In September 2020, while the world was looking the other way, busy with the pandemic and the American presidential campaign, Azerbaijan launched a military offensive against Armenian forces in Nagorno-Karabakh. The war lasted 44 days. When it was over, Armenia had lost control of territory it had held since the nineties, thousands of soldiers were dead, and military analysts around the world had to rewrite a couple of chapters in their manuals.
What happened in the South Caucasus was not merely a regional war between two small countries with a conflict frozen since the Soviet collapse. It was a laboratory. The first conventional conflict between two regular armies where unmanned aerial vehicles, drones, proved they could determine the outcome. Not as a supplement. Not as an auxiliary tool. As the decisive factor.
A Frozen Conflict That Thawed
To understand what happened you need to step back a bit. Nagorno-Karabakh is a mountainous enclave, ethnically majority Armenian, that formally belongs to Azerbaijan. In the early nineties, after a bloody war that left some 30,000 dead, Armenia took control of the territory and seven surrounding Azerbaijani districts. A ceasefire froze the situation in 1994. For more than two decades, the line of contact remained more or less stable, with periodic skirmishes and negotiations that went nowhere.
Armenia, backed by Russia, relied on classical deterrence: trenches, artillery, T-72 tanks inherited from the Soviet Union, and an air defence system that looked respectable on paper. Azerbaijan, enriched by Caspian oil, spent years modernising its army. It bought Russian, Turkish and Israeli weapons. But above all it did something nobody saw coming, or that many preferred not to see: it bet heavily on drones.
In June 2020, barely three months before the war, Azerbaijan completed the purchase of the Bayraktar TB2, the Turkish tactical drone that had already proved its effectiveness in Syria and Libya. To that it added a considerable arsenal of Israeli loitering munitions, the so-called "suicide drones": the IAI Harop, the Orbiter 1K and the SkyStriker. Platforms designed to loiter over an area of operations until they detect a target and crash into it with an explosive payload.
44 Days
The Azerbaijani offensive began on September 27. The first days revealed a pattern that would repeat until the end: TB2 drones, flying at altitudes Armenian air defences could not reach, identified positions, guided artillery and launched laser-guided munitions. Meanwhile, Israeli Harops dove like smart missiles onto radars and air defence systems.
The Armenians had, in theory, a defensive umbrella: Strela-10, Osa, Kub, Krug, and the more capable S-300 and Tor systems. But those systems were designed in the Soviet era to face very different threats—fast aircraft flying high, not small, slow machines operating at intermediate altitudes. The Strela and Osa lacked the range to engage the TB2. The S-300 and Krug were optimised for bigger, faster targets: to their radars, the drones were practically invisible. Only the Tor posed a real threat to the TB2, but Armenia had just a handful of units, perhaps six, an insufficient number to cover the front.
Azerbaijan also had an ingenious and brutal idea: it modified old Soviet An-2 biplanes with remote control systems and sent them as decoys. When the Armenians tried to shoot them down, they revealed the positions of their anti-aircraft batteries. The Israeli drones did the rest.
According to data verified by the Dutch group Oryx, which documents military losses with photographic evidence, Armenia lost more than 240 tanks, around 80 armoured vehicles, dozens of artillery systems and multiple rocket launchers, and a significant number of trucks and logistical vehicles. The asymmetry was brutal: Azerbaijan, according to the same sources, lost some 60 armoured units. The difference is explained, to a large extent, by unmanned aerial capability.
President Aliyev claimed his forces destroyed Armenian equipment worth over a billion dollars using the drones he had purchased from Turkey. Those figures should be taken with a grain of salt, as with any wartime declaration by a head of state, but the general proportion is not seriously disputed.
What the Drones Actually Did
It is worth being precise about what exactly the drones did and what they did not. The dominant narrative after the war, amplified by the spectacular videos Azerbaijan published on social media and on giant screens in Baku, was that the TB2s had won the war single-handedly. That is a simplification.
What the drones did was destroy the Armenian capacity to operate with armour and artillery in the open. Every time a tank column moved, every time an artillery battery repositioned, there was a drone watching from above. Reconnaissance was permanent, 24 hours a day, 7 days a week. And the response was immediate: a guided munition from the TB2, a Harop in a dive, or coordinates transmitted to Azerbaijani rocket artillery.
This forced the Armenians to hide. And an army that hides cannot manoeuvre, cannot counterattack, cannot sustain a dynamic defensive line. Static trenches became death traps.
What the drones did not do was take territory. That was done by Azerbaijani infantry, which advanced from the south through the less mountainous terrain, combining artillery fire with ground movement. The capture of Shusha, the strategic city whose fall precipitated the November 10 ceasefire, was a special forces operation over mountainous ground, not a drone strike.
The war was an exercise in combined arms, not an air show. But without the drones, Azerbaijan would not have been able to neutralise Armenian defences with the speed and economy of force that it did. That is the central lesson.
Turkey Did Not Just Export Drones: It Exported Doctrine
A detail that often goes unnoticed is that Turkey did not merely sell hardware. It provided operators, training, operational doctrine and, crucially, electronic warfare systems. The Turkish KORAL system was used to jam Armenian radars, blinding their air defences just before the drones struck. In other words: it was not the drone alone, but the complete ecosystem of detection, jamming and attack that made the difference.
There were also reports of Turkish F-16s at Azerbaijani bases, officially deployed for joint exercises just before the war and conveniently never sent home. Their probable function was not to fight but to deter the Armenian air force from intervening.
For Turkey, the Nagorno-Karabakh war was the definitive advertisement. Before 2020, the TB2 had already operated in Syria against Assad's forces and in Libya against Haftar's troops. But Nagorno-Karabakh was the case where everything worked, where the drone demonstrated that a country with limited resources could nullify the advantages of a conventional army equipped with Soviet-era materiel.
The commercial impact was immediate. Baykar, the company that manufactures the TB2, run by Selçuk Bayraktar (Erdogan's son-in-law), went from being a modest manufacturer to the world's largest exporter of combat drones. Its exports reached $1.8 billion in 2023 and 2024, and $2.2 billion in 2025. The company has contracts with 36 countries for the TB2 and 16 for its heavier drone, the Akıncı. Turkey controls, according to estimates from the Center for a New American Security, around 65% of the global combat drone export market.
What makes the TB2 so attractive is not that it is the best drone in the world. It is not. Its range is limited, its payload modest, its speed that of a light aircraft. Against modern and well-integrated air defences, it is vulnerable. What makes it attractive is its price. A package of six TB2s with ground station, weapons, equipment and training cost around $67 million according to the Turkish-Polish contract of 2021. For comparison: twelve American MQ-9B Protectors were sold to Australia for $1.6 billion. That cost ratio changes the rules of the game for any country on a tight budget.
What Conventional Armies Refused to See
The reaction of the world's major armies to the 2020 war was curious. There were analyses, reports, conferences, academic papers. And then, in many cases, nothing. The problem is that acknowledging the implications of Nagorno-Karabakh requires questioning multibillion-dollar investments in conventional platforms.
If a six-million-dollar drone can destroy a fifty-million-dollar tank, what is the point of continuing to buy tanks in traditional quantities? If a swarm of loitering munitions can neutralise an air defence system worth several hundred million, how much sense does it make to build your entire defensive doctrine around those systems?
These are uncomfortable questions. And like all uncomfortable questions, it was easier to minimise them than to answer them. It was said that Armenia had obsolete defences (true), that it had not operated its systems correctly (probable), that the terrain was peculiar (debatable), that a war between serious powers would be different (perhaps). All of that is partially valid. But it dodges the central point: that cheap, expendable, easily produced aerial platforms can nullify the advantages of heavy materiel.
Russia, which should have paid more attention than anyone, given that Armenia was its ally and many of the destroyed systems were Russian-made, did not learn the lesson. When it invaded Ukraine in February 2022, it sent armoured columns down predictable routes with scant anti-drone cover and suffered catastrophic losses in the first weeks. The same TB2s that had demolished Armenians in the Caucasus appeared over Russian columns in Ukraine, and the result was similar. Until Moscow adapted its tactics, incorporated electronic warfare on a massive scale and pushed the TB2 into a more marginal role, the lesson cost thousands of lives and pieces of equipment.
Taiwan, Korea, and What Comes Next
If Nagorno-Karabakh proved anything that transcends the specificity of the Caucasus, it is that drones democratise the capacity for aerial attack. You no longer need a fifth-generation fighter air force to project power from the sky. A small country with enough money can buy capabilities that twenty years ago were the exclusive province of great powers.
This has direct implications for the scenarios that most concern military planners in Washington.
Taiwan understood quickly. Since 2022, the island has accelerated a massive programme of domestic drone production. The government launched the "Drone National Team," an initiative to integrate private manufacturers into the defence industrial base. By 2024, Taiwan was producing some 10,000 units a year. The target for 2028 is 180,000 annually. In 2025, the Ministry of Defence announced the acquisition of nearly 50,000 drones over two years, with a budget of $1.5 billion.
The logic is straightforward: against a Chinese invasion force crossing the strait, Taiwan cannot compete in the number of ships, aircraft or soldiers. But if it can mass-produce cheap drones, drones that attack landing craft, that locate amphibious positions, that saturate the invader's air defences, the equation changes. A drone costing a few hundred thousand dollars sinking a landing craft carrying hundreds of soldiers is an asymmetry any military planner understands.
The philosophy Taipei is adopting, treating drones as disposable ammunition rather than prized platforms, descends directly from what was seen in Nagorno-Karabakh and then in Ukraine. The idea is that in a high-intensity conflict, thousands of drones will be consumed per day. What matters is not the individual sophistication of each device but the industrial capacity to produce them faster than the enemy can destroy them.
On the Korean Peninsula, the calculus is similar though the context differs. North Korea has the largest concentration of conventional artillery in the world aimed at Seoul. The ability to identify and destroy those positions with drones, before they fire or immediately after, is a force multiplier South Korea cannot afford to ignore.
The Real Lesson
Nagorno-Karabakh did not herald the end of tanks or the end of artillery, as some enthusiasts hastily proclaimed. What it heralded was the end of impunity for ground movement without air superiority. In the past, an army could move armour and artillery with a degree of protection if the sky was contested. Today, if the adversary has drones, the sky is never empty. Something is always watching.
The deepest consequence is economic. Drones compress the cost of air power in a way that alters strategic logic. An F-35 costs over a hundred million dollars. A TB2 costs a fraction. Obviously they are not equivalent; the F-35 can do things the TB2 cannot dream of. But to destroy a tank on a battlefield, a cheap drone is often enough. And if the drone gets shot down, you lost a few million. If the fighter gets shot down, you lost a fortune and a pilot trained over years.
This does not mean manned combat aircraft are going to disappear. It means the composition of armed forces is going to change: fewer exquisite platforms, more expendable ones. Less concentration, more distribution. The war of the future looks less like a chess match with valuable pieces and more like a flood of disposable pawns, each carrying an explosive charge and an electronic eye.
Nagorno-Karabakh was the first large-scale demonstration of this principle. Ukraine was the confirmation. Whatever comes next, if it comes, will be the logical consequence of a change that began in the mountains of the Caucasus, in the autumn of 2020, while the world was looking the other way.