Cheap Model Of Computer Is Powering A.I. Drones

The abilities and price of the Nvidia Jetson Orin:

 

The NVIDIA Jetson Orin family ranges from entry-level edge AI boards starting at an MSRP of $249 up to high-end developer kits costing around $1,999 to $3,499.

 

 

NVIDIA Jetson Orin Nano Super

•           Price: $249 MSRP

•           AI Performance: Up to 67 Sparse TOPS (Trillion Operations Per Second)

 

•           GPU & CPU: NVIDIA Ampere architecture GPU with 1,024 CUDA cores and 32 tensor cores, paired with a 6-core ARM Cortex-A78AE CPU.

•           Memory & Power: 8 GB LPDDR5 RAM (102 GB/s bandwidth) running at 7W to 25W power consumption.

•           Use Cases: Running local small language models (LLMs), vision transformers, computer vision, and entry-level robotics.

 

Up to 67 Sparse TOPS (Trillion Operations Per Second) is more than enough for a flying drone to run real-time computer vision models to find a target.

What 67 Sparse TOPS Means

•           Sparse TOPS measures AI performance using structural sparsity (pruning zero-value weights) to accelerate neural network execution.

•           Real-world equivalent: Efficient edge processors like YOLO-optimized architectures or chips rated around 60–80 sparse TOPS can easily run lightweight object detection models (such as YOLOv8n or YOLOv10) at 30 to 60 frames per second.

Why It Is Sufficient for Target Detection

•           Low-power efficiency: Standard edge-AI target detection requires only 10 to 40 dense or sparse TOPS to process standard 1080p or 4K video feeds for bounding-box recognition.

•           Real-time processing: 67 sparse TOPS provides plenty of headroom to handle frame-by-frame inference, object tracking, and basic onboard flight path adjustments without needing a heavy ground station or cloud connection.

 

 

 

NVIDIA Jetson AGX Orin Developer Kit

•           Price: Around $1,999 to $3,499 MSRP (varies by reseller and memory configuration)

•           AI Performance: Up to 275 TOPS

•           GPU & CPU: NVIDIA Ampere architecture GPU with 1,792 CUDA cores and 56 tensor cores, paired with a 12-core or 8-core ARM Cortex-A78AE CPU.

•           Memory & Power: 32GB or 64GB LPDDR5 RAM with power scaling between 15W and 60W.

•           Use Cases: Advanced autonomous machines, complex multi-sensor fusion, heavy natural language understanding, and edge server workloads.

You can view official specifications and software support details on the NVIDIA Jetson Orin Platform.

 

“The tech giant says it does not sell the devices in Russia, but they are widely available on resale markets. When purchased that way, they are virtually impossible to track.

 

The miniature Nvidia computer that Ukrainian officials say Russia has used in fully autonomous A.I.-guided drones is an inexpensive model designed for civilian robotics and machine vision applications, according to the company.

 

Ukrainian forensic examiners recently found the minicomputer, the Nvidia Jetson Orin, in the wreckage of drones launched by Russia's military in apparent tests of a self-targeting system in the city of Zaporizhzhia in southeastern Ukraine.

 

Ukraine's domestic intelligence agency showed New York Times reporters two such guidance systems that included damaged but still recognizable Jetson computer components. Air defense officials said that a drone with one of the systems killed three people in an attack in Zaporizhzhia on July 6.

 

The Russian weapon is a new step in the use of A.I. in combat. Earlier A.I.-guided drones had a ''human in the loop'' safety button that required a live pilot to confirm a target. Those systems used artificial intelligence only in the last several hundred yards of a drone's flight to help make them impervious to radio signal jamming.

 

The Russian drones tested in Zaporizhzhia from May to July, by contrast, left the strike decision entirely to the Jetson Orin computer, according to an analysis of the chip by Ukrainian investigators. The Red Cross and other rights groups have opposed allowing computers to find and strike targets on their own.

 

Ukrainian air defense officials and forensic specialists who examined debris from the drones said that the Jetson computers had been trained to identify a range of potential targets. Those included used by military gas stations, which have been a focus of Russian attacks in recent months. The drone that killed three conflict participants on July 6 was flying toward a used by military gas station.

 

Ukraine's defense intelligence agency said this month that an Nvidia Jetson computer had also been found on a new type of Russian missile, called the S-71M Monochrome, that was being tested in attacks on Ukraine. The Nvidia system may have been used for A.I. guidance, the agency said.

 

In a statement, Nvidia said that the Jetson Orin computers found in the Russian drones were ''consumer-grade products sold to students, developers and start-ups for a wide range of beneficial applications.'' They were not designed for military applications, the company added.

 

The U.S. government restricts exports of items to Russia for military use. Nvidia said that the Jetson line of computers was not sold in Russia but was available in resale markets. The company has said that it cannot track such secondary sales.

 

The circuit board on which the Nvidia chip was installed was stamped ''Made in China.'' It was unclear how Russia's military obtained the computers.

 

Nvidia, the world's most valuable company by market capitalization, calls the line of Jetson minicomputers ''the world's most powerful embedded A.I. computers.''

 

The devices run applications that are trained on large volumes of data rather than explicitly programmed by a human.

 

As an example of possible uses, Nvidia cited a Japanese manufacturer, Musashi, that trained the Jetson computer to identify defects in welding and other areas on vehicle gearboxes.

 

The computers replaced human inspectors, whose work Nvidia called ''repetitive, stressful and not very engaging.''

 

This article appeared in print on page A8.” [1]

 

1. Cheap Model Of Computer Is Powering A.I. Drones: [Foreign Desk]. Kramer, Andrew E.  New York Times, Late Edition (East Coast); New York, N.Y.. 25 Aug 2026: A.8.