Nvidia Launches Medical Physics Simulator to Tackle Robotics Data Gap

Healthcare robots must learn how catheters, guidewires and surgical tools behave when they contact tissue, while accounting for friction, anatomical variation and noisy imaging. Yet clinical data are difficult to collect at scale, and rare failures cannot be reproduced on demand. Nvidia is treating these machines as embodied AI systems that need physical experience, using simulation to expand the data available for training, testing and regulatory evaluation.

Nvidia on July 22, 2026, unveiled Medical Physics Simulation, an open-source, GPU-accelerated framework within Isaac for Healthcare. Nvidia said a benchmark running 8,192 robot-training environments in parallel cut training time from more than five hours to under two minutes. CMR Surgical contributed nearly 500 hours of anonymized clinical data, while Johnson & Johnson MedTech, XCath and Medtronic Structural Heart are among the organizations applying or exploring the technology. Nvidia disclosed no investment amount.