AMD is acquiring World Labs, Fei-Fei Li’s AI startup, for approximately $8.2 billion in stock. The deal is expected to close by the end of 2026; Li herself will join AMD as Chief Scientist and will report directly to CEO Lisa Su. For AMD, this is its biggest move yet to catch up with Nvidia in the race for AI hardware and compatible software.
The Deal at a Glance
The figures announced by AMD and World Labs on September 28, 2026, are as follows:
| Details | Specification |
|---|---|
| Purchase price | Approx. $8.2 billion, entirely in AMD stock |
| Expected closing date | End of 2026, subject to regulatory approvals |
| Founded | 2024 by Fei-Fei Li |
| Headquarters | San Francisco; the team will remain on-site |
| Funding to date | $230 million in seed funding, followed by an additional $1 billion in February 2026 (with AMD and Nvidia among the investors) |
| New Role for Fei-Fei Li | Executive Vice President & Chief Scientist at AMD, reports to Lisa Su |
Noteworthy: AMD and Nvidia had both participated in the same funding round for World Labs just this past February. Now AMD is acquiring the startup outright, further intensifying the competition between the two chipmakers in the AI sector.
What does World Labs actually do?
World Labs is less than two years old but already has a fairly clear focus: so-called spatial intelligence models that generate, reconstruct, and simulate interactive 3D environments from text, images, or videos. Its first public product is called Marble and is aimed at entertainment applications as well as training robots in simulated worlds.
That’s exactly where the appeal lies for AMD. Anyone looking to train robots or autonomous systems needs vast amounts of simulation data, because real-world training runs or movements would be too expensive and too slow. World Labs’ so-called world models are designed to fill exactly this gap and are regarded in the industry as a key building block for “Physical AI”—that is, AI that not only predicts text but also navigates physical spaces.
Why AMD is making its move right now
AMD justifies the acquisition by pointing out that AI workloads are currently shifting significantly: away from pure language model training, toward reasoning, robotics, simulation, and, indeed, Physical AI. Anyone who wants to serve these applications must tailor hardware, software, and systems to them from the ground up—and that’s exactly what World Labs’ modeling expertise is intended to provide.
It’s also interesting to note that the two companies had already been collaborating prior to the acquisition. According to consistent reports, AMD and World Labs had launched a partnership focused on inference optimization and training on AMD hardware. The acquisition thus seems less like a spontaneous move to snap up a trendy startup and more like the logical continuation of an existing relationship.
Fei-Fei Li: From ImageNet to AMD’s Chief Scientist
For AMD, the personnel decision is at least as important as the technology. Fei-Fei Li is considered one of the most influential figures in AI research: She co-developed ImageNet, the dataset that made the breakthrough in deep learning starting in 2012 possible in the first place; served as head of AI research at Google Cloud; and leads the Institute for Human-Centered AI at Stanford University. As a result, she is often referred to in the industry as the “Godmother of AI.”
“Fei-Fei and the World Labs team bring exceptional research capabilities and modeling expertise,” AMD CEO Lisa Su was quoted as saying in the press release. Li herself explained the move as follows: Joining AMD gives her team the resources and technical depth to advance their own research more quickly.
Context: Another Showdown with Nvidia
$8.2 billion is a large sum by AMD’s standards, but not an unprecedented one. In 2022, the company had already acquired Xilinx for approximately $35 billion to strengthen its position in the programmable chip sector. The World Labs deal is smaller but is just as specifically aimed at filling a gap in the portfolio that Nvidia has long been trying to close with its own world-modeling initiatives.
There is also movement among competitors in the market for physical AI and robotics: Just recently, for example, Qualcomm made a similar move with its acquisition of PickNik Robotics. The major chipmakers are apparently betting that robotics and simulation will be the next major growth area for AI, now that the hype surrounding pure language models has lost some steam.
What Happens Next
Before the deal is finalized, it must still go through the usual regulatory approvals; given the scale of the transaction, this is likely to be closely monitored, especially in the U.S. According to AMD, World Labs will remain as a team in San Francisco; the company has not yet provided operational details regarding its integration into AMD’s software stack.
For AMD shareholders and competition watchers, one question in particular is likely to remain intriguing: Whether Li’s research will actually result in products that set AMD’s chips apart from the competition, or whether the deal was ultimately just an expensive signal to the market.
