Does Russia’s Smartest Weapon Have An American Brain?

Russian military hardware has seriously underperformed in this conflict. Their tanks explode catastrophically when hit, their missiles have a failure rate of 50% or more , their air defense is a running joke, their flagship cruiser was sunk by an opponent with no navy, and their Air Force has been missing in action throughout the war. One Russian weapon though has performed dangerously well, the Lancet-3M loitering munition which has been picking off Ukrainian artillery from long range. It now seems this success may be due to components imported from the U.S.

Russian forces often post kill videos from Lancet kamikaze attacks on social media. These show a vehicle seen from a reconnaissance drone, then the same vehicle from the incoming Lancet’s camera, then back to the first drone to see the white cruciform of the Lancet slam into the target and explode. While some targets may not be destroyed, the accuracy of the system is unquestionable.

Made by a subsidiary of the Kalashnikov concern, the latest version is the Lancet-3M or “Product 52” which has an eight-foot wingspan, weighs about 35 pounds and cruises at about 70 mph. The operator seeks and identifies targets the drone’s camera. The drone dives at 190 mph to impact with an armor-piercing warhead. The makers claim a range of 25 miles.

The makers also claim Lancet has an autonomous mode in which it flies to a given location and then visually searches the area for targets using an object-recognition algorithm. In this mode is does not need any link to the operator and is immune to jamming.

Independent analysts Oryx have been keeping a tally of videos showing strikes by Lancets. They are often used to take out artillery – Oryx notes 43 strikes on artillery pieces, in particular U.S.-supplied 155mm howitzers, as well as other high-value targets behind the front lines including 7 surface-to-air missile systems, 11 radars and 2 command posts. This is in addition to 11 tanks and 9 other armored vehicles. (Ukraine’s nearest equivalent is probably the shadowy, U.S.-supplied Phoenix Ghost about which little has been revealed though hundreds have been deployed).

Russia’s dependence on Western electronics is well established. While Russian officials like to boast their equipment is purely Russian, a joint report by last August RUSI and Reuters which examined 27 captured Russian systems in Ukraine including cruise missiles and drones identified at least 450 unique microelectronic components from companies in the US, Europe and East Asia.

Last week Roman Kashchuk, a Ukrainian sniper training instructor, posted images on social media of a crashed Lancet-3 with caption “Something this evening for pathologists. This is a captured Lancet.” (my rough translation).

The images include what looks like a U.S.-made NVIDIA Jetson TX2 single-board computer and a Xilinx
XLNX
Zynq chip (Xilinx Adaptive is owned by AMD). These findings cannot be verified but seem credible.

“It’s not surprising that this particular Russian drone has imported components, given that NVIDIA
NVDA

DIA
is a global leader and its technology is considered the best in the world,” Samuel Bendett an expert Russian drones and adviser to both the CNA and CNAS, told Forbes. “Russia’s different acquisition and supply chain pipelines were also well documented last year and this year, while NVIDIA has had a long-term presence in the Russian high-tech ecosystem for years.”

The Jetson TX2 is a credit-card sized unit which the makers describe as “the fastest, most power-efficient embedded AI computing device. This 7.5-watt supercomputer on a module brings true AI computing at the edge.” It is one of a new generation of processors designed to bring advanced machine learning and AI to mobile devices like drones.

For example, most drone require the operator to fly them and point the camera at an object of interest. But newer consumer drones like the Skydio 2, powered by an NVIDIA Jetson 2, fly themselves – you just launch the drone and tell it to follow you as you ski, skateboard or bike, automatically avoiding obstacles and keeping you in the frame – video here.

The exact role of the AI-chip aboard the Lancet is unclear, according to Zak Kallenborn of CSIS’s Strategic Technology Program. He suggests that it might be able to highlight objects of interest, in other words possible targets, for the operator. Alternatively, it may be enhancing the guidance system.

“For example, the Lancet might use the AI algorithms to help track the target, and ensure the Lancet is on the correct course to blow it up,” Kallenborn told Forbes.

A third possibility is that the TX2 is the brains behind the autonomous mode and provides a genuine ‘killer robot’ capability to independently seek and destroy targets, though it may be somewhat underpowered.

“In the killer robot case, it’s not clear this chip has the umph to do that effectively, but it seems plausible,” says Kallenborn.

Sanctions should prevent Russia from getting this type of high-tech gear. NGO Conflict Armament Research has been tracing the supply routes that lead to Western components in Russian weapons with a view to ensure that sanctions are enforced and that agents in third countries cannot simply bypass them, but tightening the screws will take some time.

“Its likely that the Russian government and its institutions have multiple avenues for procuring this technology via allies, third parties and other schemes,” says Bendett.

In theory it would be possible for Russia to build their own processors. Latvia-based Atlas Dynamics build all their components themselves. Reinventing the wheel is a costly, time-consuming process; their FGPA-based autonomy system took five years to develop, but it means they are not reliant on NVIDIA, Chinse manufacturers or any other third party. It also means that no other country, such as the U.S. or China, can impose export restrictions or delay their supplying drones to Ukraine.

The Kalashnikov Concern is no longer state-run, but it is now 75% owned by former Minister Alan Lushnikov who is also the company’s President. It likely suffers the endemic corruption, bureaucracy and inefficiency of the Russian defense industry; that the company has its own anti-corruption hotline says something. Given that they chose to rely heavily on components from potentially hostile states, the company’s ability to innovate its way out of shortages has to be in doubt.

Russia’s use of the Jetson TX2 on a front-line weapon should be an indication of just how powerful commercial AI hardware has become. Meanwhile NVIDIA continues to bring out even more powerful modules, like the Jetson Orin capable of 275 trillion operations per second compared to 1.3 trillion for the TX2. Such modules will be freely available to Ukrainian developers and those who supply them as they build their drone army and introduce ever more capable drones.

As sanctions take hold, Russia will have fewer and fewer smart drones, while Ukraine is likely to deploy them by the thousand.