Russia has developed a new loitering munition called the V2U that, according to Ukrainian reports, can use AI to navigate and strike targets and may have a swarm capability. If these reports are true, the V2U marks a significant step forward in Russian autonomous drone development.
Ukraine’s military intelligence directorate, or HUR, released information about the V2U on June 9. The HUR said it first spotted the drone in September 2024 during a “Tanker Day” event in Kazan, the capital of Russia’s Republic of Tatarstan. Which company developed the system is unclear.
The V2U’s first known combat employment occurred in February 2025 in Ukraine’s Sumy region. Since then, Russia reportedly has used the loitering munition with increasing frequency. In mid-May, Ukrainian expert Serhii “Flash” Beskrestnov, who works closely with the Ukrainian military and has inspected V2Us recovered in Ukraine, said Russia was using as many as 30 to 50 of these drones per day across several different areas of the front, training the drone’s AI in the process.
The Russians appear to have tweaked the V2U’s design over time, as shown in the photos below.
The loitering munition is catapult-launched and carries a KOFZBCh-3 warhead, a 2.9-kilogram shaped-charge high-explosive fragmentation incendiary munition. The warheads shown below appear to have been manufactured in 2024.
The drone is propeller-driven with a commercial off-the-shelf electric motor made in China, powered by Chinese-made batteries. Its precise range is unclear. Beskrestnov has offered estimates of 40-60 and 80 kilometers. There is reportedly also a gas-powered variant with a longer range of at least 100 kilometers. The Russians may have used this version in a strike in Odesa in May.
The V2U has inertial and satellite guidance systems and can also be manually controlled in first-person view (FPV) mode via LTE wireless broadband communication. The drone contains a router from the Russian company Microdrive (with Chinese components), coupled with a SIM card from a Ukrainian provider, the HUR explained.
More interestingly, the V2U is believed to have a terrain-matching capability for navigation in jammed environments. The HUR suggested that the drone can use computer vision to compare footage from its high-quality camera against imagery pre-loaded on a 128-gigabyte Chinese-made solid-state drive. In addition, the loitering munition carries a downward-facing LiDAR sensor, also from China, which could be part of a terrain contour matching system, as Beskrestnov reported.
Some of Ukraine’s long-range one-way attack drones reportedly possess similar capabilities. According to Beskrestnov, the Russians have also started training the V2U to maneuver during flight to avoid Ukrainian anti-aircraft FPV drones.
Moreover, according to the HUR, the V2U can “autonomously search for and select targets using artificial intelligence.” The drone’s brain is based on a Chinese-made Leetop A603 carrier board with a Jetson Orin AI module from the US company Nvidia—one of several Western components found in the drone despite Western sanctions and export controls.
Russia reportedly has employed the V2U in groups of two to around half a dozen drones. Beskrestnov’s accounts of V2U strikes in Ukraine suggest it has some swarm capability, meaning these loitering munitions can work together toward a shared objective, though the HUR did not address this question. Presumably, the drones could communicate with each other via their onboard LTE or Wi-Fi devices.
In one instance in May, Beskrestnov reported strikes by a “swarm of seven” V2U loitering munitions in the village of Velykyi Burluk in the Kharkiv region. He said the drones apparently noticed a cluster of vehicles near a Nova Poshta facility (Ukraine’s version of FedEx), as well as people at a nearby market, and decided to attack them. Before diving toward their targets, the drones formed into a circle, using distinct colors painted on their wings to help recognize each other and stay in formation, Beskrestnov said.
Assuming the targets were, in fact, civilian in nature, and the drones indeed identified them autonomously, this anecdote could suggest the V2U’s AI cannot yet reliably distinguish between certain military and civilian targets. It could also be programmed to treat civilian vehicles as legitimate targets, given that Ukrainian troops frequently use such vehicles (as do the Russians).
Whatever the case, the AI will likely improve with time as the developers gather more real-world data on which to train it.
