How experience in Ukraine is redefining military balances with 5 million drones per year

The deploy­ment of drones and robots in com­bat, i.e. the “dron­isa­tion” of armed con­flicts, is one of the major tech­no­lo­gic­al and tac­tic­al chal­lenges of mil­it­ary research. Since the early 2000s, the grow­ing use of aer­i­al drones has been doc­u­mented in Kosovo, Afgh­anistan and Iraq. These stud­ies show a clear accel­er­a­tion in their use since 2015, par­tic­u­larly dur­ing con­flicts in Ukraine, the Middle East and the Cau­cas­us. How­ever, des­pite this intens­i­fic­a­tion, the exist­ing lit­er­at­ure remains largely scattered between tech­no­lo­gic­al ana­lyses of plat­forms, loc­al tac­tic­al stud­ies and often norm­at­ive stra­tegic inter­pret­a­tions. This frag­ment­a­tion lim­its the abil­ity to pro­duce a com­pre­hens­ive under­stand­ing of the effects of this dron­isa­tion. Added to this is the nov­elty of aquat­ic drones – above and below water – and land robots, which are poorly doc­u­mented due to their rel­at­ively recent appearance.

Net assess­ment, a mul­tidiscip­lin­ary ana­lys­is meth­od ori­gin­at­ing in the United States that aims to under­stand stra­tegic com­pet­i­tion to gain advant­ages, was developed by Andrew W. Mar­shall¹ in the 1970s with­in the US Depart­ment of Defence. Its cur­rent sci­entif­ic pur­pose is not so much to eval­u­ate an isol­ated cap­ab­il­ity, but to com­pare the tra­ject­or­ies of oppos­ing sys­tems over the long term, integ­rat­ing tech­no­logy, organ­isa­tion, industry and stra­tegic adapt­a­tion. Applied to dron­isa­tion, this approach leads to a spe­cif­ic research ques­tion: to what extent does the massive intro­duc­tion of drones alter the rel­at­ive net mil­it­ary value of the act­ors involved, when con­sidered as a sys­tem­ic phe­nomen­on sub­ject to attri­tion, sat­ur­a­tion and adverse co-evolution?

This art­icle is part of an applied research approach and lies at the inter­sec­tion of socio-tech­nic­al sys­tems engin­eer­ing and the polit­ic­al eco­nomy of mil­it­ary power. Thus, the thes­is under study is that dron­isa­tion is not only an autonom­ous tac­tic­al break, but a revolu­tion in mil­it­ary oper­a­tions. It is also an accel­er­at­or of sys­tem­ic dynam­ics iden­ti­fied and char­ac­ter­ised by dimin­ish­ing mar­gin­al returns, a shift in com­pet­i­tion towards industry and a reduc­tion in the dur­a­tion of rel­at­ive tech­no­lo­gic­al advantages.

Recent con­flicts provide a body of quant­it­at­ive data that can be used to ana­lyse dron­isa­tion as an indus­tri­al and sys­tem­ic phe­nomen­on. In the case of Ukraine, the author­it­ies have released fig­ures show­ing that Ukraine pro­duced 800,000 drones in 2023, 2 mil­lion in 2024, and aims to pro­duce 5 mil­lion drones in 2025, includ­ing 4.5 mil­lion FPV (First Per­son View) drones. Sim­il­ar fig­ures have been released by Russia.

FPV drones are very fast and designed for strik­ing fixed and mov­ing tar­gets. They are piloted in immer­sion mode, using vir­tu­al real­ity goggles. Small in size and of civil­ian ori­gin, mil­it­ar­ised with explos­ive charges and resi­li­ent links, these drones have a very low pro­duc­tion cost ($300 to $3,000) and can reach speeds of 100 to 500 km/h. Since the sum­mer of 2023, they have been respons­ible for more than 75% of the destruc­tion of armoured vehicles and bunkers in Ukraine. They are also neut­ral­ising more and more ground com­batants and oth­er drones (drone hunter drones). Pro­duced in their mil­lions each year by Ukraine and its allies, as well as by Rus­sia, they have become an essen­tial new weapon in close com­bat, with strikes car­ried out up to 30 kilo­metres away and fibre optic con­trol to avoid elec­tron­ic war­fare countermeasures.

Monthly ana­lyses indic­ate a shift from a few hun­dred drones engaged per month to sev­er­al thou­sand dur­ing cer­tain periods

These mil­lions of drones should be inter­preted as annu­al pro­duc­tion and acquis­i­tion flows, not as sim­ul­tan­eous oper­a­tion­al stocks, sug­gest­ing increas­ing employ­ment rates, high and con­tinu­ous attri­tion rates, but also increas­ingly massive use of effect­ive weapons that cost much less than mis­siles (anti-tank, cruise, bal­list­ic mis­siles, etc.) and which some­times need to be deployed sim­ul­tan­eously in their hun­dreds to break through enemy defences.

On the Rus­si­an side, data from inde­pend­ent mon­it­or­ing and open sources show a rap­id increase in the use of long-range drones such as the Shahed or deriv­at­ive sys­tems between 2024 and 2025. Monthly ana­lyses indic­ate a shift from a few hun­dred drones engaged per month to sev­er­al thou­sand dur­ing cer­tain peri­ods, with occa­sion­al attacks involving more than 600 drones in a single night-time raid deep behind enemy lines. On the Ukrain­i­an side, the use of long-range drones for recon­nais­sance and deep strikes is also grow­ing exponentially.

At the same time, for close com­bat, the arrival of FPV drones on the bat­tle­field from the sum­mer of 2023 onwards has led to increas­ing use on both sides, with up to 10,000 FPV drones being deployed per day. The latest offi­cial Ukrain­i­an data from 7^thFeb­ru­ary 2026, released by Gen­er­al Oleksandr Syrsky, Com­mand­er-in-Chief of the Ukrain­i­an Armed Forces, indic­ates that in Janu­ary 2026, Ukrain­i­an defence forces’ drone units reduced the Rus­si­an army by nearly 29,700 sol­diers (93.7% of the 31,700 Rus­si­an sol­diers repor­ted killed by the Ukrain­i­an army that month), while Rus­sia only man­aged to recruit 22,000 men dur­ing the same month. They also show that in Janu­ary 2026, Ukrain­i­an drones des­troyed 66,200 tar­gets (armoured vehicles, light vehicles, bunkers, com­mand posts, logist­ics depots, etc.). Finally, accord­ing to Ukrain­i­an intel­li­gence data, in 2026, Rus­sia plans to increase the num­ber of mil­it­ary per­son­nel using drones by 79,000, reach­ing a total of 165,000.

This increase in volume reflects the indus­tri­al­isa­tion of drone use, con­firm­ing Singer’s obser­va­tions² on the lower­ing of bar­ri­ers to entry for cer­tain air capabilities.

How­ever, sys­tems engin­eer­ing ana­lys­is requires link­age of these increas­ingly large num­bers of drones to the oper­a­tion­al archi­tec­tures in which they oper­ate. Drones must be con­sidered as a sub­sys­tem integ­rated into a socio-tech­nic­al sys­tem com­pris­ing com­mu­nic­a­tion net­works, digit­ised com­mand and con­trol chains, nav­ig­a­tion sys­tems, inform­a­tion pro­cessing resources, human oper­at­ors and logist­ics chains. The observed per­form­ance is not pro­por­tion­al to the num­ber of drones deployed, which is con­sist­ent with the non-lin­ear prop­er­ties of com­plex sys­tems described by Hughes³.

Avail­able feed­back shows that elec­tron­ic war­fare, par­tic­u­larly jam­ming, is one of the main causes of sys­tem­ic fail­ure. Ana­lyses pub­lished in 2024 and 2025 indic­ate that, in cer­tain oper­a­tion­al sequences, losses due to jam­ming, loss of con­nec­tion or degrad­a­tion of the GNSS sig­nal exceeded 30% of the drones deployed, regard­less of their level of tech­nic­al soph­ist­ic­a­tion. And dur­ing cer­tain com­bat sequences in Ukraine or Rus­sia, up to 80–95% of the drones used can be jointly neut­ral­ised by elec­tron­ic war­fare, nav­ig­a­tion war­fare and oth­er anti-drone meas­ures (in par­tic­u­lar kin­et­ic means deployed from the ground or in the air), not to men­tion tech­nic­al fail­ures and “friendly fire” destruc­tion, with a sat­ur­ated elec­tro­mag­net­ic spec­trum and dif­fi­cult vec­tor iden­ti­fic­a­tion. The effect­ive mil­it­ary value is there­fore dom­in­ated by net­work inter­faces and depend­en­cies rather than the intrins­ic per­form­ance of the platforms.

Fur­ther­more, the con­tinu­ous increase in the volumes engaged leads to sat­ur­a­tion effects not only on enemy defences, but also on one’s own sys­tems. Com­mu­nic­a­tion net­works, decision-mak­ing chains and human oper­at­ors are finite resources. Bey­ond cer­tain thresholds, the deploy­ment of addi­tion­al drones no longer res­ults in a pro­por­tion­al gain in mil­it­ary effect, illus­trat­ing the beha­viour of dimin­ish­ing returns typ­ic­al of com­plex socio-tech­nic­al sys­tems. This is undoubtedly where the rise of arti­fi­cial intel­li­gence, both in mis­sion pre­par­a­tion and back office oper­a­tions, as well as in embed­ded data pro­cessing mod­ules, is bring­ing sig­ni­fic­ant improve­ments that should logic­ally grow very rap­idly over the next few years.

The dis­tinc­tion between tech­no­lo­gic­al matur­ity and cap­ab­il­ity matur­ity is a cent­ral point of the ana­lys­is. The drones used in recent con­flicts often achieve high levels of tech­no­lo­gic­al matur­ity as defined by the Tech­no­logy Read­i­ness Levels form­al­ised by NASA in the 1990s [6]⁴. How­ever, as Saus­er et al.⁵ have shown with Sys­tem Read­i­ness Levels, the rel­ev­ant matur­ity for a com­plex sys­tem is spread across sev­er­al dimen­sions, includ­ing integ­ra­tion, oper­ab­il­ity and sustainability.

Empir­ic­al data from dron­isa­tion con­firms this dis­tinc­tion. Tech­nic­ally mature sys­tems may have low net mil­it­ary value if they are not integ­rated into resi­li­ent archi­tec­tures or if they can­not be pro­duced, renewed and improved at the rate imposed by attri­tion, adversity and coun­ter­meas­ures. Con­versely, tech­nic­ally simple drones that are mass-pro­duced, eas­ily repair­able and quickly adapt­able can main­tain high­er rel­at­ive effect­ive­ness over time. This is typ­ic­ally the case with the Ira­ni­an Shahed drone, pro­duced in Rus­sia under the name Ger­an, which is low-cost ($20,000 to $80,000 depend­ing on the pro­peller or tur­bojet ver­sion), uses basic tech­no­logy, is con­stantly being improved and will soon have four years of oper­a­tion­al effect­ive­ness in the con­flict in Ukraine.

Com­par­at­ive ana­lys­is of oppos­ing tra­ject­or­ies reveals that the ini­tial advant­ages asso­ci­ated with the intro­duc­tion of new types of drones are quickly off­set by countermeasures 

From a net assess­ment per­spect­ive, com­par­at­ive ana­lys­is of oppos­ing tra­ject­or­ies reveals that the ini­tial advant­ages asso­ci­ated with the intro­duc­tion of new types of drones are quickly off­set by coun­ter­meas­ures. The meas­ur­able con­sequence is an increase in the volumes required to achieve an equi­val­ent effect, shift­ing com­pet­i­tion from the tech­no­lo­gic­al domain to the indus­tri­al and organ­isa­tion­al domain. The estim­ated unit costs for Rus­si­an or Ukrain­i­an long-range drones, often ran­ging between $30,000 and $50,000 accord­ing to sources pub­lished between 2023 and 2025, must be con­sidered in rela­tion to the observed pen­et­ra­tion and attri­tion rates.

This dynam­ic con­firms the con­clu­sions⁶ that tech­no­lo­gic­al dif­fu­sion tends to favour defence and reduce the dur­a­tion of offens­ive advant­ages. Dron­isa­tion accen­tu­ates this trend by accel­er­at­ing adapt­a­tion cycles and mak­ing indus­tri­al sus­tain­ab­il­ity essen­tial. The decis­ive cap­ab­il­ity then becomes the abil­ity to main­tain a faster and more robust adapt­a­tion tra­ject­ory than that of the adversary, while pro­du­cing very large quant­it­ies of drones, rather than the occa­sion­al pos­ses­sion of more power­ful systems.

Empir­ic­al data from recent con­flicts con­firms that the mil­it­ary value of drones depends less on their indi­vidu­al per­form­ance, which has cer­tainly improved sig­ni­fic­antly over the last five years, than on their archi­tec­tur­al integ­ra­tion into a com­bat sys­tem, their over­all cap­ab­il­ity matur­ity and the indus­tri­al sus­tain­ab­il­ity of adversari­al trajectories.

In the con­text of net assess­ment, dron­isa­tion appears to reveal the fun­da­ment­al prop­er­ties of socio-tech­nic­al sys­tems in con­tested envir­on­ments: dimin­ish­ing mar­gin­al returns, crit­ic­al depend­ence on inter­faces, cent­ral­ity of industry and reduc­tion in the dur­a­tion of tech­no­lo­gic­al advant­ages. This inter­pret­a­tion is con­sist­ent with the sem­in­al works of Mar­shall⁷, Hughes and Horow­itz⁸, while provid­ing a unique empir­ic­al basis linked to the scale of data avail­able since 2022.

The lim­it­a­tions of the ana­lys­is lie mainly in the het­ero­gen­eous qual­ity of open data and the inab­il­ity to access com­plete met­rics on total cost and actu­al attri­tion. These lim­it­a­tions do not inval­id­ate the approach but high­light the need to con­tin­ue applied research on a con­sol­id­ated empir­ic­al basis. For both research and decision sup­port, the chal­lenge is not so much to determ­ine wheth­er drones are “decis­ive,” but to under­stand under what sys­tem­ic con­di­tions they will be less so or even cease to be so.