World Labs has announced a $1 billion funding round, marking one of the largest capital infusions yet into the emerging category of spatial intelligence. The round includes strategic and financial backing from Autodesk, NVIDIA, and AMD, among others, signaling conviction that AI systems capable of reasoning about real, three-dimensional worlds will define the next phase of computing.

World Labs’ founding team brings deep roots in neural rendering and radiance field research. The startup, co-founded by NeRF first author Ben Mildenhall, has quickly ramped up its capabilities, bringing promptable generative 3D, which they deliver to consumers through Gaussian splatting.

At the center of World Labs’ strategy is Marble, a system designed to generate spatially coherent, persistent 3D environments from images, video, or text. Just recently, the team released an API for Marble and also populated users’ accounts with test credits. While the company frames Marble broadly as a “world model,” its trajectory increasingly overlaps with one of the most practical representation formats to emerge in the last two years, gaussian splatting.

The scale of this funding round suggests World Labs is continuing to move toward infrastructure. Gaussian splatting fits that shift in several ways.

As Marble targets storytelling, simulation, and robotics, explorable worlds provide tremendous value. Additionally, Gaussian splats’ compatibility with modern GPU pipelines aligns closely with the interests of partners like NVIDIA and AMD.

The combination of capital, research pedigree, and representation choices suggests a clear direction. Gaussian splatting is rapidly becoming a foundational layer for spatial AI systems that need to see, remember, and operate inside real-world geometry.

With $1 billion in new funding, World Labs now has the runway to turn that representation into a platform—and, potentially, into one building the strongest world modeling effort of this decade. Congrats Ben and team!

Try Marble here.