The Information Machine

NVIDIA Jetson Brings Agentic AI to the Physical World

NVIDIA Blog · Chen Su · 2026-06-02

NVIDIA announced JetPack 7.2 and NemoClaw support on Jetson at COMPUTEX, delivering agentic AI capabilities to edge hardware for robotics, industrial automation, and smart city applications, with real-world deployments already live across humanoid robots, delivery drones, and AI vending machines.

Open original ↗

Appears in

Extraction

Topics: edge-ainvidia-jetsonagentic-airoboticsphysical-ai

Claims

  • JetPack 7.2 adds agentic AI skills, Yocto OS support, CUDA 13, and MIG support on Jetson Thor for deterministic workloads.
  • Jetson AGX Orin 32GB performance is boosted to 241 TOPS, a 20% increase above original spec.
  • NemoClaw now deploys to Jetson with a single command, bringing agentic AI from servers to production edge hardware.
  • SandStar achieved nearly 40% memory optimization migrating from 16GB to 8GB Jetson devices using NemoClaw without performance loss.
  • NoTraffic reduced memory usage 29% through CUDA static compilation and targeted kernel pruning for real-time traffic AI.

Key quotes

Agentic AI is here, and Jetson's programmability and high performance enable developers to instantly deploy physical AI agents in production at the edge.
NemoClaw started in the data center. Now it runs in a retail store, a humanoid robot on a factory floor, a traffic system at a busy intersection. The era of physical AI agents has just begun.