Physicl Raises $10M Seed to Turn 3D Rendering Into Robot Training Fuel

Physicl has raised $10 million in seed funding from Inovia Capital to build more realistic simulations for training robots. The startup wants to convert messy real-world data into high-fidelity 3D environments that robot makers can mine for synthetic training data.

Founder and CEO Alex de Vigan spent the last decade building nfinite, a company that helped brands and retailers create 3D models of their products for online display. Generative AI changed that market. Transformer-based models pushed nfinite toward generative imagery, but de Vigan saw a new use for the rendering pipeline his team had already built.

"What we were super unique about doing in 3D, AI was going to replace, but something was missing in AI: its ability to understand, interact, and manipulate space and objects," de Vigan said.

Why do robot developers need better simulations?

The problem sits at the center of physical AI research. Robots still struggle with tasks like handling diverse, small objects. Training models on video has produced frustrating returns. Gathering data through remote operation — humans piloting robots to collect examples — is expensive.

Simulated digital worlds look like the fix, but only if they obey real-world physics. That has proven difficult. De Vigan argues the pipeline his previous company built can now do the job: it transforms "unstructured, incomplete, heterogeneous real-world information into a high-fidelity physical representation… with correct geometry, materials, and physics."

The result, he says, is training material that behaves the way real objects behave.

What is Physicl actually selling?

The company is entering a crowded category. Startups are collecting, labeling and generating data in different modalities as customers hunt for the formula that makes robots deliver on their promise.

"The space is very young and and and very blurry sometimes," de Vigan admits.

Physicl is focusing on post-training models on difficult edge cases. Its first product, Studio, ships with:

• 10,000 3D environments
• 100,000 assets representing various objects
• Tools to build custom scenes that generate synthetic data for specific problems

One use case: a robot that fails to detect a cup on a cluttered counter-top, or spots it only when shadow falls across the scene. Studio lets engineers generate exactly those variations at scale.

De Vigan positions Physicl as more than a data vendor. He calls it tooling for ongoing synthetic data generation. "What we created 10 years ago, without knowing it, was basically an agentic workflow to transform unstructured information into high fidelity 3D," he said. He expects customers to bring trillions of images or millions of hours of LIDAR capture, convert them into high-quality 3D representations, and spin off billions of permutations to train their models.

Who is already using it?

SpAItial, a world model builder, worked with Physicl as an early customer to improve the fidelity of its products. Its latest tool, Echo-3, was released last week and generates 3D simulations from 2D imagery.

SpAItial CEO and founder Matthias Niessner, also a professor of robotics at the Technical University of Munich, said his company needed high-fidelity synthetic data of the kind Physicl provides to train that conversion process effectively.

The manual alternative is slow. "The real challenge is, you have to curate the data by hand, and this is a very long cycle," Niessner said. "To us what seems very attractive is you can just do this on the fly […] You're running Claude, and it checks, 'Oh, I need more data of this in my training,' and then it auto generates that data, depending on the training state. That, to me, is super exciting, and this is exactly what Physicl is offering."

What comes next?

Physicl's bet is that the robot industry's bottleneck shifts from model architecture to data supply, and that on-demand synthetic worlds become standard infrastructure for physical AI teams. If de Vigan's decade-old rendering pipeline proves it can scale from product catalogues to trillions of images, the $10 million seed round may mark the point where e-commerce 3D tech found its second life in robotics.