AMD agrees to buy World Labs for $8.2 billion in an all-stock transaction that would bring Fei-Fei Li’s spatial-intelligence research group inside the chipmaker. The definitive agreement pushes AMD beyond supplying processors and software into ownership of a lab building models that generate, reconstruct and simulate interactive three-dimensional environments.

 

The companies expect the acquisition to close by the end of 2026, subject to regulatory approvals and customary conditions. Li would become AMD’s executive vice president and chief scientist, reporting to chief executive Lisa Su, while the World Labs team would continue its model research.

 

The agreement establishes four headline terms:

  • Approximately $8.2 billion in AMD stock
  • Targeted closing by the end of 2026
  • Fei-Fei Li as AMD chief scientist
  • Continued World Labs model research

 

AMD Agrees to Buy World Labs for $8.2 Billion

AMD described the transaction as a definitive agreement rather than a completed purchase. The distinction matters because shareholder value will depend on AMD’s share price and because the closing still requires regulatory clearance and other customary conditions.

 

An all-stock structure lets AMD preserve cash, but it transfers part of the economic risk to its shareholders through the shares issued for the acquisition. AMD has not presented the $8.2 billion as a cash payment, research budget or guaranteed future revenue from World Labs.

 

The deal follows an existing technical relationship. Reuters reported that AMD and World Labs began working together last year on model training and inference optimization using AMD GPUs. AMD also invested in a $1 billion World Labs financing round earlier in 2026.

 

That history gives the acquisition more technical substance than a cold purchase of an unfamiliar startup. The two teams have already tested how World Labs workloads behave on AMD hardware, creating a foundation for deeper co-design if the transaction closes.

 

World Labs Brings Spatial Intelligence to AMD

World Labs develops spatial-intelligence models that can generate, reconstruct and simulate interactive 3D environments from text, images and video. It also works on robotic learning and simulation, areas that require models to reason about geometry, movement and physical interactions rather than only predict language.

 

Those workloads can demand a different combination of compute, memory capacity, bandwidth, networking and software than a conventional large language model. AMD says direct access to the researchers building them should help the company understand how future applications will stress complete computing systems.

 

The company is therefore buying expertise as well as software. World Labs can inform decisions across accelerator architecture, CPU design, interconnects, ROCm software and system configuration, potentially shortening the feedback loop between model experiments and AMD’s hardware road map.

 

Physical AI remains an emerging market rather than a mature revenue category. Robotics and simulation systems must still meet demanding standards for latency, reliability, safety and cost, so the acquisition does not guarantee that World Labs research will quickly translate into commercial hardware sales.

 

Related Coverage

 

Fei-Fei Li Will Become AMD Chief Scientist

Li co-founded World Labs in 2024 after decades of work in computer vision and human-centered AI. She is widely associated with ImageNet, the large labeled-image project that helped demonstrate how data scale could improve visual recognition and accelerate modern deep learning.

 

After closing, Li would report directly to Su rather than placing World Labs beneath a distant product group. That reporting line suggests AMD wants model research to influence company-wide technical decisions, not operate only as an application team attached to its current accelerator portfolio.

 

The World Labs researchers are expected to keep advancing AI models inside AMD. Retaining that mission will be central to the deal’s value because highly specialized research teams can lose momentum when acquisition integration replaces experimentation with short-term product deadlines.

 

AMD says the combination will support an open AI ecosystem. The claim is strategic, not a promise that every World Labs model, dataset or training method will become open source. The practical test will be what the combined company releases and how broadly its tools work beyond AMD hardware.

 

World Labs Expands AMD Beyond AI Chips

The acquisition would give AMD a model-development operation at a time when chipmakers increasingly compete as full-stack AI companies. Customers now evaluate accelerators together with CPUs, networking, compilers, libraries and reference models because weaknesses in any layer can limit the performance of the whole system.

 

Nvidia has built a strong position through CUDA, integrated systems and its own model research. AMD has answered with Instinct accelerators, EPYC processors, ROCm software and system partnerships. World Labs adds direct visibility into the workloads that future robots, simulators and creative tools may generate.

 

The strategic benefit could work in both directions. AMD can give World Labs more engineering resources and optimized infrastructure, while the lab can expose gaps in AMD’s hardware and software before those limitations reach outside customers.

 

Regulatory approval, researcher retention and technical integration are now the major checkpoints. If the transaction closes on schedule, the clearest early evidence of success will be new World Labs research optimized on AMD systems and visible changes to the company’s hardware and software plans for physical AI.