Nvidia has upgraded its entry-level robotics computer, promising twice the inference performance of the previous generation for small machines and drones.
The Jetson Orin Nano 2 module delivers 78 TOPS of AI throughput with an eight-core Arm CPU, Tensor cores and 8GB of LPDDR5X memory, all inside the same compact 69.6mm by 45mm footprint as the existing Nano and Super boards. Running in its 15W mode, the module draws 40 percent less power than the Orin Nano Super while matching its performance.
Nvidia pairs the hardware with its Jetson agent skills software stack, which lets developers run large language models and vision language models efficiently on the edge, including Nvidia Cosmos models aimed at physical AI. The company positions the module for home robots, vision inspection systems, delivery drones and the carrier boards built around them.
Early adopters named by Nvidia include machine vision firm Cognex, construction equipment maker Doosan Bobcat and home robotics company Matic. Drone delivery company Wing says it will move its fleet from the older Orin Nano Super to take advantage of the new module’s real-time perception and reasoning capabilities.
The Jetson Orin Nano 2 module and its developer kit are scheduled to ship in the first half of 2027. The upgrade arrives as Nvidia pushes its edge lineup to capture a wave of robotics and autonomous systems demand building on its data center dominance.
