Fei-Fei Li’s startup World Labs has secured a $200 million investment from Autodesk as part of a broader $1 billion funding round that includes backing from AMD, Emerson Collective, Fidelity, Nvidia, and other investors.
World Labs declined to disclose its latest valuation. The company previously emerged from stealth in 2024 with $230 million raised at a $1 billion valuation. Reports last month suggested it was targeting a valuation of around $5 billion in the new round.
The investment formalizes a partnership between World Labs and Autodesk focused on spatial AI. The two companies plan to explore how World Labs’ world models, AI systems capable of generating and reasoning about immersive 3D environments, can integrate with Autodesk’s design tools. Initial efforts will center on media and entertainment use cases.
World Labs’ first product, Marble, launched in November. It allows users to create editable and downloadable 3D environments generated by AI. The technology is designed to move beyond image generation by modeling geometry, physics, and spatial relationships in ways that enable interactive and functional environments.
Linking World Models With Design Software
Autodesk is one of the largest developers of 3D computer-aided design software, with tools widely used in architecture, engineering, construction, manufacturing, and entertainment. Its platform underpins workflows for designing buildings, products, and animated content.
Daron Green, Autodesk’s chief scientist, said the partnership remains in early stages. Potential integrations could involve Autodesk consuming World Labs’ models or vice versa. For example, users might sketch an office layout using a world model and then refine specific objects, such as furniture components, using Autodesk’s precision design tools.
Conversely, objects designed within Autodesk’s software could be placed into AI-generated environments created through World Labs’ prompts. Green said the agreement does not involve data sharing between the companies.
Autodesk already develops AI models for character animation, which Green described as conceptually close to world models. These systems simulate how characters move within physical constraints such as terrain and time. Integrating broader environmental reasoning could allow animated entities to interact with dynamic digital worlds rather than isolated scenes.
Toward Physical AI
The collaboration aligns with Autodesk’s push into what it calls neural CAD, generative AI models trained on geometric data that can produce functional 3D designs with an understanding of structural and physical constraints. These models are being integrated into Autodesk’s product design and architecture software.
World Labs’ technology could extend that capability beyond individual design files toward comprehensive digital representations of physical spaces. Li has described the effort as building “physical AI” that reconciles language-based reasoning with spatial and physical understanding.
Industry peers, including Google DeepMind and Runway, are also investing in world models, often targeting gaming and interactive entertainment as initial markets.
For World Labs, Autodesk’s investment signals commercial validation from a major software provider embedded in global design workflows. For Autodesk, the deal represents a step toward combining large language models, world models, and geometric AI systems into unified tools that can reason not only about text, but about physical environments and functional design.
