ROBBRIEF
The new low power AI computer puts 78 TOPS (trillions of operations per second) of AI compute and 8GB of memory into a 15 watt chip, slotting into existing Orin designs for drones and robots.
What used to require a data-center rack now runs on a 15-watt chip: NVIDIA's Jetson Orin Nano 2 puts AI compute that previously needed datacenter-scale resources into a compact module built for drones, small robots, and vision systems that need on-device AI without a cloud connection.
The device is the runtime "robot brain" within NVIDIA's broader robotics stack — Omniverse and Cosmos for simulation, Jetson for inference at the edge. Deepu Talla, NVIDIA's vice president of robotics and edge AI, pointed to a year of model compression: NVIDIA's data-center class models a year ago topped out at roughly 600 billion to 1 trillion parameters, and he said the same workloads now fit a Jetson-class edge device. "One year ago, frontier models were 600 billion or 1 trillion parameter models," Talla told The Robot Report. "A year later, we're able to bring that level of accuracy all the way down to intro-level edge AI products in Orin Nano 2."
NVIDIA says the module delivers 78 TOPS of AI compute with 8GB of memory and an eight-core Arm CPU in the same compact footprint as its predecessor, doubling inference performance over the Jetson Orin Nano Super while drawing 40% less power at 15 watts. The 15-watt envelope puts Nano 2 in the same power range as a low-power laptop chip, aimed at drones, small robots, and vision systems that need on-device AI without a cloud connection. Nano 2 is a drop-in for existing Orin designs, so current customers can swap the chip without reworking their boards, per SiliconAngle.
The performance and power claims are vendor-reported. Independent benchmarks on the Jetson AI Lab archive and third-party testing of Nano 2 are not yet public, and NVIDIA has not disclosed pricing or a ship date.