Physicl raises $10 million for targeted physical AI training data
The AMW Read
The seed round and Studio launch add a data provider focused on physical AI edge cases, with impact concentrated in a specialized market.
Physicl raises $10 million for targeted physical AI training data
Physicl has raised $10 million in seed funding led by Inovia Capital, with participation from US Venture Partners. The company also introduced Studio, a tool for generating training and evaluation data around specific model failures. Studio offers 10,000 ground-truth 3D environments and more than 100,000 assets, according to the company. Physicl plans to use the funding to expand its team and develop the platform.
Physicl operates in the data infrastructure layer serving robotics and world-model developers. Its approach addresses a problem in physical AI: models can encounter rare combinations of lighting, viewpoints and objects that are difficult to capture at scale. Studio lets researchers recreate such conditions and generate variations with reference data for training and evaluation. The company says it works with Meta, World Labs, Epic Games and SpAItial. It also says SpAItial more than halved a model's room reconstruction error rate using Physicl data; the provided report does not specify the test conditions.
For builders, the practical question is whether gains on generated scenes transfer to real-world conditions. Teams can use recreated failures to test changes against known weaknesses, then check performance on real scenes. For investors, the SpAItial result is an early example of a targeted use case, though the reported result alone does not establish performance across other models or deployments.