Physical AI
NVIDIA Omniverse
Accelerated libraries and microservices for developing physical AI simulation applications and agentic simulation workflows.
Overview
What is NVIDIA Omniverse?
NVIDIA Omniverse™ is a collection of accelerated libraries and microservices for developing physical AI simulation applications and agentic workflows. Agents and software developers can use NVIDIA Omniverse™ capabilities as prebuilt tools to build, test, and refine their own solutions and agentic simulation workflows.
Technology
Omniverse Capabilities
NVIDIA Omniverse libraries, microservices, and APIs provide the foundation for data interoperability, scalable GPU-accelerated physics, and physically based rendering. These capabilities can be integrated into applications and services, or called as tools by physical AI simulation agents.
Physics and Motion
GPU-accelerated physics libraries built on NVIDIA PhysX® deliver scalable, USD-native physics for complex simulation, robotics, and industrial digital twin workflows.
Rendering and Simulation
Sensor simulation and physically-based, real-time rendering libraries built on NVIDIA RTX™ help generate synthetic data and simulate physical AI environments at scale.
Runtime and Deployment
Optimized data architecture and runtime services support the development, deployment, and scaling of physical AI simulation applications and agentic workflows.
Use Cases
How Omniverse Powers Physical AI Simulation Workflows
See how Omniverse libraries, microservices, and tools support simulation applications, services, and agentic workflows across industrial digital twins, synthetic data generation, robotics, and autonomous vehicle simulation.
-
Industrial Facility Digital Twins
-
Synthetic Data Generation
-
Robot
Simulation -
Autonomous Vehicle Simulation
-
Robot
Learning
Industrial Facility Digital Twins
Leverage Omniverse libraries to develop advanced virtual factory solutions and bring data interoperability, physically based visualization, generative AI, and real-time collaboration to your software.
Delta Electronics
Synthetic Data Generation
Developers can save significant training time and reduce costs by using synthetic data alongside real-world data to create carefully labeled datasets for training multimodal physical AI models. And now, with agent skills powered by NVIDIA Cosmos™, developers can turn coding agents into synthetic data experts for physical AI development.
Robot Simulation
Physical AI-powered robots and robot fleets must autonomously sense, plan, and execute complex tasks in the physical world. These include safely and efficiently transporting and manipulating objects in dynamic, unpredictable environments.
Autonomous Vehicle Simulation
With NVIDIA Cosmos, conditioned on Omniverse physics libraries, simulation developers can enhance their AV simulation workflows with high-fidelity, diverse sensor data and realistic behavior to train perception models and validate the AV software stack.
Robot Learning
Preprogrammed robots struggle with unexpected changes, while AI-driven robots use simulation-based learning to adapt to dynamic environments. This enables them to refine capabilities such as navigation and manipulation, improving performance in a wide range of scenarios.
Agility, Apptronik, Fourier Intelligence, Unitree
Starting Options
Ways to Build With NVIDIA Omniverse
Success Stories
Physical AI Simulation in Practice
NVIDIA Omniverse is used across the physical AI ecosystem to build simulation applications, generate synthetic data, and support digital twin and robotics workflows.
Ecosystem
Industry Leaders Adopt Omniverse
Resources
The Latest From Omniverse
Next Steps
Stay Up to Date on NVIDIA Omniverse News
Get the latest news, breakthroughs, and more sent straight to your inbox.
Frequently Asked Questions
NVIDIA Omniverse™ is a collection of libraries and microservices for developing physical AI applications such as industrial digital twins and robotics simulation. Leveraging NVIDIA’s deep expertise in accelerated computing and AI, Omniverse libraries enable software makers to integrate pre-built functionality into their solutions. These libraries include developer tools, GPU-accelerated libraries, and technologies packaged as microservices and cloud APIs for streamlined development and deployment.
There are three ways to start developing with NVIDIA Omniverse libraries:
- Use agent skills to make Omniverse capabilities callable as tools by physical AI simulation agents.
- Use Omniverse libraries and microservices to integrate prebuilt physical AI simulation capabilities into applications and services.
- Use NVIDIA Blueprints as reference workflows to jump-start physical AI applications, including industrial digital twins and robotics simulations.
Explore Omniverse libraries, agent skills, and NVIDIA Blueprints to get started here.
Physical AI simulation agents can call NVIDIA Omniverse™ libraries, microservices, and APIs to enable scene data interoperability, scalable GPU-accelerated physics, physically based rendering, sensor simulation, and runtime workflows. These callable capabilities help agents generate, test, and refine simulation workflows.
NVIDIA Omniverse Launcher (Deprecated)
The Omniverse Launcher was deprecated on October 1, 2025 to better align with developers and their expected development workflows.
Many of the Omniverse applications, tools, and assets that used to live in Launcher will transition to the following locations:
- Kit, Apps, Samples, Tools, and Templates are available on GitHub and the NGC Catalog.
- Connectors are available in the NGC Catalog.
- Extensions are available directly from the associated vendor's web pages or NVIDIA OpenUSD Ecosystem Catalog.
- Content and Assets are available on NVIDIA Documentation Hub (OmniDocs).
Visit Omniverse legacy tools and NVIDIA developer forums for more details.
Nucleus Workstation
Nucleus Workstation on Launcher was deprecated on October 1, 2025. Developers wishing to continue using Nucleus can obtain the Enterprise Nucleus Server software from the NGC Catalog (login required). The Enterprise Nucleus Server is free for testing and development, but an enterprise license is required for production use and includes enterprise support. See Omniverse legacy tools for more details.
NVIDIA Cosmos is a world model (WFM) development platform. At its core are Cosmos WFMs that generate world states as videos using multimodal input.
Developers can input Omniverse simulations as instructional videos to the Cosmos Transfer WFM model to generate controllable, photorealistic synthetic data.
Together, Omniverse provides the simulation environment before and after training, while Cosmos photoreal controllable synthetic data to train physical AI models.