Russia’s drone campaign against Ukraine has stopped being a curiosity and become an industrialized part of modern warfare. What began as a stream of imported Iranian Shahed loitering munitions has evolved into a layered, high‑volume campaign that mixes mass production, incremental engineering changes, paired tactics, and simple doctrinal insight: if you can build and launch enough cheap strike drones, you can grind down an opponent’s sensors, interceptors, and patience. The change looks like learning. It is learning, but it is not the sort of elegant, autonomous leap often hyped in the press.
The mechanics of that learning are straightforward and low tech. Russia localized Shahed production and rebranded the kits as Geran; it has ramped output dramatically, sometimes measured in the thousands per month, then turned those parts into nightly saturation packages. That industrial pivot is the real multiplier. Numbers matter because modern air defenses have finite interceptors, sensors, and trained crews. Flood the sky and the math shifts in favor of the attacker. The data from 2024–2025 shows orders of magnitude increases in launches and in average drones per strike package.
But quantity alone does not explain the rising number of successful penetrations. Over 2024 and into 2025 Russia made deliberate, iterative changes to the drones themselves and to how they are employed. Engineers moved fuel tanks into fuselages, reinforced engine compartments, and experimented with higher altitude and higher speed variants. They began pairing slow, bigger loitering munitions with smaller, quieter Lancet loitering munitions and reconnaissance platforms to complicate Ukrainian detection and targeting. Those adjustments reduce the single‑shot vulnerability that defined the earliest Shaheds and create tactical pairings where one munition draws fire or reveals a counter‑measure and another exploits the gap. These are practical, battlefield‑driven modifications, not miracles of machine learning.
We have also seen doctrinal adaptations. Russian strike packages increasingly combine decoy and imitation UAVs with larger strike airframes and missile salvos. Decoys force radars and interceptors to distribute attention. Missile volleys create windows of opportunity for the cheaper drones to close. The goal is resource denial: make Ukraine burn expensive interceptors on decoys and missiles so cheaper, massed drones can get through. ISW and other analysts logged the rapid increase in drones per package in early 2025 and documented the explicit use of decoys and saturation waves to overwhelm layered air defense umbrellas. That is not algorithmic genius; it is sound weapons engineering and operational design.
Claims that Russia has suddenly fielded AI swarms that outthink defenders should be met with skepticism. There are reports that some Lancet variants and reconnaissance platforms have received improved autonomy or machine‑aided target recognition, and developers have promoted enhanced EW resistance and alternative navigation channels. However, ISW reporting and open imagery suggest these capabilities remain incremental and operationally constrained. Autonomy helps reduce operator load in very specific tasks like visual target selection, but autonomy does not replace the logistics, communications security, and maintenance tail that defines real combat scale. In short, adding limited ML to a loitering munition can help select a target in the last seconds, but it does not solve the strategic problems of supply chains, attrition, or complex coordinated maneuvers at scale.
The economics are brutal and unsentimental. Cost analyses show that Shahed‑class drones are cheap relative to interceptors and many air defense missiles. When an attacker can produce tens of thousands of low‑cost strike drones, the attrition calculus tilts in favor of repeated harassment and depletion of air defense stocks. That is the cruel logic behind the nightly drone barrage. Don’t confuse cost effectiveness with precision or ethical legitimacy. Cheap strike drones can still cause disproportionate civilian harm when used against populated infrastructure and contribute to prolonged suffering and reconstruction burdens.
Ukraine’s response has been adaptive and layered. Mobile point defense systems, machine‑gun teams repurposed for low‑altitude interceptions, electronic warfare campaigns, and a growing set of indigenous interceptor drones have raised the cost of penetration. Ukrainian industrial responses have also tried to match production where possible, and Western donations of sensors, interceptors, and EW equipment have helped blunt some assaults. Yet as analysts have pointed out, no defense is immune to sustained massed attacks when the attacker controls steady production and launches. That is the tough reality Kyiv faces.
There are practical limits to Russia’s learning curve. Scaling factories, maintaining supply chains for specialized components, training crews, and keeping launch networks secure are nontrivial challenges. High launch rates require reliable logistics and fresh crews with clear maintenance practices. Leaking production schedules and covert procurement networks to skirt sanctions have helped Moscow, but they also create single points of vulnerability analysts can target with intelligence and interdiction. Moreover, the more you rely on massed, low‑tech attrition tactics, the less you can accomplish surgically. Strategic effects are uneven: energy infrastructure and morale are legitimate targets in a war of attrition, but decisive operational outcomes require combined arms effects that massed kamikaze drones alone cannot deliver.
If there is a lesson for militaries watching this conflict, it is a prosaic one. Effective use of unmanned strike systems is not a single breakthrough in autonomy or an off‑the‑rack doctrine. It is the product of iterative engineering, production scale, tactical pairing, and a willingness to accept an attritional model. Defenders respond by redistributing costs onto the attacker, innovating countermeasures, and searching for asymmetric counters like inexpensive interceptors and improved EW. For those of us in the trenches of prototyping and field testing, the human factors remain central: maintenance, logistics, training, and clear doctrine matter far more than buzzwords about autonomous swarms.
The Russian drone learning curve is real, and it is grounded in the same practical forces that drive any wartime adaptation. It favors states that can couple industrial scale with battlefield testing. It is not an inevitability that autonomy will dominate; it is an argument for investing in resilient, affordable air defense networks, sensible procurement that balances numbers with quality, and political strategies that reduce the appetite for open‑ended attrition. In short, watch the factories, not the hype.