Key Takeaways:
NVIDIA is giving AI agents the tools to build simulation-ready worlds, opening its Omniverse libraries for RTX sensor simulation, physics and asset validation on GitHub.
NVIDIA is embedding physical AI capabilities into the 3D tools developers already use, releasing three Omniverse libraries under its Agent Toolkit that let AI agents prepare simulation-ready environments for robots, factories and autonomous systems.
"The physical AI era will be built in simulation first," Jensen Huang, founder and CEO of NVIDIA, said. "NVIDIA Agent Toolkit with Omniverse libraries brings AI agents into the 3D tools developers already use, helping build the simulation-ready worlds where robots, factories and autonomous systems are trained and tested long before they reach the real world."
The libraries — ovrtx for RTX sensor simulation, ovphysx for GPU-accelerated physics, and CAD-to-SimReady skills for asset validation — are openly available on GitHub. A new blueprint for integrating them into Blender is also live. SideFX and PTC are among the first software makers to adopt the technology, embedding agent-ready sensor simulation and physics into Houdini and Onshape workflows respectively. The libraries give AI agents callable tools to inspect scenes, flag issues, test physics and prepare 3D assets for simulation — tasks that previously required manual work by technical artists and engineers.
Simulation Infrastructure as a Competitive Moat
Preparing 3D content for physical AI simulation requires more than realistic visuals. Assets need correct structure, materials, scale, labels, sensors and physical properties — a process NVIDIA calls "SimReady." The ovrtx library generates camera, lidar, radar and other sensor outputs from 3D scenes, letting developers test how autonomous systems perceive virtual environments. Ovphysx applies GPU-accelerated physics for collisions, mass, friction and motion, so teams can validate object and system interactions before physical deployment. The CAD-to-SimReady skills convert computer-aided design data into OpenUSD-based assets with the properties needed for virtual testing.
"Procedural 3D creation is essential to building the complex, controllable worlds needed for simulation, robotics and industrial AI," Kim Davidson, president and CEO of SideFX, said. "With NVIDIA Omnibus libraries and OpenUSD, SideFX is exploring how agent-ready tools can support Houdini workflows, helping technical artists review, test and prepare procedural content for simulation while staying in control of the creative process."
PTC's Onshape platform is using OpenUSD and ovrtx to connect cloud-native design workflows with physical simulation. "Engineering teams are seeking more connected ways to design, collaborate and simulate throughout the development process," Neil Barua, president and CEO of PTC, said.
Startups including Palatial, Lightwheel, ForgeCAD and Moonlake AI — all part of NVIDIA's Inception program — are also building on the libraries. Palatial is using CAD-to-SimReady skills to automate asset creation and validation at scale. Lightwheel's SimReadyGen technology generates physically accurate SimReady assets from text prompts using Omnibus Content Agents powered by OpenUSD.
Local AI Systems Bring Simulation to the Desktop
The workflows built with Omnibus libraries can run locally across NVIDIA's hardware stack, from compact RTX-powered systems to the deskside DGX Station. RTX Spark systems will be available this fall from ASUS, Dell Technologies, HP, Lenovo, Microsoft Surface and MSI, with models from Acer and GIGABYTE to follow. DGX Station systems — powered by the GB300 processor with up to 20 petaflops of FP4 AI compute and 748 gigabytes of coherent memory — are available to order from ASUS, Dell, GIGABYTE, HP, MSI, Supermicro and Exxact, with pricing starting around $85,000.
The local deployment option matters for enterprises working with sensitive design data that cannot be processed through external cloud models. NVIDIA's NemoClaw stack pairs the Nemotron 3 Ultra open model with Omnibus tools and the OpenShell secure runtime, enabling agent-driven workflows without an internet connection.
NVIDIA also introduced Synthetic Video Detector, a NIM microservice that analyzes footage frame by frame and flags synthetic content with up to 92 percent accuracy on uncompressed video in internal testing. Wowza is embedding the tool into livestreaming workflows across more than 35,000 deployments in 170 countries.
The broader SIGGRAPH slate included 21 accepted technical papers from NVIDIA Research, covering real-time character motion, 3D capture cleanup and new physics and material systems for applications ranging from games and films to robotics and factory digital twins.
For investors, the Omnibus libraries expansion deepens NVIDIA's software ecosystem at a time when its data center segment — which generated $62 billion in revenue over the past four quarters — faces growing competition from custom ASICs at Amazon, Google and Microsoft. By making AI agents productive inside the 3D applications that engineers and technical artists already use, NVIDIA creates switching costs that extend beyond GPU procurement. The company's shares trade at roughly 35 times forward earnings, reflecting the market's expectation that its software platform can sustain hardware margins as the physical AI market develops.
This article is for informational purposes only and does not constitute investment advice.
