Physicl Launches the Data Infrastructure Layer for Physical AI at NVIDIA GTC
Emerging from stealth to scale world-ready data for robotics, world models, and embodied AI
SAN JOSE, Calif., March 17, 2026 /PRNewswire/ - Physicl today emerged from stealth at NVIDIA GTC, introducing a new data infrastructure platform purpose-built for Physical AI and robotics.
Launched by members of the team behind Nfinite — the company known for building one of the largest pipelines of high-fidelity 3D digital twins and recently collaborating with Getty Images on AI-ready visual datasets — Physicl focuses exclusively on the next frontier: scaling world-ready data for embodied intelligence.
As the AI industry moves beyond language and image models toward systems that interact with space, a new bottleneck has emerged: physics-grounded, simulation-ready 3D data.
"Every major advance in AI has required a new data layer," said Alex de Vigan, CEO of Physicl. "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."
Powering the Next Wave of Physical Intelligence
Physicl supports three converging pathways:
- Robotics — training embodied systems to navigate and manipulate environments through physics-accurate simulation
- World Models — enabling spatial reasoning and causal understanding using structured 3D environments
- Vision-Language Models (VLMs) — grounding multimodal systems in physically coherent, simulation-ready data
Built on nearly a decade of experience digitizing the physical world, Physicl's infrastructure enables continuous, production-grade data pipelines, moving Physical AI beyond one-off datasets and toward scalable, world-ready infrastructure.
The platform is already supporting teams at leading technology and AI organizations, including Meta, DeepMind, World Labs, and Getty Images, integrating three tightly connected infrastructure layers:
- Data Normalization — converting visual inputs into structured, spatially consistent 3D representations
- Physics-Aware Data Augmentation — generating large-scale, IP-safe domain randomization grounded in physical realism
- DataSim Pipelines — enabling simulation-ready environments for embodied AI training and sim-to-real transfer
Physicl Leverages NVIDIA Technology: The Data Layer for the Omniverse Ecosystem
As Physical AI scales, the team is building on NVIDIA's Physical AI stack — NVIDIA Omniverse, NVIDIA Isaac Sim, NVIDIA Isaac Lab, and NVIDIA Cosmos. Physical AI developers need production-grade 3D environments to train and evaluate their models. Today, most of that data is built in-house, slowly and at high cost.
Physicl is purpose-built to fill this gap:
- Omniverse-ready assets — Simulation-ready 3D environments in OpenUSD, designed to plug directly into Omniverse workflows
- Isaac Sim & Isaac Lab — Physically-accurate environments optimized for robotic manipulation, navigation, and long-horizon task training
- Cosmos-compatible data — Structured 3D environments aligned with world foundation model post-training requirements
Physicl launches with a library of millions of simulation-ready 3D assets and environments, providing a continuous supply of simulation-ready 3D data — much like Getty Images serves licensed visual content or Scale AI provides labeled training data.
While GPUs power the models, Physical AI requires a fundamentally different type of training input than language systems.
Industry leaders have increasingly emphasized that the next generation of AI will require exponentially larger volumes of high-quality data. For systems that must interact with space, that data must also encode geometry, physics, and spatial relationships.
"NVIDIA has built the compute and simulation infrastructure for Physical AI," Alex continued. "Physicl is designed to be the dedicated data layer powering the next wave of AI and robotics"
NVIDIA GTC Presence
Physicl is showcasing its platform at NVIDIA GTC (March 16–19, San Jose, California), demonstrating how simulation-ready assets can be deployed inside NVIDIA Omniverse and Isaac Lab to accelerate robotics and Physical AI development.
Attendees can visit Physicl at Booth 3307 to learn more.
Developers and researchers interested in early beta access can apply here: https://www.physicl.ai/
About Physicl
Physicl is building the data infrastructure layer for Physical AI — enabling robots and AI models to perceive, simulate, and interact with the physical world. Physicl delivers simulation-ready 3D assets, physics-accurate data augmentation, and human-validated environments used to train robotic systems, world models, and embodied AI.
Website: https://www.physicl.ai/
SOURCE Physicl
Share this article