Nvidia's Jetson Orin Nano 2 doubles edge inference with 40% lower power draw
Nvidia just refreshed the little computer that powers a huge slice of the world's robots — and the efficiency numbers are the real story.
Nvidia announced the Jetson Orin Nano 2, a new entry-level robotics computer that doubles the inference performance of its predecessor while drawing 40% less power in 15-watt mode. The module packs 78 trillion operations per second of AI compute, 8GB of memory and an eight-core Arm CPU into the same compact form factor as the previous generation, with gains coming from improved Tensor Cores and higher memory bandwidth. It will be available as a module and developer kit in the first half of 2027.
The pitch from Nvidia's Deepu Talla, vice president of robotics and edge AI, is that small and medium frontier models have now reached the accuracy of last year's largest frontier models — which means real-time reasoning no longer needs a data center on the other end of a wireless link. That is the shift this hardware is built to catch: when a delivery drone or a home robot can run a vision-language model onboard, it can perceive, decide and react in milliseconds instead of waiting for a round trip to the cloud. Wing, Alphabet's drone-delivery subsidiary, is already exploring the chip for exactly that reason, and Matic wants it for cleaning robots that hold a live semantic map of your home.
Our take: the edge-AI market has quietly become Nvidia's defensive moat while everyone watches the data-center arms race. More than 3 million developers already build on the Jetson stack, and partners including Cognex, Doosan Bobcat, Matic and Wing are signed up for this generation. Every robot that starts on cheap, familiar Jetson hardware is a sale Nvidia keeps when that robot ships at scale — and a market where rivals like Qualcomm, Hailo and a wave of Chinese NPUs have so far struggled to break the ecosystem lock-in.
What to watch: whether first-half 2027 availability holds amid the global memory shortage, and if the doubled inference headroom is enough to run frontier-class models fully offline by launch.
Would you trust a delivery drone making real-time decisions entirely onboard — or do you want a human in the loop? Tell us in the comments.
