AI models that turn text prompts into realistic images are now common. The harder frontier is building foundation models that can generate complete, coherent 3D online environments from text.
SpAItial is entering that frontier with an unusually large seed round for a European startup. Founded by Matthias Niessner, one of Europe’s most prominent AI 3D model researchers, the company has raised $13 million to pursue text-to-3D worlds that are not only visual, but eventually interactive.
Why SpAItial is attracting attention early
SpAItial does not yet have a large public product portfolio. What it has shown so far is a recently released teaser video demonstrating how a text prompt could generate a 3D room.
Even so, the company has secured a seed round led by Earlybird Venture Capital, with participation from Speedinvest and several high-profile angels. Earlybird Venture Capital is described as a prominent European early-stage investor and has backed UiPath through its Digital East Fund, as well as Peak Games.
The funding is notable because SpAItial is still early. The bet is less about a finished product and more about the team, the market opening, and the belief that 3D foundation models are moving from research into commercial reality.
Niessner has taken an entrepreneurial leave of absence from his visual computing & AI lab at the Technical University of Munich to build the company. He was also formerly a co-founder at Synthesia, the realistic AI avatar startup valued at $2.1 billion.
The team is built around 3D AI experience
SpAItial’s founding team brings together people who have worked directly on advanced 3D and AI systems. Ricardo Martin-Brualla previously worked on Google’s 3D teleconferencing platform, now called Beam. David Novotny spent six years at Meta, where he led the company’s text-to-3D asset generation project.
Luke Rogers, a former Cazoo executive, is also a co-founder and is helping on the business side. Rogers and Niessner were once roommates in Palo Alto while Niessner was a visiting assistant professor at Stanford.
That mix matters because the challenge is not only to create good-looking 3D spaces. The larger ambition is to generate environments that can support interaction, realism, and eventually useful applications across different industries.
Niessner framed the problem clearly: “I don’t just want to have a 3D world. I also want this world to behave like the real world. I want it to be interactable and [let you] do stuff in it, and nobody has really cracked that yet,” he said.
The market is promising but still undefined
The potential demand for photorealistic 3D environments is not fully proven. The source describes a broad promise around a “trillion-dollar” opportunity ranging from digital twins to augmented reality, but also notes that this breadth can make go-to-market strategy difficult.
The most obvious application is video game creation. If a model can generate a usable 3D space from a prompt, it could change how game worlds are created and who can create them.
But gaming is not the only possible use. The technology could also apply to entertainment, 3D visualizations used in construction, and eventually real-world use cases such as robotic training.
For now, SpAItial is trying to avoid narrowing too early. Niessner wants developers to license the foundation model and build downstream applications for specific needs.
That approach makes the model itself the platform. Instead of trying to guess every final product, SpAItial can work with partners and learn where early model quality is already useful and where higher quality will be required.
Partners and APIs come before mass hiring
One of SpAItial’s early roadmap tasks is to identify partners that can work with earlier models. Some potential users may be able to experiment soon, while others may need to wait for more capable output.
Niessner said, “We want to at least work with a few partners,” and added, “and see how they can use the APIs.”
Compared with other well-funded AI startups, SpAItial is placing revenue higher on its agenda. But before revenue can become meaningful, the company will need to spend on compute and hiring.
The hiring plan is focused on quality rather than scale. According to Niessner, “the team is not going to grow to hundreds of people right away; it’s just not happening, and we don’t need that.”
That restraint is important in a field where compute costs and technical hiring can rise quickly. SpAItial’s approach suggests a company trying to stay close to research while testing commercial demand through partners and APIs.
From 3D rooms to interactive worlds
SpAItial is not alone. Odyssey has raised $27 million and is pursuing entertainment use cases. World Labs, founded by AI pioneer Fei-Fei Li, is already valued at over $1 billion.
Niessner sees the field as less crowded than other foundation model categories, especially when measured against the bigger goal he is pursuing. The target is not just generating a 3D scene, but making that scene act in ways users can manipulate.
The company is working on larger and more interactive 3D spaces. One example from the source is a glass that can shatter realistically. That kind of behavior points to a different level of world generation than static 3D output.
Niessner calls the larger goal the “holy grail”: a 10-year-old could type in some text and make their own video game in 10 minutes. In his view, that may be more achievable than simply letting users create 3D objects, because most gaming platforms still tightly control what third parties can add.
There is another possible direction as well. If gaming platforms build similar tools themselves, SpAItial could end up focusing on areas such as replacing CAD. The next chapter in 3D generation is still open, and SpAItial is positioning itself early in the race.