beyond generating text and images into reasoning, simulation, robotics, and Physical AI, compute infrastructure needs will become far more diverse. Acquiring World Labs helps AMD understand where AI model development is heading, and apply that understanding to shape its future hardware, software, and systems direction.
Who Is World Labs, and Why Does AMD Care?
World Labs was founded in 2024 by Fei-Fei Li, a well-known AI researcher in the computer vision field and one of the key figures behind ImageNet, the large-scale image dataset that played a major role in early computer vision and deep learning development.
World Labs' goal is to develop AI with spatial intelligence — the ability to understand space, position, objects, and the relationships between things within an environment.
This differs from AI that works with text alone, since it requires understanding "where things are," "what shape they have," and "what happens if something moves or the environment changes."
This technology can be applied across many industries — from simulation, computer vision, gaming, and visual effects to robotics and AI systems that need to work in the real world.
What Are World Models?
One of World Labs' core technologies is world models.
A world model can be described simply as an AI model that tries to build a "model of the world," letting AI learn from and predict what will happen within that environment.
For example, if you want to train a robot to work in a factory, instead of repeatedly testing a physical robot — which is expensive and risks damage — developers can build a virtual factory and let AI practice within a simulation. AI can learn from repeated trials such as:
- Picking up and placing objects
- Moving through confined spaces
- Avoiding obstacles
- Working alongside machinery
- Responding to moving objects
- Following sequential steps within a factory
Once AI has learned enough from simulation, the trained model can be applied to real-world systems. This approach reduces the cost and risk of training AI in the real world, and is one reason world models matter so much for robotics and Physical AI development.
World Labs' Atlas: Building 3D Worlds From Images
One of World Labs' most notable technologies is Atlas, a multimodal world model that can build and simulate 3D environments from input data.
Atlas can take image data and construct virtual environments viewable from multiple angles, complete with spatial and depth information that helps AI better understand where objects are positioned within a space.
One notable use case is building a virtual factory or warehouse from just a handful of images. Once a virtual environment exists, developers can use it for simulation and to train AI that controls robots.
Atlas can also simulate moving elements such as conveyor belts or objects moving through a space, making simulations closer to real-world conditions. Atlas's technology isn't limited to generating 3D imagery — it can serve as a foundation for building environments AI uses to learn about the world.
Why World Models Matter for Robotics
Robotics is one of the markets positioned to benefit directly from world models.
Modern robot development isn't just about hardware design — it also requires AI that can control a robot to perceive its environment and make appropriate decisions. The core challenge is that letting a robot learn from the real world can be slow, expensive, and risky for the hardware.
Simulation plays a critical role here, since it lets AI test large numbers of scenarios simultaneously. If a world model can build environments that are realistic and understand spatial relationships well, it helps developers generate training data for AI much faster. This is one area where AMD can combine its compute expertise with World Labs' AI modeling capability.
AMD Is Thinking Beyond GPUs, Toward AI Infrastructure
The World Labs acquisition also reflects AMD's broader strategic direction. AMD is a major player in CPUs, GPUs, AI accelerators, and data center compute systems, but competition in the AI market is shifting from a pure focus on chip performance toward building full-stack AI infrastructure.
Next-generation AI needs:
Understanding how future AI models will actually work is therefore critical to hardware design. AMD says World Labs' modeling expertise will help the company understand where AI workloads are heading, and apply that understanding to shape AMD's technology roadmap.
AMD Already Has a Foundation in Robotics
The World Labs acquisition doesn't come out of nowhere — AMD already has businesses connected to robotics.
In 2022, AMD acquired Xilinx for approximately $50 billion. Xilinx specializes in FPGAs (Field-Programmable Gate Arrays) — chips that can be customized for specific tasks and are widely used in industrial and robotics systems.
AMD also has its Ryzen AI Embedded X100 Series, designed for embedded and robotics applications, combining CPU, GPU, and AI accelerators into a single platform built for more demanding operating environments.
Combining these capabilities with World Labs' world models gives AMD the opportunity to connect the full chain — from AI models to compute hardware — more tightly than before.
The Deal Follows an Earlier Investment in World Labs
This acquisition builds on an existing relationship between AMD and World Labs.
AMD had previously participated in a $1 billion funding round for World Labs in early 2026, a round that also included investment from Nvidia. At the time, World Labs was valued at approximately $5 billion, before AMD later announced its acquisition agreement.
Going from investor to acquirer in under a year shows just how much value AMD places on World Labs' AI modeling and spatial intelligence capabilities.
Fei-Fei Li to Join AMD as Chief Scientist
Another key part of this deal is the role of Fei-Fei Li.
Once the acquisition closes, Li will join AMD as Executive Vice President and Chief Scientist, reporting directly to AMD CEO Lisa Su. Meanwhile, the World Labs team will continue its AI model research and development.
Gaining not just the company and its research team but also a leading AI model researcher like Li carries strategic weight, since the next phase of AI competition won't be determined by hardware alone — it will also depend on the ability to understand where AI models are heading and design compute systems that fit those models.
From Generative AI to Physical AI
In the early days of the generative AI wave, most users became familiar with AI that could generate text, images, audio, or video. But the next direction is shifting toward Physical AI.
Physical AI is the idea that AI doesn't just operate in the digital world — it can perceive, understand, and respond to the real environment through devices such as robots, vehicles, automated systems, and industrial machinery.
For Physical AI, understanding the world in 3D matters far more than AI that only understands text or two-dimensional images. AI needs to know where objects are, how far apart they are, how they're moving, and how the environment will change if a given action occurs. This is exactly where world models and spatial intelligence come in.
What the AMD-World Labs Deal Signals for the AI Industry
The $8.2 billion acquisition of World Labs shows that the competitive landscape of the AI market is expanding.
Competition is no longer limited to building large language models or manufacturing high-performance GPUs — it's extending toward models that can understand the real world, build simulations, and use AI to control physical systems.
For AMD, combining World Labs' expertise with its CPUs, GPUs, AI accelerators, networking, and embedded technology could help the company position itself more comprehensively in the AI infrastructure market. On the other side, World Labs will gain engineering and compute resources from AMD to continue developing its AI models.
The deal therefore brings together AI model research, compute hardware, software, simulation, and robotics in one package.
Summary
AMD has announced its acquisition of World Labs for approximately $8.2 billion in an all-stock deal, expected to close by the end of 2026 pending regulatory approval.
At the heart of the deal is bringing World Labs' expertise in world models and spatial intelligence into AMD's AI infrastructure. World Labs' technology can build and simulate 3D environments and has real potential across robotics, simulation, and Physical AI, while AMD brings the hardware and compute infrastructure to support these workloads.
Once the deal closes, Fei-Fei Li will join AMD as Executive Vice President and Chief Scientist, and the World Labs team will continue its AI research under AMD.
This deal is another important signal of the shift from AI that operates purely in the digital world to AI that can understand, simulate, and interact with the physical world — a foundation that will underpin robotics and Physical AI going forward.
