Fei-Fei Li’s World Labs has successfully secured a $200 million investment from the prominent software design company Autodesk. This funding contributes to a larger $1 billion investment round, which also includes contributions from notable backers such as AMD, Emerson Collective, Fidelity, and Nvidia, among others.
Having debuted from stealth in 2024 with $230 million in funding and a $1 billion valuation, World Labs chose not to disclose whether this latest investment round increased its current valuation. However, reports from a month prior indicated that the company was aiming for a $5 billion valuation.
This new partnership between World Labs and Autodesk will involve the two entities collaborating to investigate how World Labs’ advanced AI models, capable of generating and interpreting immersive 3D environments, can integrate with and complement Autodesk’s existing toolsets, and vice versa. The initial focus for this collaboration will be on applications within the entertainment sector.
For World Labs, Autodesk’s substantial investment serves as a clear indication of its product’s commercial viability. The startup’s inaugural world model offering, Marble, launched last November, empowers users to create 3D environments that are both editable and downloadable.
Autodesk stands as a leading developer of 3D CAD (computer-aided design) software, with its platforms underpinning crucial workflows across architecture, engineering, construction, manufacturing, and entertainment. This deep-rooted focus on the constructed world makes the investment in sophisticated spatial AI a natural and strategic extension of its core business operations.
As Fei-Fei Li articulated in a press statement: “Autodesk has a long history of enabling individuals to think spatially and solve real-world challenges. Together, we are united by a shared objective: developing physical AI that enhances human creativity and provides more powerful instruments to designers, builders, and creators.”
Under the terms of the agreement, Autodesk will also take on an advisory role for World Labs, and both companies will engage in collaborative efforts at the “research and model level.”
Daron Green, Autodesk’s chief scientist, informed TechCrunch that the partnership is in its preliminary stages, meaning the exact nature and scope of its implementation are still being determined.
“You could envision us utilizing their models or them utilizing our models in diverse contexts,” Green stated.
He speculated that users might begin with a world-model-generated sketch within World Labs (for instance, an office layout) and subsequently delve into specific design elements (like the design of a desk), where Autodesk’s technology would then become relevant.
“Conversely, you might want to integrate an object designed within our [platform] into a context you generate using one of [World Labs’s] prompts,” Green added.
Green further clarified that the agreement does not encompass data sharing between the two companies.
Green indicated that the initial focus for both companies will be on use cases within media and entertainment. This strategy aligns with most other developers of world models, including Google DeepMind and Runway, who frequently target gaming and interactive entertainment as their primary market entry points.
Autodesk already maintains collaborations with most major media production companies and has prior experience in training models specifically for character animation.
“These models are quite similar to world models,” Green explained. “They involve characterizing an animal within an environment, responding to physical constraints like time or specific terrain it needs to navigate. This demonstrates a physical understanding within the model, and one can easily see how that could be integrated [with World Labs’s technology]. You’re not just animating the dog; you’re providing it with a world where it can now interact.”
This collaboration with World Labs reinforces Autodesk’s broader initiative to embed more advanced AI functionalities across its entire software portfolio. The company is actively developing “neural CAD,” an innovative generative AI model type trained on geometric data that can comprehend components and entire systems. In essence, it can create functional 3D models, not merely static images, complete with an understanding of how these designs would perform in real-world scenarios.
Autodesk’s neural CAD models are progressively being integrated into its product design and architecture software, marking a step towards more sophisticated spatial intelligence. However, World Labs’ models possess the potential to extend this capability beyond individual design files, moving towards comprehensive digital representations of the physical world.
Green believes that a future convergence of diverse AI systems, including large language models, world models, and neural CAD, will ultimately lead to improved designs for Autodesk’s clientele.
“For AI to achieve true utility, it must comprehend worlds, not merely words,” Li declared in her statement. “Worlds are governed by geometry, physics, and dynamics, and the reconciliation of semantic, spatial, and physical understanding represents the next significant frontier for AI.”
This article was initially published on February 18, 2026, at 6 AM PST. It has since been updated to provide additional details regarding World Labs’s latest funding round.