Fei-Fei Li spent years building a case that machines need to understand the physical world before they can be trusted to act in it. On Monday, that argument became a chip company’s bet. AMD agreed to buy World Labs, the startup Li co-founded, for about $8.2 billion in an all-stock deal, and the transaction will make Li, a Stanford professor and one of the best-known figures in artificial intelligence, a billionaire.
Her net worth will rise to at least $1.6 billion once the deal closes, up by more than half, according to the Bloomberg Billionaires Index. The gain comes mostly from her stake in a company that did not exist three years ago.
Li’s standing in the field predates the startup by a long way. Born in Beijing and raised partly in New Jersey, she built ImageNet at Princeton and Stanford, a dataset of millions of labeled images that helped make deep learning practical more than a decade ago. She later co-founded Stanford’s Institute for Human-Centered AI, a perch from which she argued that AI should serve people rather than replace them. She is sometimes called the godmother of AI, a title she has worn uneasily in public.
World Labs works on what Li calls spatial intelligence: models that build a three-dimensional understanding of the world from images, video and text, so that machines can reason about objects, rooms and movement rather than just label them. She has described it as the missing piece between a model that can caption a photo and a machine that can walk into a room and act. That framing has made the company a magnet for attention even as its commercial products remain early.
Li founded World Labs in 2024, and it raised roughly $1 billion before the sale, with early backing from investors including Andreessen Horowitz and Radical Ventures. Within months it was valued above $1 billion, a rise driven more by the scarcity of top researchers than by revenue, which remains small.
Under the deal, Li will join AMD as executive vice president and chief scientist, reporting to chief executive Lisa Su. The two companies already had a working relationship: AMD and World Labs formed an inference-optimization and training partnership last year, and Li appeared at AMD’s CES presentation earlier this year. The acquisition is expected to close by the end of 2026, subject to regulatory approval.
For AMD, the logic is straightforward. The company has spent Su’s tenure rebuilding itself into the closest thing Nvidia has to a rival in AI chips, and it has argued that understanding how frontier models behave is the way to shape its next generation of hardware. Buying World Labs puts Li, and a team that studies precisely that, inside the company rather than across a partnership table.
The price reflects how fast a young AI lab can become valuable. At $8.2 billion, the deal values World Labs at roughly eight times the money it had raised, a multiple that says more about the market for researchers who can push models toward physical reasoning than about the startup’s sales.
The transaction caps an unusually rapid rise. Li left an academic post to build the company, watched it become a center of attention within two years, and now stands to convert that attention into a fortune measured against the founders who built the last generation of technology companies. Few of them reached a ten-figure net worth this quickly.
Analysts said the deal is as much a hiring move as an acquisition. The chip industry is fighting over a small pool of people who can push models toward physical reasoning, and securing Li alongside the World Labs team gives AMD a research bench it could not easily assemble on its own.
A deeper bet runs under the transaction. The next wave of AI systems, the ones meant to run robots, cameras and vehicles, will need models that understand space and physics the way Li’s lab is teaching them to. AMD is wagering that whoever supplies the chips for that wave will have to know what those models need, and that owning the research gives it a line Nvidia cannot match on the same terms.
For Li, the sale converts years of argument into a concrete stake. She has spent much of her career outside industry, building datasets and institutes rather than companies. World Labs was the bridge between the two, and Monday’s deal is the point where the academic position and the commercial one became the same thing.
The open question is whether spatial intelligence becomes a product or a research advantage. AMD is buying a group whose work is still, by the standards of the industry, early, and whose models have yet to prove they can do what Li has promised. The company is betting that the person who framed the problem is worth the price of the lab she built to solve it.
