On 9 July, eight Iranian ballistic missiles were fired at Muwaffaq Salti Air Base in Jordan. All eight were intercepted. Nobody was hurt, nothing was damaged, and the incident barely registered outside the region.

Eight days later the same base was attacked again, and this time things got through. Three Americans are dead.

The difference between those two nights is not really about radar or missile performance. It is about a spreadsheet, and the spreadsheet is losing.

The exchange rate

Air defence has always cost more than attack. What has changed in the last four years is the size of the gap.

A Shahed-type one-way attack drone is a lawnmower engine, a wing, a warhead and a satellite navigation receiver. Open-source estimates put it between twenty and fifty thousand dollars. It is slow, it is loud, it is not stealthy and it is easy to shoot down.

Shooting it down with a PAC-3 MSE costs around four point two million dollars.

You can win every single engagement on that exchange and still lose, because the attacker is buying a hundred of something for the price of one of yours. No defence budget in the world is structured to sustain that ratio indefinitely.

Why you cannot just use something cheaper

The obvious response is to stop using million-dollar missiles on cheap drones, and every serious air force is working on exactly that. Guns, lasers, cheap interceptors and electronic attack are all in development or in service.

The difficulty is that the decision has to be made in seconds, on incomplete information, and the cost of being wrong is asymmetric in the other direction.

A radar track is not labelled. The operator sees speed, altitude, heading and radar cross-section. From that he has to judge whether the inbound is a fifty-thousand-dollar drone he can afford to engage cheaply or leave alone, or a two-million-dollar ballistic missile that will kill people if it arrives. Get that wrong in the expensive direction and you have wasted an interceptor. Get it wrong in the cheap direction and you are writing letters to families.

Faced with that, operators default to the expensive answer, because that is the rational choice for the individual engagement even when it is the losing choice for the campaign.

The saturation trick

Attackers understand this perfectly, which is why modern strike packages are deliberately mixed.

Send only ballistic missiles and the defence engages all of them with the right weapon. Send only drones and the defence learns to handle them cheaply. Send both, in the same raid, arriving together, and you force the defender into an impossible allocation problem in real time.

The drones are not really trying to hit anything. They are there to absorb interceptors, occupy radar channels and consume the defender's attention for the ninety seconds that matter. Behind them, the ballistic missiles arrive at a defence that has already spent part of its magazine on decoys.

That is the most plausible explanation for the difference between 9 July and 17 July at the same air base. Not that the defences got worse. That the attack got smarter.

Magazine depth

Underneath the cost problem is a harder one that money cannot quickly solve: you cannot buy interceptors faster than the factories make them.

Global Patriot interceptor production runs at roughly 850 rounds a year, projected to rise to around 1,130 by 2027, with only about half of that going to Europe. Diehl builds perhaps 500 to 600 IRIS-T SL a year, heading toward 800 to 1,000. Those are the real constraints, and they are measured in years of factory throughput.

Now set that against consumption. A single heavy night of defending one country can expend dozens of rounds. Reporting through 2025 suggested US Army interceptor stocks had fallen to somewhere near a quarter of required levels after sustained combat expenditure and large-scale transfers to Ukraine, with analysts converging on a restocking timeline of three to eight years for that expenditure alone.

That is the title CSIS gave an entire study of the problem, and the framing is the important part. The right question is not what the interceptor costs. It is what the thing behind you is worth. Four million dollars to protect an empty stretch of desert is waste. Four million dollars to protect a hangar with aircrew in it is trivially good value.

The trouble is that an attacker who understands this can bankrupt you by making you defend everything.

What changes

Three things are moving, none of them fast enough.

The first is cheap effectors: gun-based systems, small interceptors and modified air-to-air missiles priced in the tens of thousands rather than the millions. The second is directed energy, where the marginal cost of a shot is essentially electricity, and which is finally reaching deployable power levels after decades of promise. The third is electronic attack, defeating the guidance rather than the airframe, which is cheap and reversible and is a large part of why every serious air force is currently buying stand-off jammers.

That last point is the oldest answer of all, and it is worth noting that it is the one being used right now. The cheapest way to stop being shot at is to destroy the thing doing the shooting, which is why the F-16s went back to the Gulf this week.

Until the cheap defensive options arrive at scale, the arithmetic stays ugly. Every night that a defence works perfectly, it gets more expensive. Every night it does not, somebody dies. That is the actual state of air defence in 2026, and no amount of technology has yet made it a good trade.

Sources: Center for Strategic and International Studies, RUSI, Modern War Institute, Military Times, US Army budget documents, Bruegel, Air and Space Forces Magazine.