Key Points:
- Nvidia introduced the Jetson Orin Nano 2, an entry-level edge AI computer delivering 78 TOPS of processing power for smart robotics and drones.
- The new module doubles AI inference performance while cutting power consumption by 40% in its 15-watt operating mode.
- The system packs 8GB of unified memory and an eight-core Arm CPU, enabling real-time on-device execution of large language and vision models.
- Early commercial adopters include Alphabet’s drone delivery subsidiary Wing, industrial equipment maker Doosan Bobcat, and Cognex.
Semiconductor leader Nvidia is expanding its robotics computing lineup to bring advanced artificial intelligence directly onto physical machines. The company officially introduced the Jetson Orin Nano 2, a compact edge computing module designed to serve as the onboard brain for smart robots, automated inspection devices, and autonomous delivery drones. The next-generation hardware doubles artificial intelligence inference performance over its predecessor while maintaining an identical, compact physical footprint.
The Jetson Orin Nano 2 delivers a major leap in computing density for entry-level embedded devices. Packing 78 trillion operations per second (TOPS) of artificial intelligence compute, the module integrates an upgraded graphics architecture with enhanced Tensor Cores, an eight-core Arm central processing unit, and 8 gigabytes of high-bandwidth unified memory. This hardware combination provides the computational muscle required to process high-resolution video streams, sensor data, and neural networks simultaneously in real time.
A primary engineering breakthrough in the new module is its dramatic improvement in power efficiency. When operating at equivalent computational loads to the prior-generation Jetson Orin Nano Super, the new system consumes 40% less electrical power, lowering total system draw from approximately 25 watts to 15 watts. In space- and battery-constrained machines like lightweight aerial drones and battery-operated mobile robots, reduced power consumption translates directly into extended battery runtime and lower thermal heat dissipation.
The hardware upgrade arrives as machine learning models become smaller, faster, and more efficient. While running multimodal artificial intelligence historically required massive cloud data centers, compressed frontier models now run directly at the edge. The Jetson Orin Nano 2 natively supports memory-efficient open models, including Nvidia’s proprietary Cosmos and Nemotron architectures alongside Google’s Gemma 4 and Alibaba’s Qwen 3. Executing vision-language models locally allows robots to interpret visual scenes and voice commands without transmitting private data over wireless networks or suffering from network latency.
The real-world utility of the new processor was demonstrated through interactive robotics platforms. In technical demonstrations, a single computing board powered two small interactive robots, enabling them to understand spoken human dialogue, map physical surroundings, and respond verbally completely offline. Commercial partners are moving quickly to test the module in industrial fleets: Alphabet’s drone delivery subsidiary Wing is evaluating the system for its autonomous aerial delivery drones, while industrial machinery maker Doosan Bobcat and smart home robotics developer Matic are integrating the processor into autonomous cleaning and construction tools.
The new processor forms a critical link in Nvidia’s comprehensive three-computer robotics strategy. Under this development framework, robotics companies use hyperscale data centers to train complex foundation models, test and validate robotic movements inside virtual digital-twin simulations using Omniverse and Cosmos, and deploy the perfected software onto physical machines powered by Jetson processors. This end-to-end software pipeline eliminates the risks and costs of testing experimental code on expensive physical hardware.
The launch leverages an extensive developer ecosystem that spans more than 3 million active robotics engineers and programmers worldwide. Hardware ecosystem partners—including Connect Tech, Seeed Studio, AAEON, ADLINK, and Advantech—are actively developing custom carrier boards, rugged enclosures, and industrial sensor kits based on the new module. This broad hardware support ensures that system integrators can deploy the technology across factory automation, agricultural robotics, and medical equipment seamlessly.
Nvidia announced that production modules and comprehensive developer kits for the Jetson Orin Nano 2 are scheduled for commercial release in the first half of 2027. By lowering the financial and power barriers to deploying generative artificial intelligence at the edge, the company is democratizing physical AI for developers, researchers, and industrial manufacturers.
As the global robotics industry shifts from pre-programmed, rigid automation toward flexible, physically intelligent machines, on-device computing power has become the ultimate differentiator. By packaging 78 TOPS of artificial intelligence compute into a low-power, 15-watt board, Nvidia is transforming how robots perceive, reason, and interact with the physical world. The Jetson Orin Nano 2 establishes a versatile computing foundation that will accelerate the deployment of autonomous machines across commercial industries for years to come.
