NVIDIA Jetson Moves Agentic AI from Servers to the Physical Edge

NVIDIA Jetson AGX Orin 32GB just received a 20% performance boost, climbing to 241 TOPS of AI compute. This upgrade is part of a broader shift to bring agentic AI—previously confined to data centers and workstations—directly into the physical world of robotics, inspection, and industrial automation.

Software foundation for deterministic edge AI

JetPack 7.2 introduces critical architectural changes for production-grade edge deployments. For industrial users, Yocto-based OS support provides a leaner, more customizable Linux foundation, which is essential for memory-constrained environments. SandStar, for example, utilized these optimizations to achieve nearly 40% memory reduction, allowing them to migrate from 16GB to 8GB devices without sacrificing performance.

On the high-end spectrum, NVIDIA Jetson Thor now supports Multi-Instance GPU (MIG) alongside a real-time kernel. This allows developers to reserve dedicated GPU resources for deterministic workloads, such as robot perception systems, ensuring that critical sensor processing is never interrupted by unrelated AI inference tasks. Existing hardware also benefits, as CUDA 13 arrives on the Jetson Orin platform.

Deploying autonomous agents with NemoClaw

The integration of the NemoClaw agentic AI framework onto the Jetson stack enables developers to deploy autonomous agents with a single command. This moves beyond simple reactive programming into true agentic workflows where robots can reason, perceive, and act.

Humanoid robots like Solomon are already using NemoClaw to coordinate complex AI agents, integrating reasoning, perception, sensor fusion, locomotion, and manipulation into a single unified workflow. In manufacturing, Advantech is deploying an "agentic factory brain" using Jetson Thor and NemoClaw to automate robot fleet management and intelligent defect detection.

Beyond heavy industry, these agentic capabilities are scaling into smart cities and retail. Rebotnix is deploying smart city cameras with agentic reasoning for faster decision-making, while companies like Zipline use Yocto-based Jetson Orin NX modules to power autonomous delivery drones with real-time environmental awareness.

As NemoClaw moves from the data center into retail stores, factory floors, and humanoid robots, the era of physical AI agents is transitioning from experimental research to production-scale deployment.

Source: NVIDIA Jetson Brings Agentic AI to the Physical World
Domain: blogs.nvidia.com