NVIDIA Jetson Orin Nano 2 brings physical AI to drones and robots
| Source: AI News (ainews.com)
Tags: NVIDIA, Jetson Orin Nano 2, edge AI, robotics, physical AI, drones, computer vision
NVIDIA's Jetson Orin Nano 2 delivers 78 TOPS and 8GB RAM in a compact edge board, doubling inference performance over its predecessor while using 40% less power — enabling drones, robots, and vision systems to run LLMs like Gemma 4 and Qwen 3 locally at 15 watts.
Details
NVIDIA has launched the Jetson Orin Nano 2, targeting developers who want generative AI running on edge devices rather than in data centers. The board packs 78 TOPS of AI compute, 8GB of memory, and an eight-core Arm CPU — twice the inference throughput of the Jetson Orin Nano Super, with 40% lower power draw at the same performance level in its 15-watt operating mode. The case for the launch rests on a shift in small-model capability: NVIDIA argues that compact and mid-size models now match the accuracy previously requiring the largest frontier models. That lets edge hardware interpret language and images in real time without cloud round-trips — relevant for inspection drones, delivery robots, and industrial vision systems where latency and connectivity matter. Deepu Talla, VP of Robotics and Edge AI at NVIDIA, framed it as putting 'breakthrough' performance within reach of millions of developers. Cognex, Doosan Bobcat, and Matic are named as early adopters. The board runs NVIDIA's open software stack with support for Cosmos, Nemotron, Gemma 4, and Qwen 3 models optimized for memory-efficient inference. NVIDIA's robotics developer base exceeds 3 million, and the Jetson Orin Nano 2 slots in as an entry-level option in that ecosystem — a meaningful on-ramp for teams wanting physical AI without enterprise-grade hardware budgets.