NVIDIA Jetson

NVIDIA's family of embedded AI computing platforms built for robots and edge devices.

Jetson is NVIDIA's line of embedded computing platforms, packaging an Arm CPU, an NVIDIA GPU, and memory into a small, low-power core module, paired with a carrier board (which breaks out various I/O) or a developer kit. It lets a robot run neural networks on board without a connection to the cloud, making it one of the most common choices for a robot's onboard compute. The lineup began with the TK1 in 2014, and has since gone through TX1/TX2, Xavier, Nano, and Orin, up to 2025's Thor, built on the Blackwell architecture. All of them run the same JetPack software stack (bundling Ubuntu, CUDA, TensorRT, and more), so a model trained on a server is relatively easy to deploy onto one. The main specs to weigh when choosing a module are compute performance, memory, and power draw.

ExampleMany humanoid and quadruped robots mount a Jetson Orin inside the body as the onboard computer running perception and policy models.

Also called

Jetson, Jetson Series

Related

NVIDIA Jetson Orin · NVIDIA Jetson Thor · Onboard Compute Platform · On-Device / Edge Deployment · NVIDIA JetPack SDK · NVIDIA TensorRT

Sources

NVIDIA Embedded Systems for Autonomous Machines
Nvidia Jetson - Wikipedia

As of

2026-09