AMD acquires World Labs, with Fei-Fei Li joining as Executive VP and Chief Scientist
- World Labs signed a definitive agreement to join AMD, with the deal expected to close by the end of 2026 subject to regulatory approvals and other customary closing conditions.
- Dr. Fei-Fei Li will join AMD as Executive Vice President and Chief Scientist, working directly with CEO Dr. Lisa Su.
- Justin Johnson and Ben Mildenhall will keep leading the World Labs team as it joins AMD to form a frontier research organization, aimed at an end-to-end open AI ecosystem spanning hardware, software, platforms, and open models.
- The two companies started a deep technical partnership last year, beginning with model training and inference optimization on AMD GPUs, before deciding a full combination was a natural fit.
- World Labs was founded in 2024 and frames its work around AI for the spatial and physical world; the announcement says scaling that work requires getting closer to the hardware.
Hacker News opinions
What is AMD even getting out of this? Talent, and a shot at building the next generation of multimodal models. They're stacking up bets, they picked up Taalas recently too.
The real bet here is that conventional LLM development plateaus or becomes commoditized, so world models are where the value lands.
I'd never heard of them before this, but their world models are actually transformers. That's encouraging, it means the conversation has shifted to assuming transformers hold emergent world models.
Where does this go? My bet is world models become the new media, replacing TV, movies, music and video games.
Seems like AMD wants to jumpstart a model and simulation ecosystem to compete with Nvidia, who's been at this for about a decade. Might also be a response to NV acquiring HF. Downside is it could kill the interesting stuff World Labs was doing with Marble and Atlas.
Hard to do what World Labs wanted without a serious amount of compute, so honestly this looks like a good match.
Throwaway here because I don't want the flame war. I black box test these things for real design clients, and World Labs output was never usable, distortions and errors everywhere. It rode the gaussian splat hype train with great demos, but it felt like vaporware.
It's sad you had to make a throwaway just to say that. People shouldn't be scared to post honest opinions about tech here as long as it's legal speech.
Is a two year old company worth $8 billion? The name Fei-Fei Li alone is probably worth a chunk of that. It's a public bet, not a valuation based on product economics.
A company is worth what someone will pay for it. OpenRouter sold for $7.5B. These people don't evaluate companies by value, they evaluate by what a buyer has to pay.
AMD grew revenue 50% year over year last quarter, to $11.5B. If this team keeps that going or grows it, then yeah.
Neolabs keep moving down the stack. Neoclouds want to do neolab things, and now chipmakers want to do neolab things. Nobody actually knows what the right integrations are yet.
AMD is still struggling to establish itself in AI, which makes it a worthy competitor. Agglomerating talent and resources is probably good, knowing little else.
This was absurdly soon, and the Taalas acquisition was fast too. AMD may be prepping for ultra fast inference and embodied AI inference.
If the sigmoid approacheth, the first one to get a good model into silicon at acceptable benchmarks makes truckloads. Most people don't need SOTA for most of their problems.