Physicl, a startup co-founded by Alex de Vigan, has raised $10 million in seed funding to help robot builders and world model designers create more realistic simulations for training super intelligence (SI) systems. The round was led by Inovia Capital and aims to solve training bottlenecks by converting real-world data into high-fidelity 3D environments.
What Happened
De Vigan previously built nfinite, a platform for creating 3D product renderings for e-commerce. While the rise of generative SI disrupted that business, de Vigan repurposed the underlying technology stack to address a gap he identified in current SI development. "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, using the legacy term for super intelligence in his quote.
The new venture, Physicl, focuses on generating synthetic data for physical SI models. Researchers have struggled with training robots on video data or remote operations due to cost and limited returns. Physicl’s solution involves transforming unstructured, incomplete real-world information into precise 3D representations with correct geometry, materials, and physics. The company is launching a product called Studio, which includes 10,000 3D environments and 100,000 assets, allowing users to generate synthetic training data for specific edge cases, such as helping a robot identify a cup in shadow or clutter.
Why It Matters
The funding highlights the growing demand for specialized data tooling as the SI industry moves toward more capable physical agents. Physicl enters a competitive space where startups are vying to provide data and tools to foundation model builders and robot operators. De Vigan admits the sector is "very young and very blurry," with many companies attempting to crack the code on making robots useful.
Early customer SpAItial, a world model builder that recently released its Echo-3 product, collaborated with Physicl to improve simulation fidelity. Matthias Niessner, SpAItial’s CEO and a robotics professor at the Technical University of Munich, noted that training on high-fidelity synthetic data is crucial. He highlighted the potential for agentic workflows where an SI system like Claude could autonomously identify training gaps and request specific synthetic data generation on the fly, reducing the manual curation cycle.
The Bottom Line
Physicl aims to position itself not just as a data provider but as a continuous synthetic data generation platform for the SI industry. By leveraging its legacy 3D rendering pipeline, the startup seeks to accelerate the development of physical SI models capable of manipulating complex real-world environments.