SAN FRANCISCO – Chipmaker Advanced Micro Devices agreed to acquire World Labs for US$8.2 billion S($10.5 billion), gaining an artificial intelligence start-up founded by industry pioneer and researcher Fei-Fei Li.
The all-stock transaction is expected to close by the end of the year, subject to regulatory approvals, the companies said in a statement on Sept 28.
AMD, a key challenger to Nvidia in the market for AI chips, is adding a team of leading researchers and their models, along with insights into how the artificial intelligence field is evolving. That should help the company plan its future hardware releases.
“The more you understand end to end, the better system you are going to build,” AMD chief executive officer Lisa Su said in a Bloomberg Television interview that is slated to air on the afternoon of Sept 28.
“That’s why we are acquiring World Labs,” Su said. “To have access to the world-class talent Fei-Fei has brought together and bring it together with AMD’s capabilities around hardware, software and systems.”
World Labs develops so-called world models: AI software that can run applications in the physical world.
The models generate and reconstruct three-dimensional environments, which the start-up says can be used to help advance work on robots, scientific discovery and factory equipment.
Li, who served as CEO of World Labs, is a renowned researcher and computer scientist at Stanford University.
As part of the agreement, she will join AMD as an executive vice president and chief scientist and report directly to Su.
“As our company grows, so does our ambition,” Li said in the Bloomberg Television interview. And that has increased the start-up’s need for computing power, she said.
“At this point, if we can join forces with AMD, we really are giving ourselves, AMD and the ecosystem an opportunity to accelerate the flywheel between software and hardware development,” she said.
The deal will also help bolster AMD’s offerings by allowing it to provide a variety of AI models to customers.
The company has been expanding its line-up in a bid to sell more products and services to data centre clients, its biggest source of revenue.
The idea is to offer the necessary elements to rapidly set up the facilities and put them into operation.
“You are going to have both open as well as proprietary models,” Su said. “AMD has been a contributor there. I think we are going to expand that contribution.”
Ultimately, the goal is to build better AI, Su added.
“Our ambition is to define the future of AI compute.”
AMD’s larger rival Nvidia pioneered the approach of bundling equipment and services with AI chips to help clients get up and running quickly.
It is now using the strategy to target new kinds of customers – going beyond the cloud computing giants that still fuel much of its growth.
Nvidia recently agreed to acquire another high-profile AI start-up, Hugging Face, in a deal valued at about US$13 billion.
World Labs, launched in 2024, has about 70 employees. That includes a team of robotics experts from its July acquisition of SceniX, a start-up that builds software to train robots in realistic virtual worlds.
During Su’s 12-year tenure as AMD’s leader, the chipmaker has sought to be recognised for its technical capabilities and escape its reputation as an also-ran in Silicon Valley.
In the company’s core processor market, AMD has already emerged from the shadow of Intel and now has products that are rated the best in the industry.
In the market for artificial intelligence accelerators – chips used to train and run AI models – AMD remains a distant second to Nvidia.
But Su has rolled out new products that she says can outperform offerings from its larger rival.
Investors see the company as a key beneficiary of AI spending and have bid up its shares nearly threefold in 2026.
That has given AMD a valuation of roughly US$1 trillion, a distinction shared with only about a dozen other tech titans.
Nvidia remains the world’s most valuable company, with a market capitalisation above US$5 trillion.
But its shares are only up about 22% in 2026, underperforming nearly all major chipmakers in 2026. BLOOMBERG
