NVIDIA

英伟达Essential

The GPU giant, supplying a full stack of compute and software for training, simulating, and deploying robots.

NVIDIA was founded in 1993 by Jensen Huang and colleagues, headquartered in Santa Clara, California, and became the leading supplier of AI training compute through its GPUs. In embodied AI it promotes a 'three computers' framing: DGX clusters for training models, Omniverse and Isaac Sim/Isaac Lab for simulation and synthetic data, and Jetson Thor for running models onboard the robot. It also builds its own models: the GR00T N1 series of humanoid foundation models (starting March 2025) and the Cosmos world foundation model (with the third generation, Cosmos 3, announced at GTC in March 2026); at the same event it also previewed GR00T N2, shifting toward a 'world action model' approach, planned to open up before year's end. Research is led by teams including the GEAR lab. On September 3, 2026, NVIDIA announced it would acquire Hugging Face, pending regulatory approval.

ExampleMany quadruped and humanoid reinforcement-learning locomotion policies are trained in Isaac Lab across thousands of parallel simulated environments, then deployed onto Jetson hardware.

Also called: Nvidia
Related: NVIDIA Three-Computer Solution · NVIDIA Isaac Lab · NVIDIA Isaac GR00T N1 · NVIDIA Cosmos · NVIDIA Jetson Thor · NVIDIA Generalist Embodied Agent Research Lab
Sources: NVIDIA and Global Robotics Leaders Take Physical AI to the Real World (GTC 2026)
Jetson Thor
Hugging Face - Wikipedia（英伟达收购报道） (Chinese)
As of: 2026-09