Fei-Fei Li becomes AMD's chief scientist after the $8.2 billion acquisition of World Labs
AMD announced on Monday, September 28, 2026 that it has agreed to acquire the AI lab World Labs, founded by Fei-Fei Li, for $8.2 billion in an all-stock deal. Li herself becomes executive vice president and chief scientist at AMD, reporting directly to CEO Lisa Su. According to CNBC, the deal is AMD's second-largest ever, after the acquisition of Xilinx for roughly $50 billion in 2022, and is expected to close later this year.
What World Labs Does — and Why AMD Wants It
World Labs, based in San Francisco, develops what is known as a world model: a model capable of simulating three-dimensional environments. Earlier this year, Li and Lisa Su showcased the model Marble in a demo, where it generated a complete 3D scene from just a few images.
The venture is about more than impressive demos. "Intelligent agents, whether they're robots, 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 the presentation, according to CNBC. The logic is that agents can be trained in simulated environments with an understanding of physics before being released into the physical world.
For AMD, this is a crossroads between chips and models. 24/7 Wall St. describes the acquisition as Su's bet on "physical AI" — AI that operates in and on the physical world — and points out that world-model workloads are memory-intensive, which in turn is driven by precisely the kind of hardware AMD sells.
Who Is Fei-Fei Li?
Li is a Stanford professor who previously worked at Google, where she led AI research. She is considered one of the most central figures in modern AI research and is referred to in commentary around the acquisition as the "godmother of AI." She now becomes AMD's chief scientist and executive vice president, reporting directly to Su — an unusually senior role for an incoming profile that prioritizes research over product management.
In a comment on the acquisition, carried in syndicated coverage with reference to AMD's press release, Li said: "Joining AMD will give our team the resources and engineering depth to accelerate our research and help define the infrastructure needed for the next era of AI."
The Financial Logic: Stock Instead of Cash
The price looks high on paper. It is reported to be roughly 64 percent above the $5 billion valuation World Labs had just seven months earlier — figures Insider Monkey via Yahoo Finance gives with the caveat "reportedly," without independent confirmation. But AMD is paying entirely in stock. With a market capitalization of approximately $1 trillion, the new shares increase the share count by less than 1 percent.
The structure also preserves liquidity. AMD had $13.1 billion in cash and short-term investments at the end of the second quarter, and an all-stock settlement keeps that money available for the rollout of the Helios platform, according to 24/7 Wall St.
The stock market greeted the news with a shrug: the shares fell 0.4 percent in trading outside regular market hours, according to Stocktwits coverage via Yahoo Finance.
Su on the Strategy: Understand the Models to Build the Right Hardware
Lisa Su tied the acquisition directly to AMD's position in the AI market, where the company is chasing Nvidia. "Building the computing platforms for the next generation of AI requires a deep understanding of how the models are evolving," Su said, quoted from the company's press release.
Su also pointed to progress in the software stack around AMD's chips: more than 3 million models now run out of the box on AMD hardware, and contributions to the open-source project ROCm have increased more than tenfold over the past year, according to 24/7 Wall St.
AMD's senior vice president for AI, Vamsi Boppana, described the World Labs team as "a horizontal models capability for the company" — that is, not a bet on a single AI segment — and as "a core for attracting more talent." It should be noted that these quotes reach the public secondhand, through Patrick Moorhead's X post as republished by Stocktwits, not through a direct interview.
The Promised Answer on the Competition Question
The most sensitive detail in a deal where a chipmaker acquires a model lab is the relationship to AMD's own customers: hyperscalers and AI companies that buy GPUs to train competing models. AMD has said the company does not intend to compete with its own AI customers — a statement Insider Monkey calls justified concern management, and which Boppana, according to the secondhand account, called "not the intent at all" when it comes to frontier models.
That means in practice that World Labs will likely function as a research and tooling engine that makes AMD hardware more attractive, not as a frontier lab meant to outcompete AMD's customers. But the line between "horizontal models capability" and direct competition is unclear, and no detailed frameworks have so far been made public for how it will be enforced.
What Remains Unresolved
Several matters remain open. First, the transaction is not closed: AMD plans to keep World Labs separate from the chip business until closing later this year, and the status of antitrust approval appears unresolved in the available coverage. Second, the entire story rests on secondary sources — CNBC, Yahoo Finance/Insider Monkey, 24/7 Wall St. and syndicated coverage — and AMD's press release and investor materials have not themselves been reviewed here. The reporting on the premium of roughly 64 percent and the $5 billion valuation should be read as independently unconfirmed figures.
There is also speculation that should stay where it belongs: some retail investors interpret Li's placement directly under Su as a sign that she could be a possible successor as CEO. These are undocumented market comments, not information from AMD.
The deeper question is commercial: whether world models actually become a large enough workload to drive chip sales, or whether the acquisition is primarily a differentiation and talent play. Li's own argument — safer robots and vehicles trained in simulation — suggests the former. The answer will come at the earliest when Marble and its successors meet real customer needs after the deal closes.
Why It Matters
For AMD, this is an attempt to move the competition with Nvidia from chip-level specifications to the whole stack: hardware, software ecosystem, and now also an understanding of where the models are heading. World models sit at the intersection of both — they are memory-hungry enough to need AMD's hardware, and they require a forward-looking platform strategy to be well served.
For the industry, the signal effect may be the greatest: one of AI research's most prominent names is now choosing a chip company that challenges Nvidia, in a chief scientist role. How it is digested — whether World Labs remains "horizontal," and whether memory-hungry world-model work actually translates into AMD revenue — is something none of the sources can answer yet.

