As I understand it, currently all drones require a human operator who can only operate one at a time. And except for some special operations behind enemy lines, you must be fairly close to the target, as within a few km. The fiber optic ones, even closer
So your 50 drone swarm is going to need 50 operators, fairly close to the front. Who are also vulnerable to enemy counter drones and glide bombs - the latter is a real problem for Ukraine
I haven’t seen any evidence of a “swarm” on combat footage from Ukraine war, I have seen a few drones hitting a single target, especially armored vehicles in fairly quick succession, like a few seconds, It looked like independent operators all picking the obvious high value target, not some intentional “swarm”
Tech may change this in the future but we’re not there quite yet
First, you don't need AI operators, you just need a swarm. The operators are reusable!
>Ukraine reported the largest single-day drone attack by Russia on July 9, 2025, where Russia targeted Ukraine with a record 728 drones. This surpasses earlier attacks, including one on May 26, 2025, when Russia launched 355 drones.
With that many pilots, that is a swarm.
Next, analysis of last months AI driven attack was performed by many drones with no human terminal guidance - they were jammed and expected to be!
>“Our models are being trained to recognise targets to understand target prioritisation,” he says. “We do not have full autonomy yet. We use the human factor where we need to, but we are developing different scenarios for taking autonomy further.
> “We are also testing some autonomous drones, which we have not announced and are probably not planning to announce, but they have a high degree of autonomy, and they can potentially combine themselves into swarms. We are still facing technical problems and hurdles, but we already see a path forward on this.”
One Final Note - Most of the info you ever hear about military tech is only the things people are allowed to discuss publicly. The battlefield is also a hell of a lab, and 3d printers and open source flight software (and open source AI models) are amazing.
Those 725 drones were spread across a fairly big geographic area, and didn’t hit all at once. Also they operate more like cruise missiles, not the FPV drones it seemed like the article was referring too
“Swarm” to me means more than just number. It’s number, concentration, and tactics, like a swarm of bees… the problem is they are concentrated and hitting from many directions, While individually they are not that bad, when they use this tactic it is very effective, Which is how they can drive 500 pound bears away from their hive.
Otherwise “swarms” have been a thing for along time. Would you call an 19th century infantry regiment (let’s say about 600-1000 soldiers) a “swarm”. Or how about those formations of B17/B24s/Lancasteres in WW2 which would attack in similar numbers (hundreds). I would say no, partly because they didn’t use a swarming tactic
Argue about the definition of swarm (the distance between units and level of coordination) all you want, but ultimately it's irrelevant given the addition information.
Massive coordination is going into attacks across hundreds or thousands of Km. Multiple layers of drones, electronic warfare, recon, airspace deconfliction, etc. Highly orchestrated. Large numbers that are overwhelming systems designed to defeat them, like a swarm of locust.
Note that the one drone is not the other; when they talk about Russian mass drone attacks, they will refer to Shahed etc drones, which are autonomous, not unlike the WW2 V1 "drones".
But yeah, drone swarms with fewer operators will be, probably already is a thing. But what I've seen so far, they're just not very useful; drones look to be generally used on individual targets, if there's a bigger or more targets, they'll use something bigger like a HIMARS, glide bomb, or if it's closer by, an artillery strike.
Drone swarms primary purpose is to overwhelm defenses.
Many argue drone swarms require some level of orchestration and control, others say a certain level of automation is required.
I'm aware of the differences in many drone classifications.
HIMARS was made largely impotent by GPS jamming. Glide bombs have limited range (barring exceptions for stuff like JASSM-ER but that is massive increase in cost) and detection and fire by counter battery. Artillery strike requires fairly close proximity but a bit more of rocket assisted.
Spent time doing military things with a lot of ordinance and a lot of drones.
Ukraine is having pretty substantial manpower problems in its armed forces. If fully autonomous drones against mobile targets had been figured out, they would been deployed and there would be no need for the more expensive / shorter range fiber optic drones and you wouldn't hear about the manpower issue as much
If we were in a war with widespread conscription that needed self-driving vehicles, I'd bet we'd a) see a lot more investment in them and b) a lot less concern about edge cases.
So your 50 drone swarm is going to need 50 operators, fairly close to the front. Who are also vulnerable to enemy counter drones and glide bombs - the latter is a real problem for Ukraine
I haven’t seen any evidence of a “swarm” on combat footage from Ukraine war, I have seen a few drones hitting a single target, especially armored vehicles in fairly quick succession, like a few seconds, It looked like independent operators all picking the obvious high value target, not some intentional “swarm”
Tech may change this in the future but we’re not there quite yet