The Daily Ledger

StartupsNews

Physicl raises $10M for robot training simulations

The startup, led by nfinite veteran Alex de Vigan, has closed a $10 million seed round and launched Studio, a library of 3D environments for generating synthetic robot-training data.

The Daily Ledger Staff · Updated 2 min read

Seed funding for simulation

Physicl, a startup building realistic simulations for robots, has raised $10 million in seed funding led by Inovia Capital, TechCrunch reported on 6 November 2026. The company aims to help designers of world models and builders of robots tackle the problems they face when training physical AI systems.

The training problem

According to the report, researchers building physical AI models have been frustrated by two approaches. Training on video yields poor returns, while collecting data through remote operation is expensive. Simulation is seen as promising, but making it faithful to real-world physics is difficult.

Physicl's founder and chief executive, Alex de Vigan, previously spent about a decade building nfinite, a company that let brands and retailers create 3D models of their products for online display. When transformer-based AI models arrived, nfinite moved toward generative imagery. De Vigan saw value in the 3D rendering stack his team had built, and said that although AI "was going to replace" that work, "something was missing in AI."

What Studio offers

Physicl is releasing a product called Studio, which includes 10,000 3D environments and 100,000 assets representing different objects. Users can assemble scenes to generate synthetic training data for specific problems. TechCrunch gives the example of a robot that struggles to spot a cup on a cluttered countertop or in shadow.

The company focuses on post-training models on difficult edge cases. De Vigan acknowledged that the field is busy, saying "The space is very young and very blurry sometimes."

An early customer

SpAItial, which builds world models that generate 3D simulations from 2D imagery, was an early customer. Its latest product, Echo-3, was released the week before the report. SpAItial's chief executive, Matthias Niessner, who is also a robotics professor at the Technical University of Munich, said the company needed high-fidelity synthetic data to train its tool.

Niessner said the real challenge is that data must be curated by hand, which is a very long cycle. He described a future in which a model works out what training data it needs and generates it automatically during training, calling that prospect "super exciting" and saying Physicl offers exactly that.

Looking ahead

De Vigan describes Physicl's earlier pipeline as an agentic workflow that turns unstructured information into high-fidelity 3D. He imagines customers feeding in trillions of images or millions of hours of LiDAR data to generate billions of training permutations. Whether simulation can close the gap with real-world data remains an open question for the industry.

Sources

This story was prepared with the help of automated drafting tools and was reviewed and fact-checked by our editors before publication. See our Editorial Policy.

Report a problem with this story · Copyright & takedown