AMD to Acquire Fei-Fei Li's World Labs for About $8.2 Billion in All-Stock Deal, Its Second-Largest Ever

AMD said Monday it has agreed to acquire World Labs, the San Francisco AI lab founded by Stanford professor and former Google AI research leader Fei-Fei Li, in an all-stock transaction valued at about $8.2 billion, according to CNBC. World Labs develops "world models," which can be used to simulate 3D environments, CNBC reported. The deal marks AMD's second-largest acquisition on record, trailing only the roughly $50 billion it paid for Xilinx in 2022, per CNBC. AMD plans to keep World Labs operating separately from its chipmaking business until the transaction closes later this year, CNBC reported.

From investor to owner

AMD was not a stranger to World Labs before Monday's announcement. The chipmaker was among the backers of a $1 billion funding round World Labs disclosed in February 2026, which also included Autodesk, Emerson Collective, Fidelity Management & Research Company, Nvidia and Sea, according to World Labs' own announcement and coverage from Trending Topics. Autodesk separately disclosed it contributed $200 million to that round, per Frontier Enterprise. Public reporting reviewed for this story does not disclose the size of AMD's individual stake in that round or the valuation attached to it.

AMD's venture arm had already staked out its interest in the company's underlying technology months before the buyout. In a March 2026 post, AMD Ventures described spatial intelligence as "the missing pillar that unlocks new workflows and entirely new categories across virtual and physical domains," arguing that it upgrades world models "from generating a scene to generating environments you can enter, edit, and act inside."

What World Labs actually sells today

World Labs' commercial product is Marble, a generative world model that launched in November 2025 after a limited beta and is offered through free and paid tiers, according to The AI Insider. World Labs opened Marble to the general public alongside the February 2026 funding announcement, per Trending Topics. The platform lets users generate spatially coherent, persistent 3D environments from text, images, video or 3D layouts, with outputs exportable as Gaussian splats, meshes or video for use in existing workflows, per Frontier Enterprise. The company is targeting early use cases in gaming, visual effects, virtual reality and robotics, The AI Insider reported. Public reporting reviewed for this story does not disclose Marble's revenue, customer count or paying-user base.

AMD's stated rationale for the purchase centers on using that research to inform chip design rather than on Marble as a standalone product line. CNBC reported that World Labs' work on next-generation world models will let AMD plan years in advance for the capabilities its AI chips will need to support. Li and AMD CEO Lisa Su gave a joint demonstration earlier this year in which Marble built a three-dimensional scene from a handful of images. "Intelligent agents, whether it's robots or vehicles or even tools, can learn inside very-rich physics-aware digital worlds before they even need to be deployed into the real one, making them much safer," Li said at that presentation, per CNBC's account.

Li's new role, and the ties between the two leaders

Under the agreement, Li will join AMD as chief scientist and an executive vice president, CNBC reported. In a social media post Monday, Li called Su a "great friend" and wrote, "We're unwavering in our mission, and even more excited to continue building a world leading frontier research organization that is durable for the decades ahead," according to CNBC's reporting of the post. The story does not specify whether other members of the World Labs team will join AMD or how the research organization will be structured once the deal closes.

Part of a broader AI dealmaking wave

The World Labs purchase extends a run of AI-related acquisitions AMD has made in recent years. It paid $665 million for Silo AI in 2024 and $4.9 billion for ZT Systems the same year, then picked up several smaller AI companies, including inference-software maker MK1, CNBC reported. In August 2026, AMD announced a deal to acquire Toronto-based inference chip startup Taalas at an undisclosed price, according to AMD's investor relations release.

The $8.2 billion price tag also sits inside a broader pattern of large technology companies paying up for AI talent and research. CNBC noted that OpenAI paid about $6.4 billion for Jony Ive's device startup io last year, Meta spent $14 billion in 2025 for a minority stake in Scale AI in a deal that brought in founder Alexandr Wang to run a new AI division, and Nvidia has committed a combined $33 billion to Groq and Hugging Face, buying assets from the former in December and agreeing to acquire the latter. On our reading of those CNBC-cited figures — $6.4 billion, $8.2 billion, $14 billion and $33 billion — AMD's outlay ranks toward the lower half of that group rather than at the top. That comparison is our interpretation; AMD did not size the deal against other transactions in its own statement.

Public reporting reviewed for this story does not include AMD share-price reaction to the announcement, sell-side analyst commentary on the deal's valuation, or any AMD guidance on whether the acquisition will be accretive or dilutive to earnings. AMD has not disclosed regulatory conditions attached to the deal beyond stating the transaction is expected to close later this year, per CNBC.

Bottom Line

AMD is paying roughly $8.2 billion in stock, its second-largest deal on record, to bring in-house a research lab it had already backed financially, along with its founder, who becomes the company's chief scientist and an executive vice president. The stated goal, per CNBC, is to use World Labs' work on next-generation world models to plan years in advance for what AMD's AI chips need to be capable of, rather than any product plan the company has detailed for Marble. What the deal means for AMD's near-term financials, and how the market and analysts value it relative to comparable AI acquisitions, has not been detailed in the reporting reviewed for this story.