Jetson Orin Nano 2 consumes much less energy on the identical efficiency degree of its predecessor. Supply: NVIDIA
As AI fashions turn out to be extra environment friendly, extra gadgets can turn out to be autonomous, however builders want compact, energy-efficient computer systems constructed for edge AI, asserted NVIDIA Corp. The corporate immediately launched NVIDIA Jetson Orin Nano 2, a brand new entry-level pc for edge AI that it stated allows tens of millions of builders worldwide to construct programs for bodily AI purposes.
“At present’s small and medium frontier fashions have reached the accuracy of final yr’s largest frontier fashions, unlocking actual time intelligence for edge gadgets,” acknowledged Deepu Talla, vp of robotics and edge AI at NVIDIA. “The Jetson Orin Nano 2 pc places that breakthrough inside attain of tens of millions of builders, delivering the efficiency and power effectivity wanted for real-time reasoning in good drones, robots, and imaginative and prescient AI programs.”
“Over the previous three months, we’ve seen the evolution of open fashions powering bodily AI and the way essential they’ve turn out to be,” he added throughout a press briefing. “One yr in the past, frontier fashions which had been 600 billion or 1 trillion parameter fashions. A yr later, we’re in a position to convey that degree of accuracy all the best way all the way down to intro-level edge AI merchandise in Orin Nano 2. That is superb.”
NVIDIA Jetson Orin Nano 2 is a part of the corporate’s “three-computer, full-stack” strategy to robotics, with Omniverse with Cosmos offering simulation for testing, DJX supporting coaching, and Jetson because the “robotic mind” offering runtime deployment.
Register now and save in your cross to RoboBusiness 2026Jetson Orin Nano 2 advances accuracy, processing pace
“Jetson Orin Nano 2 delivers a big leap in AI and video processing efficiency in an economical, power-efficient system,” stated NVIDIA. The system options 78 trillion operations per second (TOPS) of AI compute, 8GB of reminiscence, and an eight-core Arm CPU.
Jetson Orin Nano 2 doubles the inference efficiency of Jetson Orin Nano Tremendous by improved Tensor Cores and better reminiscence bandwidth, whereas sustaining the identical compact kind issue. In 15-watt mode, Jetson Orin Nano 2 consumes 40% much less energy to ship the identical peak-to-peak efficiency as its predecessor, Talla advised The Robotic Report.
NVIDIA famous that its current open software program stack can run on Nano 2, which is constructed on the identical GPU structure as knowledge facilities. It’s designed to be a drop-in for current Orin prospects.
Together with Jetson agent expertise and the corporate‘s know-how ecosystem, builders can now run the newest giant language fashions (LLMs) and imaginative and prescient language fashions (VLMs) optimized for memory-efficient edge inference. These embody open fashions corresponding to NVIDIA Cosmos and Nemotron, Gemma 4, and Qwen 3.
“This now all of a sudden unlocks a degree of intelligence that was not possible — we’ve been dreaming about this for a decade in edge AI,” stated Talla throughout a press briefing. “It is a vital second in time, the place now for all of those robotics and bodily AI purposes, we are able to truly put frontier AI fashions corresponding to LLMs and VLMs on prime of the opposite autonomous capabilities. We consider it will pace up the event and deployment of AI within the bodily world.”
NVIDIA had already refreshed its higher-end Jetson choices, together with the Orin NX, Orin, and T3000/T2000 for mainstream purposes and the Jetson T5000/T4000 for superior makes use of.
Builders use NVIDIA stack for bodily AI purposes
NVIDIA stated that greater than 3 million builders are already constructing on its robotics stack. Companions corresponding to Cognex, Doosan Bobcat, Matic, and Wing are evaluating or utilizing Jetson Orin Nano 2 for real-world edge AI in compact gadgets.
“1000’s of firms ship with entry-level edge AI on Jetson, and greater than 10,000 firms are delivery or growing merchandise constructed on Jetson,” stated Talla.
Alphabet subsidiary Wing is utilizing Jetson Orin Nano Tremendous and the NVIDIA software program stack in its supply drone fleet. It plans plans to judge Jetson Orin Nano 2 to advance real-time AI notion and reasoning for sooner, safer deliveries from native companies to residential yards.
“Drone supply is determined by AI that may allow quick, dependable understanding of the actual world,” stated Dinuka Abeywardena, head of notion at Wing. “Wing is exploring Jetson Orin Nano 2 to offer us a path to extra responsive, energy-efficient drones that may assist make deliveries faster and extra reliable for purchasers.”
Matic Robots is utilizing Jetson Orin Nano 2 to allow extra intuitive interactions with its house cleansing robots. The corporate is giving the robots conversational AI, gesture detection, precision mapping and semantic understanding of the house setting, and better autonomy.
“Residence robots want to know individuals, map areas exactly, perceive the structure of objects and areas, and clear autonomously in dynamic and consistently altering environments,” stated Navneet Dalal, co-founder and CEO of Matic Robotics. “With Jetson Orin Nano 2, Matic can run state-of-the-art AI fashions on the edge in a compact house robotics platform constructed for real-time notion, interplay and navigation.”
“The issue with companion robots has been that the intelligence has not been good, however now, with the provision of all this frontier-level intelligence, which might run in actual time on Orin Nano or Orin Nano 2, we are able to allow a brand new class of purposes,” stated Talla. He additionally talked about Hugging Face‘s Reachy Mini, which makes use of a small LLM for speech recognition.
NVIDIA cited AAEON, ADLINK, Advantech, Aetina, Antmicro, Aptiv, Auvidea, AVerMedia, Chuanglebo, Join Tech, ForeCR, JWIPC, Neurealm, Plink, Realtimes, RidgeRun, RS, Seeed Studio, Tauro Tech, Twowin, TZTek and YUAN as different Jetson companions constructing provider boards, {hardware}, custom-made AI software program, and reference programs to assist prospects speed up time to market.
Talla stated he expects NVIDIA’s know-how to allow generalized navigation and dexterous manipulation, in addition to enhancements in robotic coaching, security, and independence. NVIDIA stated the Jetson Orin Nano 2 module and developer package can be accessible within the first half of 2027.
The publish Jetson Orin Nano 2 doubles inference efficiency for robotics on the sting, says NVIDIA appeared first on The Robotic Report.
