Insider Brief
- Physicl emerged from stealth at NVIDIA GTC with a platform focused on providing scalable, simulation-ready 3D data as a foundational layer for robotics and physical AI systems.
- The company’s infrastructure supports robotics, world models and vision-language models through physics-aware data pipelines, including 3D normalization, synthetic data generation and simulation environments for training and deployment.
- Physicl is integrating with NVIDIA’s Omniverse and Isaac stack and launching with millions of 3D assets, positioning itself as a data provider for physical AI as developers seek to scale beyond in-house dataset creation.
Physicl has emerged from stealth with a platform focused on what it said is a growing constraint in robotics and physical AI: the lack of scalable, simulation-ready 3D data.
Announced at NVIDIA GTC, the company said it is building a data infrastructure layer designed to support AI systems that operate in the physical world. Physicl is led by members of the team behind Nfinite and is focused on producing structured, physics-aware data that can be used to train robots and spatial AI models.
What Does Physicl Do?
Physicl said the shift from language and image models to systems that interact with real environments is exposing a data gap. Unlike text or images, physical AI systems require data that captures geometry, physics and spatial relationships, which is more difficult to generate and standardize.
“Every major advance in AI has required a new data layer,” CEO Alex de Vigan said in the announcement. “For Physical AI, that missing layer is structured, spatially consistent, physics-aware data that models can actually learn from. Physicl exists to build that foundation — enabling robots and world models to understand space, simulate environments, and ultimately operate reliably in the real world.”
According to Physicl, its platform is designed to support three core areas — robotics, world models and vision-language models — and the platform is already being used by organizations including Meta, DeepMind, World Labs and Getty Images.
At the core of the system are three infrastructure layers:
- Data normalization: Converting visual inputs into structured 3D representations
- Physics-aware augmentation: generating synthetic data grounded in physical realism
- Simulation pipelines: creating environments that can be used for training and sim-to-real transfer
Physicl is building its platform to integrate with Nvidia’s physical AI stack, including Omniverse, Isaac Sim, Isaac Lab and Cosmos. The company said most teams today still build 3D training environments in-house, a process that is time-consuming and difficult to scale.
To address that, Physicl said it is offering simulation-ready 3D environments designed for direct use in existing workflows, along with physically accurate environments tailored for robotic manipulation, navigation and longer-duration tasks. The platform also includes structured datasets aligned with the post-training requirements of large-scale world models.
It is also launching with a library of millions of simulation-ready 3D assets to serve as a data provider for physical AI, similar to image libraries and labeling platforms.
Physicl said beta early access is available to developers and it is demonstrating the platform at NVIDIA GTC, with a focus on how developers can use simulation-ready data inside Omniverse and Isaac-based tools to accelerate robotics development.
