AMD is making an $8 billion bet (0.8% of AMD valuation) to see if World Model where if we model all physics and chemistry and everything else we can accelerate the pace of major scientific discovery. Fei Fei and World Models are a leader in this area and need the trillion dollar financial power of AMD to have a good shot at taking AI to the next level.

Great news for world models and spatial-intelligence.

AMD agreed to buy Dr. Fei-Fei Li's World Labs for ~$8.2B in stock, installing Li as chief scientist.

Thats ~64% above the $5B valuation talk of World Lab for a $1B February round, which AMD and Nvidia both joined.

World… pic.twitter.com/W5kN2IUaD9

— Rohan Paul (@rohanpaul_ai) September 28, 2026

AMD is buying Fei-Fei Li’s World Labs for about $8.2 billion in stock. Li becomes AMD EVP and Chief Scientist, reporting to Lisa Su. World Labs had already partnered with AMD on training/inference and AMD Ventures had invested earlier.⁠TechCrunch

The size of the bet: AMD recently crossed ~$1 trillion market cap. $8.2B is closer to ~0.8% of that, not “less than 0.1%.” Still small versus AMD’s valuation, and small versus the $50B Xilinx deal. It is a research-and-roadmap bet, not a balance-sheet-defining one.⁠CNBC

What AMD is actually buying is not a model of all physics and chemistry. It is a team and stack for spatial intelligence. Models that reconstruct, generate, and simulate interactive 3D worlds from text/images/video, plus robotics simulation (including the SceniX acquisition). AMD’s stated reason is co-design: next-gen models tell them what chips, systems, and software must do next. Nvidia already has Cosmos world models. AMD has been thinner there.

Li’s own framing is language is not enough. The universe is made of things, not words. The hard problems (science, robotics, design) need models that reason over structure and behavior of the physical world. She explicitly lists drug discovery, creativity, learning, and next-gen robotics as the destination.