Physicl Raises $10 Million to Close Sim-to-Real Gap for Robot Training

Physicl has secured $10 million in seed funding, led by Inovia Capital, to help bridge the sim-to-real gap that frustrates robotics developers, according to TechCrunch Robotics. The company, founded by CEO Alex de Vigan, aims to deliver high-fidelity digital environments that make simulated robotic training more reflective of real-world complexity. De Vigan and his team previously built nfinite, a firm known for generating high-quality 3D e-commerce content, pivoting their tech stack after generative AI disrupted their initial business.

The Physicl platform is designed to convert unstructured real-world data—like images and lidar—into accurate, physically correct 3D environments. This allows for the generation of synthetic training datasets with precise geometry, materials, and physics. Physicl distinguishes itself by enabling customers to produce billions of permutations for model training and auto-generate targeted data when current training routines show deficiencies. The company sees itself not only as a data supplier but as a provider of ongoing synthetic data tooling for robot and AI developers.

To demonstrate impact, Physicl launched Studio, a product shipping with 10,000 3D environments and 100,000 object assets. Studio lets researchers and integrators build highly specific scenes to train robots, addressing challenges like clutter, shadowing, and fine-grained object manipulation. An early collaboration with world model developer SpAItial, which used Studio’s resources to improve its Echo-3 simulator, showcased Physicl's value for next-generation simulation requirements; SpAItial’s CEO Matthias Niessner said high-fidelity synthetic data was vital for their system’s training performance, as TechCrunch Robotics reported.

De Vigan characterizes the market as young and crowded, with many startups collecting, labeling, or generating training data across diverse formats and modalities. Physicl’s focus is on easing ongoing experimentation, reducing the hand-curation bottleneck with pipelines that automate high-volume, fine-grained data synthesis. Industry demand is driven by the persistent shortcomings of training physical AI on video alone and the prohibitive cost and latency of labeling real-world sensor data. Physicl positions itself for users who want iterative control over their data creation—whether training foundation models or testing operational edge cases.

Why it matters. Physicl’s entry intensifies competition among synthetic data platform providers, raising the bar for realism in digital simulation—a key if robots are to meet advertised performance levels outside the lab. If the firm’s approach scales, robot model builders and world simulation companies could see reduced training cycles and higher transferability to the real world, but entrenched labels and simulation incumbents may be pressured to rethink their pipelines. Deployment and customer results in 2027 will be critical to gauge adoption and efficacy, as TechCrunch Robotics documents. Reported

First published in [Robotics Weekly — 2026-11-09](https://roboticsweekly.com/newsletter/robotics-weekly-2026-11-09).