NVIDIA is pushing agentic AI beyond servers and workstations with the latest updates to its Jetson embedded platform. The release of JetPack 7.2 and the integration of NVIDIA’s NemoClaw framework bring significant advancements for developers building AI-powered applications directly on edge devices.
What Happened
At COMPUTEX, NVIDIA announced the availability of JetPack 7.2 and NemoClaw support on Jetson. JetPack 7.2 introduces agentic AI skills, Yocto project support, and NVIDIA CUDA 13 on Jetson Orin. It also delivers a performance boost to the Jetson AGX Orin 32GB module—increasing AI compute to 241 TOPS—and Multi-Instance GPU (MIG) support on Jetson Thor. These updates were showcased at the Build-a-Claw event, a hands-on workshop bringing the GTC San Jose experience to Taipei.
Image 1: NVIDIA's Asier Arrnaz shows how Build-a-Claw brings AI to the edge, a personalized, always-on assistant running right on NVIDIA Jetson.: image omitted due to site embedding policy; open the original article (NVIDIA Blog) (opens in a new tab) to view it. Photo/source: NVIDIA Blog (opens in a new tab)
The release includes a layered approach: JetPack 7.2 forms the base with OS and compute functionalities, a new layer of agent skills automates developer tasks, and NemoClaw sits at the top, providing the agentic AI framework. These agent skills can automate tasks such as Linux customization, memory optimization, and model benchmarking, significantly reducing development time.
Why It Matters
These updates are significant for developers working on edge AI applications in robotics, autonomous systems, industrial inspection, and medical devices. The combination of improved performance, a more customizable OS (through Yocto support), and automated developer workflows with agent skills promises to accelerate time to market and reduce the total cost of ownership. The addition of MIG support on Jetson Thor is particularly relevant for real-time applications like robotics, where deterministic performance is crucial. NemoClaw’s deployment on Jetson extends the reach of agentic AI, allowing for more sophisticated and autonomous behavior in physical systems.
This release highlights NVIDIA’s commitment to simplifying and accelerating the development process for edge AI. The agent skills layer, in particular, addresses a key challenge for embedded systems developers—the complexity of building and customizing a Jetson-based system. By automating common tasks, NVIDIA aims to empower developers to focus on the core AI functionality of their applications.
What To Watch
It will be important to see how developers adopt the new agent skills and integrate NemoClaw into their projects. The 20% performance increase on Jetson AGX Orin 32GB is a welcome addition, but real-world performance gains will depend on the specific application. Further, the degree to which Yocto support simplifies deployments for industrial customers remains to be seen. Looking ahead, continued improvements in agentic AI capabilities and further optimization of the Jetson platform will be key to unlocking the full potential of edge AI.