The $464.8 Million LOCUST Award Is a Queueing Test, Not an Energy Bill

Russia's September 2 overnight salvo comprised 163 aerial targets, roughly half of them jet-powered. The US Army awarded AeroVironment a $464.8 million contract to supply LOCUST laser weapons. The award marks the service's first production contract for a high-energy laser weapon system. Those facts belong in the same paragraph, because the contract will be judged against the salvo.

The case for directed energy has been an energy-price argument: advocates emphasize the electrical cost of a laser shot over the cost of the interceptor it replaces. TASS recently quoted a military expert putting the cost of an IRIS-T missile at about €400,000. Militaries are openly hunting for cheaper ways to destroy one-way attack drones than spending surface-to-air missiles on them. The pitch has been durable because it is true. But a production contract is a claim about repeated performance under conditions the seller does not choose, and once you sign one, cost per shot is no longer the question.

Production does not test the energy bill; it tests the queue.

What The Army Actually Signed

The award tasks AeroVironment with delivering dozens of LOCUST X3 systems over the next few years. The agreement also tasks the company with providing ongoing support and training under the Army's Enduring-High Energy Laser counter-drone program. The X3 is the third LOCUST system in the company's catalog. It is designed to defeat drone threats through Group 3 using a 30-kilowatt laser.

Production will be backed by a $30 million investment in the Albuquerque facility. The contract followed a March counter-drone test run with the Pentagon and the FAA. As program facts go, these are solid ones: real money and real deliveries on a schedule.

None of that is my complaint. My complaint is with an industry habit: demonstrate a capability once, under conditions the demonstrator helped arrange, then discuss it as though it were fielded capacity against a thinking adversary. A March test with the Pentagon and the FAA is a data point, and a useful one; it is not a raid. I have no evidence that LOCUST underperforms its specifications, and I am not claiming any. I am claiming that the specification measures the wrong axis, because the axis that matters is set by the other side, in arrivals per minute.

A Laser Is A Server, Not A Magazine

A missile battery can put several interceptors into the air against several tracks at once; the engagements run in parallel, and the binding constraint is magazine depth. A laser cannot parallelize. It must hold line of sight on one target for a finite dwell time to deposit enough energy to kill it, then slew to the next track, where it must reacquire and settle before the dwell clock starts again. Assume, for the model, that one aperture engages one target at a time. That turns the defended point into a single-server queue in which the attacker sets the arrival rate while physics and the weather set the service rate.

Every variable that matters lives in that service rate. Atmospheric attenuation stretches dwell, so haze and rain cut throughput exactly when a competent attacker would choose to fly. Target mix moves the clock too, because a Group 3 airframe and a small quadcopter do not cost the same seconds, and a fast target spends less time inside the engagement envelope in the first place. Readiness bites harder here than it does for launchers: a battery at 85 percent uptime still has rounds in the tubes, but a laser that is down is a queue with no server at all. When arrivals outrun service across the exposure window, targets do not wait politely; they leak.

Run A Raid Through The Queue

Take that September 2 salvo as the reference threat and hand a single defended point a modest slice of it: 30 targets across a 20-minute window, one arrival every 40 seconds on average. Assume, generously, 20 seconds of total service per kill, covering slew and acquisition as well as dwell and kill assessment. On paper the server keeps up with a two-to-one margin. Now let half the targets be jet-powered so each spends perhaps a minute inside the envelope instead of several, then stretch every dwell by half for weather and let the arrivals bunch, because arrivals always bunch. In this analysis, waiting time rises sharply as modeled utilization nears capacity, and here a target that waits too long is not late; it is through.

Every number in that paragraph is an assumption, and that is exactly the problem: I have not found the real figures in public materials about the award. Meanwhile the arrival side keeps getting cheaper to build. Chinese manufacturers are expanding production of inexpensive, partly 3D-printed interceptor drones. One producer says it already sells thousands of interceptor drones per month, including to Russia.

Those particular airframes are defensive, but the industrial lesson transfers, since a line that prints thousands of small drones a month can feed either end of this queue. Cheap interceptors attack the cost problem by adding servers in parallel; a laser attacks it by making one expensive server's shots nearly free. When arrivals bunch, server count beats shot cost, and no consumable price changes that.

The Denominator Belongs In The Contract

So the number a buyer should demand is not cost per firing. It is cost per defeated raid, computed at a disclosed arrival rate and target mix, under disclosed weather assumptions and a disclosed readiness level. Cost per firing is a numerator trick: it puts the cheapest figure in the headline and hides the one that decides whether the airfield is still usable at dawn. A serious procurement annex would specify a reference raid with size, mix, and geometry. It would add a weather table and an uptime floor, and it would state the leakage rate the seller is prepared to sign against.

That denominator also fixes the unit of purchase. If one laser leaks above some raid size, then the real product is not a laser; it is a cluster of lasers, plus cueing and whatever engages the leakers, and the cost of that cluster divided by raids actually defeated is the figure that belongs in the decision memo. Requirements written as wish lists, with no exchange-ratio arithmetic behind them, are how a service ends up owning hardware that wins every demonstration and loses every saturation night. The LOCUST buy deserves better than that fate, and the way to give it a fair chance is to score it honestly from the start.

Hold Proposals To The Same Standard, Starting With Mine

This test should hit paper studies harder than it hits production hardware, not the other way around. My company, Kibernan, has produced six complete, costed engineering programs, including DRONE WALL. It is a proposal. No Kibernan hardware has been built, flown, or fielded, and we say so in writing; every published figure in those programs carries a maturity label that marks it as modeled or specified, calculated or objective.

That discipline exists because a costed proposal is the easiest artifact in the world to make look good, and I distrust our own numbers on principle until they survive hostile arithmetic. If DRONE WALL is ever taken seriously, it should be scored against the same denominator I am asking of the LOCUST buy: cost per defeated raid at stated arrival rates and mixes, under stated weather and readiness. Anything less is marketing with units attached, whoever publishes it.

This award is the service's first production contract for a high-energy laser weapon system, and that is a real milestone; someone had to go first. The agreement tasks AeroVironment with ongoing support and training under the Army's E-HEL counter-drone program. A $30 million investment in AeroVironment's Albuquerque facility will support production. I have not found public throughput arithmetic for the program. The energy-price argument won the procurement, but the queueing argument will decide the raids, one dwell time at a time.

The queue does not read the datasheet.