AMD to Acquire World Labs in $8.2 Billion Deal to Expand Its AI Strategy

AMD is turning a strategic bet into full ownership. The chipmaker said Monday, September 28, 2026, that it has agreed to acquire World Labs, the San Francisco AI lab founded by Dr. Fei-Fei Li, in an all-stock deal worth $8.2 billion. It's a notable jump from AMD's prior role as an investor and partner to now becoming the company's owner outright.

World Labs isn't building chatbots. Founded in 2024, the two-year-old startup develops what researchers call "world models," AI systems designed to generate, reconstruct and simulate interactive 3D environments from text, images or video. Its flagship product, Marble, has been pitched both as a tool for building entertainment experiences and as a way to generate simulated environments for training robots. Earlier this year, Li and AMD CEO Lisa Su publicly demoed Marble turning a handful of images into a full 3D scene. Before this deal, World Labs had raised roughly $1 billion in funding, according to Fortune, with AMD already among its backers through an existing inference and training partnership formed last year.

Under the terms announced, the transaction is expected to close by the end of 2026, pending regulatory approval, Bloomberg reported. Li will join AMD as executive vice president and chief scientist, reporting directly to Su, a reporting line that signals how central her role is expected to be. World Labs' broader team is also joining AMD, though the companies haven't disclosed exact headcount, and the group is expected to keep focusing on AI model research rather than being folded into unrelated projects.

"Together, we'll combine World Labs' deep expertise in AI and world models with AMD's compute leadership," Su said in a post announcing the deal. World Labs, for its part, framed the rationale around the need for "close collaboration across model research, systems and compute," the kind of language that reflects AMD's broader push to tie chip design more closely to how cutting-edge AI models actually behave.

That's really the strategic logic here: AMD wants a front-row seat to how next-generation AI workloads evolve, so its hardware roadmap isn't chasing yesterday's models. World models are increasingly viewed as important for physical AI, robotics, autonomous vehicles and other systems that need to understand real-world space, an area where useful training data is scarce, making Li's synthetic-data approach potentially valuable well beyond entertainment applications.

The deal also fits a pattern that's become familiar across the AI industry this year. OpenAI paid about $6.4 billion for Jony Ive's device startup io. Nvidia spent a combined $33 billion scooping up Groq and Hugging Face. Meta put $14 billion into a minority stake in Scale AI specifically to bring in founder Alexandr Wang. Buying talent and technology outright, rather than just partnering, has become one of the AI industry's defining moves, and AMD is now following that same playbook against a considerably larger rival in Nvidia.

Citigroup analyst Atif Malik framed the acquisition as serving three goals at once: bringing in top AI research talent, strengthening AMD's position in physical AI, and helping the company build better end-to-end systems rather than just components. Whether that translates into a meaningful competitive edge against Nvidia's dominant position will likely take time to show up in AMD's actual product roadmap, but the world-model space itself is heating up, with Meta's former chief AI scientist, Yann LeCun, launching a rival effort called AMI Labs earlier this year.