NVIDIA Jetson Brings Agentic AI to Physical World

At COMPUTEX, NVIDIA announced significant updates to its Jetson platform, bringing agentic AI capabilities directly to edge devices. The release of JetPack 7.2 introduces support for agentic AI skills, Yocto project compatibility, and CUDA 13 on Jetson Orin modules, delivering substantial performance improvements for physical AI applications. This update enables AI agents to operate and make decisions in real-world environments without relying on cloud connectivity. By running directly on edge hardware, these agents can process sensor data, reason about their surroundings, and execute actions with low latency. This is particularly valuable for robotics, autonomous machines, and IoT devices that require real-time responsiveness. NemoClaw support further enhances the platform's capabilities, allowing developers to build more sophisticated agent behaviors. The Yocto project integration gives engineers greater flexibility in customizing their Linux-based embedded systems, while CUDA 13 unlocks new levels of GPU acceleration for AI inference and training on the edge. These advancements position the Jetson platform as a key enabler for the next generation of physical AI. From warehouse robots that adapt to changing layouts to agricultural drones that make real-time crop decisions, the combination of agentic AI and edge computing promises to transform industries by putting intelligent decision-making where it matters most—in the physical world.