Physicl Raises $10M to Build Realistic 3D Environments for Training Physical AI Models

Physicl Secures Seed Funding to Transform Robot Training

Physicl, a Paris-based startup specializing in training data for physical AI, has raised $10 million in seed funding led by Inovia Capital, with participation from U.S. Venture Partners2

2

. The company aims to solve critical challenges in making robots more capable by providing high-fidelity 3D environments for training AI models for robots on difficult real-world scenarios1

1

.

Founded by Alex de Vigan, who previously built nfinite—a company that raised $100 million in 2022 for creating 3D product renderings for e-commerce—Physicl repurposes a decade of expertise in 3D rendering technology to address a gap in AI development1

1

2

2

. De Vigan recognized that while generative AI could replace traditional 3D visualization, it struggled with understanding and manipulating physical space and objects—a critical limitation for robotics applications.

Addressing Edge Cases with Studio Platform

The startup's flagship product, Studio, directly tackles what researchers call edge cases—unusual situations where AI models fail, such as odd lighting conditions, unexpected object placement, or cluttered environments2

2

. Studio provides 10,000 3D environments and over 100,000 assets that can be combined into billions of different scenes2

2

.

Researchers can use the platform to rebuild specific scenarios where their models struggle, then adjust lighting, camera angles, and object positions to generate synthetic training data for robots targeting those exact weaknesses. For instance, Studio could help train a robot having difficulty identifying a cup on a cluttered countertop or recognizing objects in shadow1

1

.

"Models fail on edge cases. When researchers find these weaknesses, they need a way to generate data that specifically addresses them," de Vigan explained2

2

. This targeted approach to creating training data for physical AI represents a shift from broad data collection to precision problem-solving.

Proven Results with Early Customers

Physicl has already attracted notable clients including Meta, Nvidia, World Labs, Epic Games, Amazon Web Services, and the Technical University of Munich2

2

. The startup's impact is demonstrated through its collaboration with SpAItial, a Munich-based world model builder that develops 3D AI models from 2D imagery.

SpAItial, led by CEO and founder Matthias Niessner—also a professor of robotics at the Technical University of Munich—used Physicl's synthetic training data for robots to improve its latest release, Echo-31

1

. After incorporating Physicl's data, SpAItial cut the error rate of its room-building model by more than half2

2

.

Niessner highlighted the platform's efficiency advantage: "The real challenge is, you have to curate the data by hand, and this is a very long cycle. To us what seems very attractive is you can just do this on the fly"1

1

.

Beyond Data Provider to Agentic Workflow

While Physicl operates in a crowded space of startups providing data and tooling to foundation model builders and robot operators, de Vigan positions the company as more than just a data provider. "The space is very young and very blurry sometimes," he acknowledges, but Physicl's unique value lies in its ability to transform unstructured information into high-fidelity physical representations with correct geometry, materials, and physics1

1

.

De Vigan envisions an agentic workflow where customers feed trillions of images or millions of hours of LIDAR data into the platform, which then generates billions of permutations for training. Niessner sees particular promise in autonomous data generation: "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"1

1

.

Physicl will use the capital to expand its team and develop the platform further2

2

. The startup joins a growing wave of European companies working on physical AI technology that enables machines to understand and act effectively in real-world environments, addressing applications from robotics models to self-driving cars and 3D vision systems.