Physical AI startup Physicl has raised $10 million in seed funding led by Inovia Capital to develop realistic 3D simulations that help robots learn to navigate environments and manipulate objects. The company is also introducing Studio, a platform containing 10,000 digital environments and 100,000 object assets for generating synthetic training data.
The funding round, which also included participation from US Venture Partners, was announced on November 6, 2026. Physicl aims to help robotics developers address training limitations by recreating physical environments with accurate geometry, materials, and interactions.
Physicl Converts E-Commerce Technology Into Robotics Training Tools
Founder and CEO Alex de Vigan previously spent a decade building nfinite, a company that developed 3D product visualization technology for retailers and e-commerce businesses. As generative AI transformed the market for product imagery, de Vigan identified an opportunity to apply the company’s 3D rendering capabilities to robotics.
The technology originally developed for nfinite processes incomplete and unstructured information to create detailed digital representations of physical objects and environments. Physicl is adapting that infrastructure to help AI models understand spatial relationships and physical interactions.
Training robots using recorded video presents limitations, particularly when teaching machines to recognize and manipulate unfamiliar objects. Collecting real-world training data through remotely operated robots can also be expensive and time-consuming.
Studio Offers 10,000 Simulated Environments
Physicl’s new Studio platform allows developers to construct customized scenarios using its library of 3D environments and objects. The system generates synthetic datasets that can be used to train AI models on situations where their existing performance is unreliable.
For example, developers could recreate a cluttered kitchen counter where a robot struggles to identify a cup, then generate variations involving different lighting conditions, object positions, and surrounding obstacles. These datasets can help models learn to handle difficult situations without requiring extensive physical testing.
The company is particularly focused on post-training, where developers refine existing AI models to address specific weaknesses. Physicl expects its tools to support ongoing data generation as new problems are identified during training.
SpAItial Adopts Physicl’s Synthetic Data Technology
World-model developer SpAItial is among Physicl’s early customers, using its synthetic data to improve the accuracy of generated 3D environments. SpAItial recently introduced Echo-3, a system that creates three-dimensional simulations from two-dimensional images.
SpAItial founder and CEO Matthias Niessner, who is also a robotics professor at the Technical University of Munich, said high-quality synthetic data is important for training models that reconstruct physical environments.
Physicl plans to develop tools capable of processing large collections of images and LiDAR recordings into detailed 3D representations. These representations could then support automated generation of training scenarios tailored to the changing requirements of individual AI models.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.