AI foundation models are now aplenty, from OpenAI’s Stable Diffusion to OpenAI 4o. They can create realistic images based on a text prompt. Foundation models that can generate full, coherent 3D environments online from a simple text prompt are just now emerging.
It’s not a matter of if but when these models will be available. Matthias Niessner has taken a leave of absence to pursue his entrepreneurial dreams. The Technical University of Munich’s visual computing & AI Lab is looking for a startup in this area. SpAItialNiessner, a former co-founder of Synthesia – the realistic AI avatar startup valued $2.1 billion – has raised $13 million in a seed round that is unusually large for a European startup. The round was led Earlybird Venture Capital (a prominent European early-stage investment firm that has backed UiPath, Peak Games, and others), with Speedinvest and a number of high-profile angels participating.
This round size is even impressive when you consider that SpAItial hasn’t released much yet, other than a recently-released video. Teaser video demonstrating how a text-based prompt can generate a 3D environment.
Then there’s the team of technical experts that Niessner put together: Ricardo Martin Brualla who worked on Google’s 3D teleconferencing system, now called Beam, and David Novotny who spent six years leading Meta’s text-to-3D asset generation project.
The collective expertise of their team will give them an edge in a market that already has some competitors who have a similar focus. Odyssey raised $27 million to target entertainment use cases. World Labs is another startup founded by AI pioneer Fei Fei Li and valued at more than $1 billion. Niessner believes that this is a small amount of competition in comparison to other foundation models. He also thinks it’s a small amount compared to the “bigger vision” he, and others, are pursuing.
I don’t want to just have a 3D environment. I want this world to behave just like the real one. I want it to interactable, and [let you] be able to do stuff in it. Nobody has really cracked it yet,” he said.
Video games in the real world
No one has yet figured out what the demand is for photorealistic environments. The “trillion dollar” opportunity ranging between digital twins and augmented reality is big enough to excite VCs but also vague enough to make it difficult to determine a go-to-market plan. These models can be used for a variety of purposes, including video game creation. They could also be used for entertainment, 3D visualizations in construction and even in the real world, for robotic training. Niessner hopes to avoid this issue by allowing developers to license the foundation model in order to create downstream applications for specific purposes. He also recruited a fourth cofounder, Luke Rogers, a former Cazoo executive who was his roommate while he worked as a visiting assistant at Stanford. Luke Rogers will help him with the business side.
SpAItial will first identify partners who can work with older models and those who would have to wait until the quality is higher.
Niessner said that they wanted to work with at least a few partners and see how they could use the APIs.
In comparison to other well-funded AI startup, SpAItial puts revenue higher on its agenda. It will first have to spend money on both hiring and computing. The focus of the latter is quality, not quantity. Niessner says, “the team will not grow to hundreds of members right away; that’s just not going to happen, and we don’t need it.”
Niessner and co-founders instead are working to generate larger and more interactive spaces, where a glass, for example can shatter realistically. Niessner calls this the “holy-grail,” where a 10-year old could type some text and create their own video game within 10 minutes.
According to him, this ambitious goal could be more achievable than the seemingly easy task of letting users create 3D items. This is because most gaming platforms restrict what third parties can do. This is true unless the user decides to build it themselves like Roblox. SpAItial could be replacing CAD by then; the next chapter of 3D generation has just begun. Anna Heim is an editorial consultant and writer.
Anna Heim has been a writer for over 20 years. Since 2021, Anna Heim has been a freelance journalist at TechCrunch, covering a wide range of startup topics, including AI, fintech and insurtech, pricing and SaaS, as well as global venture capital trends.
Since May 2025, she has focused her reporting at TechCrunch on Europe’s most fascinating startup stories. She also writes TechCrunch Startups Weekly, a weekly newsletter that summarizes startup news. Anna has moderated panels, conducted onstage interviews, and participated in industry events of all sizes. These include major tech conferences like TechCrunch Disrupt and South Summit. She also has experience with other major conferences, such as TNW Conference and VivaTech. She is a former LATAM & Media editor at The Next Web and a Sciences Po Paris grad. She speaks multiple languages including French, English and Spanish.
View Bio
