On February 18, 2026, World Labs — the spatial intelligence startup founded by Stanford professor Fei-Fei Li, the computer scientist widely called the “godmother of AI” — announced a $1 billion funding round. Reuters reported the round values the company at roughly $5 billion, and the company’s own announcement lists AMD, Autodesk, Emerson Collective, Fidelity Management & Research Company, NVIDIA, and Sea among the investors. Less than eighteen months earlier, World Labs had launched with a $230 million raise in September 2024.
The bet behind the money is explicit: spatial intelligence. While frontier labs pour gigawatts of compute into language and image models, World Labs builds world models — systems that understand and generate three-dimensional space. In the 2026 model race, this is a differently shaped ticket, bought with a billion dollars of other people’s money.
A $1 Billion Round
Crunchbase’s weekly funding roundup for February 20 listed World Labs as by far the week’s largest deal, and press coverage of the round put Autodesk’s contribution at around $200 million. What makes the round interesting is less the number than the cap table. NVIDIA and AMD — the two chip vendors collecting the AI data center buildout’s revenue — are simultaneously buying equity in an application-layer model company. Autodesk brings a design-tools distribution channel; Fidelity brings institutional money. This is not a classic venture syndicate. It is a supply chain taking positions.
Li’s own history does half the storytelling. She led the creation of ImageNet, the dataset that taught machines to see and arguably lit the fuse on the deep learning era. World Labs is her argument about the next decade: having taught machines to recognize images, teach them to model space.
Spatial Intelligence and World Models
The core difference between a world model and an LLM is the object of study. An LLM learns the statistical structure of language; a world model learns the structure of physical space — geometry, persistence, what is where, and what happens when you move through it. World Labs frames the applications as storytelling, creativity, robotics, and scientific discovery.
For robotics, spatial grounding is precisely what text models cannot supply. For games and simulation, the requirement is a persistent world you can revisit from any viewpoint, not a sequence of independent frames. That gap between “generating an image of a room” and “generating a room” is the entire product thesis.
Marble: The Shipping Product
World Labs’ first product, Marble, generates what the company calls “spatially coherent, high-fidelity, and persistent” 3D worlds from images, video, or text. Around it sit a developer API, the Spark JavaScript library, and the experimental Marble Labs. This matters for reading the funding news: the billion dollars are flowing into a product line that already ships, with an API and a web-native integration path — not into a slide deck. Shipping a JS library as a first-class artifact signals who they expect to build with this: web developers.
Why Chip and Tool Vendors Are Betting
NVIDIA’s position is the instructive one. It sells GPUs by the tens of thousands into buildouts like the Meta–NVIDIA Blackwell-scale expansion, and it also takes equity in the demand side. The logic is straightforward: world models are heavy in both training and inference, and 3D generation extends what AI compute is for — beyond chat and search, into simulation and content production. New workload classes mean new demand curves.
For AMD, an equity stake in a model company is one way to ensure its accelerators are woven into that ecosystem early. For Autodesk, the calculation is nearer-term: design professionals are the obvious first market for generated 3D, and a stake is cheaper than building the capability in-house.
What It Means for Builders
First, treat 3D generation APIs as entering the seriously usable phase: game prototyping, architectural visualization, and synthetic training environments for robotics are all reasonable things to test with Marble now. Second, follow the cap table as a signal: when NVIDIA, AMD, and Autodesk invest in the same thesis, the industry is pricing 3D understanding as the next differentiation axis. Third, world models complement LLMs rather than replace them — language and reasoning in one, space and simulation in the other. The interesting products of 2026 wire the two together.
Sources
- AI pioneer Fei-Fei Li’s World Labs raises $1 billion in funding — Reuters
- Funding 2026 — World Labs
- The Week’s 10 Biggest Funding Rounds — Crunchbase News
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.