NVIDIA's Robotics computer 'Jetson Orin Nano 2' product image. /Courtesy of NVIDIA

Nvidia unveiled Jetson Orin Nano 2, an entry-level Robotics computer to equip drones and home robots with Generative AI. It doubles AI inference performance and improves power efficiency over the previous product, enabling large language models (LLM) and vision-language models (VLM) to run on small devices.

Nvidia said on the 25th that it plans to release the Jetson Orin Nano 2 module and developer kit in the first half of 2027. Jetson is Nvidia's compact computing platform that handles AI computation in robots, drones, and industrial cameras. Instead of sending data to a data center for processing, it is mainly used for Edge AI, in which the device itself recognizes images and runs AI models.

Jetson Orin Nano 2 offers 78 TOPS performance, performing 78 trillion AI operations per second, with 8 gigabytes (GB) of memory and an 8-core Arm central processing unit (CPU). With enhanced Tensor Cores and memory bandwidth, it doubles AI inference performance over the previous Jetson Orin Nano Super.

Power efficiency has also been improved. In the 15-watt (W) power mode, it delivers performance on par with the previous product while cutting power consumption by 40%. It is designed for use in drones and small robots, where battery life and heat affect product performance.

Nvidia expects the market to expand for running Generative AI directly on robots and drones without going through the cloud or data centers, as AI models have recently gotten smaller and more compute-efficient. On Jetson Orin Nano 2, AI models such as Nvidia Cosmos, Nemotron, Gemma 4, and Qwen 3 can be run.

Deepu Talla, Nvidia vice president for Robotics and Edge AI, said, "Today's small and mid-sized frontier models achieve accuracy comparable to last year's largest frontier models, enabling real-time intelligence on edge devices," and added, "Jetson Orin Nano 2 delivers the performance and energy efficiency required for real-time inference in smart drones, robots, and vision AI systems."

Cognex, Doosan Bobcat, home robot company Matic, and Wing, Alphabet's drone delivery subsidiary, are prioritizing adoption of Jetson Orin Nano 2 or considering applying it. Wing currently uses Jetson Orin Nano Super and Nvidia software in its delivery drones and plans to consider applying it to next-generation products.

According to Nvidia, more than 3 million developers are currently building products based on Nvidia Robotics software. When Nvidia released Jetson Thor, a high-performance computer for robots, in Aug. last year, it said there were more than 2 million developers. The price of Jetson Orin Nano 2 has not yet been disclosed.

※ This article has been translated by AI. Share your feedback here.