From OpenAI’s 4o to image models like Stable Diffusion, AI systems that create lifelike visuals from text prompts are now everywhere. But there’s one frontier that’s still emerging: AI 3D environments generation, where entire interactive spaces are conjured from a line of text. Now, one of Europe’s top minds in 3D AI is setting out to tackle it. Matthias Niessner, a renowned professor at the Technical University of Munich and co-founder of Synthesia, has launched a new startup called SpAItial.
With a vision to make fully immersive, responsive virtual worlds from simple prompts, SpAItial just raised a massive $13 million seed round, led by Earlybird Venture Capital and backed by Speedinvest and several prominent angel investors.
That’s a big number for a European seed round—especially for a startup that has only teased what it’s building.
Building the 3D Foundation Model Layer
SpAItial’s first teaser showed a text prompt generating a photorealistic 3D room. But what’s more impressive is the team behind the scenes. Niessner has recruited former Google and Meta researchers: Ricardo Martin-Brualla, who worked on Google’s 3D telepresence tool Beam, and David Novotny, who led Meta’s text-to-3D asset generation project.
Together, they aim to go beyond static renderings. Niessner wants the virtual spaces to behave like the real world—objects that break, react, or move when touched, and environments that users can actually interact with, not just view. “I don’t just want a 3D world,” he said. “I want one where you can do things.”
This sets SpAItial apart from other players in the space like Odyssey, which raised $27 million for gaming-focused tools, and World Labs, founded by AI legend Fei-Fei Li, already valued at over $1 billion. Despite the competition, Niessner believes that 3D AI models are still in their infancy compared to the mature ecosystems around text, image, and audio generation.
From Digital Twins to Video Games: Use Cases Are Still Shaping Up
The commercial potential of AI 3D environment generation is huge but not yet fully defined. Investors are betting on applications in gaming, entertainment, architecture, robotics, and even industrial simulations like digital twins and smart city planning. But unlike image generation, which has immediate applications in content creation, 3D output still faces hurdles—especially platform restrictions on asset imports and the complexity of interactivity.
To navigate this, SpAItial plans to license its foundation model to developers who can tailor it to specific use cases. One of the startup’s first goals is to identify early partners who can work with its initial APIs. Business operations will be led by Luke Rogers, a former Cazoo exec and longtime friend of Niessner from their Stanford days.
“We’re not trying to grow to 200 people overnight,” Niessner said. “This is about talent and precision. We’re focused on quality—on both compute and hiring.”
Revenue and Realism are Priorities from Day One
While many AI startups focus solely on model training in their early phases, SpAItial is prioritizing revenue generation early. The company aims to build scalable licensing models while investing in the heavy compute infrastructure needed for realistic, interactive rendering.
Ultimately, the vision is bold: Let anyone, even a 10-year-old, type in a prompt and build their own immersive video game in minutes. That goal—generating vast, detailed, and physics-aware 3D environments—could also disrupt traditional CAD systems, not just game engines.
And while gaming giants like Roblox might one day integrate similar tools, Niessner’s bet is that open innovation will move faster.
“Replacing CAD might come before Roblox even opens up its platform fully,” he said. “We’re building the next generation of virtual creation tools.”
With one of Europe’s largest early-stage rounds in the space and a world-class team behind it, SpAItial is poised to reshape how we create and interact with 3D environments—one prompt at a time.
