Nvidia just opened the gates to its entire physical AI infrastructure. The chipmaker unveiled a comprehensive suite of open-source models and frameworks at CES 2026 that spans the complete robotics development lifecycle - from high-fidelity simulation to edge deployment. The move gives developers direct access to the same tools powering everything from Caterpillar's voice-controlled heavy equipment to FDA-cleared surgical robots, potentially accelerating the path from lab prototypes to real-world autonomous systems.
Nvidia is betting that open-source infrastructure will define the next era of robotics development. The company's latest physical AI release represents a fundamental shift - instead of keeping its simulation and training tools behind closed doors, Nvidia is handing developers the complete toolkit it uses internally to build autonomous systems.
The timing matters. At CES 2026, the show floor became a proving ground for these tools, with companies demonstrating machines that moved beyond demos into actual deployment. Caterpillar brought its Cat AI Assistant directly into equipment cabs, letting operators adjust safety parameters by voice using Nvidia Nemotron models running on Jetson Thor edge modules. Behind those voice commands sits Omniverse-generated digital twins of entire job sites, where the company simulates traffic patterns and multi-machine workflows before deploying changes to real construction zones.
The medical robotics space is seeing similar transformation. LEM Surgical showcased its Dynamis Robotic Surgical System, which is already FDA-cleared and performing spinal procedures in clinical settings. The dual-arm humanoid surgical robot runs on Jetson AGX Thor for compute, uses Holoscan for real-time sensor processing, and trains its autonomous movements through Isaac Sim digital twin simulation. What makes this notable isn't just the hardware - it's that LEM Surgical generates its training data using , the open world model that creates physically accurate synthetic datasets without requiring thousands of hours of real surgical footage.
