NVIDIA Blueprints Scale Physical AI with Simulation

The development of physical AI systems—robots, autonomous vehicles, and smart factories—faces a fundamental scaling . Real-world data, while valuable, is messy, unpredictable, and full of edge cases, creating a bottleneck that extends beyond data collection to the entire data processing pipeline. At NVIDIA's recent GTC conference, the company addressed this by introducing new blueprints and frameworks designed to transform how these systems are built, tested, and deployed at an enterprise scale.

Central to this shift are new frontier models for physical AI, including NVIDIA Cosmos 3, NVIDIA Isaac GR00T N1.7, and NVIDIA Alpamayo 1.5. NVIDIA also released two critical blueprints: the Physical AI Data Factory Blueprint and the Omniverse DSX Blueprint for AI factory digital twin simulation. These are not standalone products but open reference architectures designed to push the state of the art in world modeling, humanoid skills, and autonomous driving by providing structured ologies for development.

The Physical AI Data Factory Blueprint is an open reference architecture built on NVIDIA Cosmos open world foundation models and the NVIDIA OSMO operator. It transforms compute into large-scale, high-quality training data by unifying data curation, augmentation, and evaluation into a single pipeline. This enables developers to generate diverse, long-tail datasets from limited real-world inputs, effectively turning world-scale compute into turnkey data production engines. Microsoft Azure and Nebius are the first cloud platforms to offer this blueprint.

Parallel to data generation, the Omniverse DSX Blueprint provides a reference architecture that unifies simulation across every layer of an AI factory through a single digital twin. Modern AI factories are complex systems spanning thermals, power grids, network load, and mechanical systems. This blueprint enables operators to optimize performance and efficiency in a physically accurate virtual environment before any physical infrastructure is installed. KION, working with Accenture and Siemens, is using a related architecture, the NVIDIA Mega Omniverse Blueprint, to build large-scale warehouse digital twins for training fleets of autonomous forklifts.

A critical technical step enabling this simulation-driven approach is the conversion of CAD files to OpenUSD. OpenUSD provides a common scene-description language that lets teams bring CAD data, simulation assets, and real-world telemetry into a shared, physically accurate view of the world. Using tools like the NVIDIA Omniverse Kit SDK and NVIDIA Isaac Sim, teams optimize 3D data for real-time rendering, simulation, and collaborative workflows. Companies like FANUC and Fauna Robotics use this CAD-to-OpenUSD workflow to speed up robotic system design and validation.

The ecosystem adoption of these s is significant. Leading physical AI developers including FieldAI, Hexagon Robotics, Linker Vision, Milestone Systems, Skild AI, and Teradyne Robotics are already using the Physical AI Data Factory Blueprint to accelerate robotics projects, vision AI agents, and autonomous vehicle programs. Furthermore, industrial robot giants ABB Robotics, FANUC, KUKA, and Yaskawa—with a combined install base of over 2 million robots—use NVIDIA Omniverse libraries and Isaac simulation frameworks to validate complex applications through digital twins and have integrated NVIDIA Jetson modules for real-time AI inference.

Robot brain development is also leveraging this ology. Developers like FieldAI and Skild AI are building their systems using NVIDIA Cosmos world models for data generation and Isaac simulation frameworks to validate policies in simulation. Generalist AI is exploring synthetic data generation with NVIDIA Cosmos. This combined approach of simulated data generation and policy validation allows robots to achieve proficiency in diverse tasks, from supply chain monitoring to food delivery, at an accelerated pace.

These developments represent a ological shift where compute itself becomes data. As Rev Lebaredian, vice president of Omniverse and simulation technologies at NVIDIA, stated, this collaboration with cloud leaders provides 'a new kind of agentic engine that transforms compute into the high-quality data required to bring the next generation of autonomous systems and robots to life.' The focus is on creating scalable, repeatable processes for physical AI development, moving from isolated deployments to sophisticated, enterprise-grade workloads across industries.