Physicl Secures $10 Million to Advance AI Robot Training
The Future of Robotics Lies in Digital Simulation
Physicl, a startup operating in the artificial intelligence sector, has successfully raised $10 million in seed funding. Backed by Inovia Capital, the company aims to refine the robot training process by creating highly realistic, physics-based digital environments.
Physicl founder and CEO Alex de Vigan is not a newcomer to the tech industry. He previously led nfinite, which provided 3D product modeling for brands. However, the rapid development of generative AI introduced new market demands, prompting de Vigan to shift his focus toward the field of robotics.
Why Simulation Is Decisive
Training robots remains a complex challenge. Relying on video footage is often insufficient, while collecting data in the real world is both slow and expensive. Researchers are working to teach robots how to manipulate small objects, yet their greatest hurdle is overcoming so-called "edge cases." For instance, a robot might struggle to identify and grasp a glass if it is placed in a shadowed area or on a cluttered surface.
Physicl’s platform, known as Studio, addresses this issue. The system includes over 10,000 3D environments and nearly 100,000 object models. This allows developers to create synthetic data that precisely replicates real-world physical laws, geometry, and materials. As a result, robots can undergo training within a safe, digital space.
Collaboration and Prospects
Industry specialists have already taken note of the startup's potential. SpAItial, a company that creates 3D simulations from 2D images, is among Physicl’s first users. Matthias Niessner, a professor at the Technical University of Munich and founder of SpAItial, notes that the ability to generate data automatically and in real-time is particularly attractive. When an AI model lacks specific data during training, the Physicl platform can generate it instantly.
Physicl does not intend to remain solely a data provider. The company’s long-term vision is to become a universal tool where users can input millions of hours of data—such as LIDAR scans—and convert them into high-quality 3D models. This approach will enable the generation of billions of variations, helping robots achieve human-like dexterity.
Although the robotics field remains in its early stages and is highly competitive, Physicl’s technology represents a significant step toward bridging the digital and physical worlds. For technology enthusiasts, this is a sign that the robots capable of solving our everyday household and industrial problems are moving ever closer to reality.