World Labs is moving deeper into the commercial world of 3D creation. The company has secured a $200 million investment from Autodesk as part of a larger $1 billion round, setting up a partnership focused on how AI systems that understand 3D environments could fit into professional design workflows.
The deal links one of the best-known names in 3D CAD software with a startup building world models, a category of AI designed to generate and reason about immersive 3D environments. For designers, builders, and creators, the central question is practical: can these systems become useful inside the tools where 3D work already happens?
The investment puts World Labs closer to commercial 3D software
Autodesk’s $200 million investment is part of a larger $1 billion round that includes backers such as AMD, Emerson Collective, Fidelity, Nvidia, and others. World Labs, founded by Fei-Fei Li, emerged from stealth in 2024 with $230 million at a $1 billion valuation.
The company declined to say whether the latest financing changed its valuation. Reports a month ago suggested it was aiming to raise at a $5 billion valuation, but the confirmed fact is the size of the new round and Autodesk’s role in it.
For World Labs, the Autodesk relationship is more than capital. It is a signal that the company’s world model technology may have a path into established professional markets, not only experimental AI demos.
World Labs released its first world model product, Marble, last November. Marble lets users create editable, downloadable 3D environments, which is directly relevant to the kinds of workflows Autodesk already supports.
Why Autodesk is a natural partner
Autodesk is one of the biggest developers of 3D CAD software. Its platform is used across architectural, engineering, construction, manufacturing, and entertainment workflows, all of which depend on digital representations of objects, spaces, and systems.
That makes spatial AI a logical area for Autodesk to watch closely. World models are not just image generators. In this context, they are meant to understand the relationships among geometry, physics, and dynamics inside a scene.
Li framed the partnership around that shared spatial focus, saying: “Autodesk has long helped people think spatially and solve real-world problems and, together, we share a clear purpose: building physical AI that augments human creativity and puts more powerful tools in the hands of designers, builders, and creators.”
As part of the deal, Autodesk will serve as an adviser to World Labs. The companies will also collaborate at the “research and model level,” though the exact shape of that work has not yet been finalized.
How world models could fit into design workflows
Daron Green, Autodesk’s chief scientist, told TechCrunch that the partnership is still in its early days. That matters because the companies are not presenting a finished product integration yet. They are exploring how their systems might interact.
One possibility is a workflow that starts with a world-model-based sketch in World Labs, such as an office layout, then moves into Autodesk tools for more detailed design work, such as the design of a desk. Another possibility is the reverse: a user could take an object designed in Autodesk’s platform and place it inside a context generated through a World Labs prompt.
Green described that flexibility directly: “You could anticipate us consuming their models or them consuming our models in different settings.”
He also said data sharing is not part of the agreement. That is an important boundary for a partnership involving professional design tools, AI models, and customer workflows.
Entertainment is the first testing ground
The companies plan to begin with media and entertainment use cases. That starting point fits a broader pattern among world model companies, including Google DeepMind and Runway, which view gaming and interactive entertainment as an initial go-to-market strategy.
Autodesk already works with most major media production companies and has been training models for character animation. Green said those systems are “close to world models” because they describe how a character, such as an animal, responds to constraints in a world.
His example points to the next step: not only animating a character, but giving that character a world it can interact with. In that framing, world models could expand animation and entertainment tools from isolated assets toward richer environments.
That does not mean the partnership is limited to entertainment forever. It means the first use cases are likely to appear where interactive 3D environments are already central to the work.
Autodesk’s larger AI direction
The World Labs partnership also fits Autodesk’s broader effort to add more AI features across its software portfolio. The company is developing “neural CAD,” a generative AI model trained on geometric data.
Neural CAD is intended to reason about components and entire systems. In plain terms, it is designed to generate working 3D models, not just images, with an understanding of how those designs would function in the real world.
Autodesk’s neural CAD models are already being integrated into the company’s product design and architecture products. World Labs’ models could extend that work beyond individual design files and toward more complete digital representations of physical environments.
Green expects different AI systems, including large language models, world models, and neural CAD, to be combined in the future to improve designs for Autodesk customers.
Li described the broader technical challenge this way: “If AI is to be truly useful, it must understand worlds, not just words.” She added: “Worlds are governed by geometry, physics, and dynamics, and reconciling the semantic, spatial, and physical is the next great frontier of AI.”
The partnership is still early, but its direction is clear. World Labs is trying to bring world models into usable 3D environments, while Autodesk is looking for ways to make AI more spatial, functional, and relevant to the people who build digital versions of the physical world.