NVIDIA unveiled JetPack 7.2 and NemoClaw support for its Jetson platform at COMPUTEX on Tuesday, pushing agentic AI into physical applications. The update brings agentic AI skills, Yocto project support, and NVIDIA CUDA 13 to the Jetson Orin lineup. A substantial performance gain on the Jetson AGX Orin 32GB module and Multi-Instance GPU (MIG) support were also announced.
Agentic AI refers to systems that can autonomously perceive, reason, and act in the real world. By bringing it to edge devices like Jetson, NVIDIA is targeting robotics, industrial automation, and smart infrastructure. The move extends the company's AI stack beyond cloud data centers into environments where low latency and local processing are critical.
JetPack 7.2 is the software development kit underpinning these capabilities. It now supports the Yocto Project, a popular framework for building custom Linux distributions for embedded systems. CUDA 13 integration ensures developers can leverage GPU acceleration for AI inference at the edge. MIG support allows partitioning of a single GPU into multiple isolated instances, improving resource utilization.
These enhancements could accelerate deployment of autonomous machines in factories, warehouses, and cities. Companies building drones, delivery robots, and surveillance systems may benefit from reduced reliance on cloud connectivity. However, the complexity of deploying and maintaining agentic AI at scale remains a hurdle for many enterprises.
Some analysts question whether edge AI hardware can match the performance of cloud-based solutions for complex reasoning tasks. NVIDIA's claims of substantial performance gains lack specific benchmarks, leaving room for skepticism until independent testing validates them.