Sixty to eighty percent
In February 2025, RUSI published field notes from Ukraine that contained two numbers worth keeping side by side. Tactical drones, mostly FPVs built from commercial parts, accounted for 60 to 70 percent of damaged and destroyed Russian systems. And 60 to 80 percent of those FPVs failed to reach their targets.
Both numbers are true at once. The cheapest weapon on the battlefield became the most lethal while most of them missed, crashed, were jammed, or never found a target. Mass made up for the misses. That is what attritable means in practice: a force that expects to lose most of its aircraft and plans around it.
The United States is now building that kind of force. This paper looks at what it takes to run one, and argues that the hard part is managing the force around the aircraft, a part that rarely makes it into the solicitation.
The mass is arriving
The Drone Dominance Program's first Gauntlet led to orders for about 30,000 small drones. The second, whose results posted September 15, ran 1,858 sorties across 23 platforms from 19 companies, and the department plans to buy about 60,000 from that pool. The goal is more than 200,000 by 2027. Every long-range platform in the second round reached 15 kilometers. In the first, only 24 percent reached 10.
The Army says it has fielded 10,000 drones and is fielding 30,000 more, enough for the squad level. The Marine Corps set out to put 10,000 small drones into infantry battalions and reconnaissance units by the end of this year. Ukraine is further along. Its units ordered more than 590,000 drones through the Brave1 Market between January and September.
Consumable is a policy word
Calling a drone attritable is easy. Making the institution treat it that way takes longer. This spring, small drones in the Army were still listed as nonexpendable property, so losing one could trigger a property-loss investigation and a soldier could have pay docked. Officials were still debating in weekly meetings whether drones were attritable, consumable, expendable, or durable.
By September the language had shifted. At Saber Junction, which used about 2,000 drones, half of them consumable, Brig. Gen. Terry Tillis described them simply: "they're like a munition." That is the right model. Nobody runs a property-loss investigation on an expended artillery round. But munitions have their own accounting. A unit knows how many rounds it drew, how many it fired, at what, and on whose authority. The attritable force needs the same discipline, applied to aircraft that are also sensors, data sources, and sometimes reusable.
Six jobs around the airframe
Every drone arrives with its controller, batteries, and the vendor's ground software. Six jobs sit around it, and none of them ships in the box. The unit has to register the aircraft and know its state. It has to train pilots to a standard. It has to task aircraft from several vendors against one set of targets. It has to sustain them through batteries, spares, and the next wave. It has to account for what was expended and why. And it has to review what each sortie saw and did.
Today each of those jobs lives in a different place: a hand receipt, a training spreadsheet, a radio net, a battery log, a property book, and a debrief whiteboard. At 200,000 aircraft, the overhead of keeping those records in sync by hand will consume the time the drones were supposed to save.
Training is the binding constraint
Col. Gregory Merkl, who runs the Army's Drone Dominance effort, was candid in September: the Army has the volume, and the work left is training. Qualified pilots have not reached the squads. His lesson from Saber Junction was that soldiers must learn to navigate the electromagnetic spectrum the way they navigate terrain.
Col. Nick Ryan at the Aviation Center of Excellence made a related point in February. A cargo drone takes far less skill than a lethal FPV, so training has to be specific to the mission. He also asked how many drones, batteries, and controllers a squad can actually carry, contrasting a load of 20 drones with a realistic rucksack maximum of about five.
The Army has chosen to decentralize, spreading drones and lessons across every unit instead of building dedicated drone formations. That puts the training burden on every squad leader. Those leaders need training records that build themselves from the same system the squad uses to fight, and debriefs that do not require an instructor to reconstruct every sortie from memory. EdgeLance's Coach scores radio traffic against doctrine for exactly this reason.
Tasking across vendors
The Drone Dominance pool already includes more than a dozen vendors, each with its own ground station. A platoon running aircraft from three of them has three ideas of what the target list is. Two operators can put two drones on one target and leave a second target uncovered without either knowing.
EdgeLance assigns work as tasks that any asset can bid on, whatever the vendor. A target is offered to the aircraft that can reach it with the right payload and endurance. The best bid holds the task on a lease. If the aircraft goes quiet, the lease lapses and the target goes back on offer. The result returns as a receipt on the task. We described this in the bomber drone post and tested the assignment logic against counter-swarm scenarios. Offers and receipts are small messages that ride the team's mesh, so tasking keeps working when the wider network does not.
Sustainment and the next wave
An attritable force fights in waves. What matters before the second wave is how many aircraft are actually ready, which is a different number from how many are in the conex. Batteries need charge time. Launch points have limited throughput. Some aircraft came back damaged, and some never came back.
EdgeLance keeps every asset on one roster with its position, payload, battery, and health, fed by each vehicle's own telemetry, and the tasking layer only offers work to aircraft that can do it. The roster is the base for planning the next wave: which aircraft and launch points will be ready, and how much coverage the unit can sustain. We are extending it to model charging, reserves, and launch-site throughput across multiple waves, so the plan reflects what will be ready instead of what is on the hand receipt.
Accounting and review
When an aircraft is expended, EdgeLance records it against the task it was expended on, with the approval that authorized the task and the track it was sent against. The property question becomes a munitions question: drawn, assigned, expended, on whose authority, with what result. That record exists because the work was done in the system, not because someone filled in a form afterward.
The same record drives the review. Every sortie's tasks, video clips, detections, and receipts sit in the mission record, so the debrief can replay the fight, see which aircraft found targets and which did not, and connect misses to causes such as jamming, pilot error, or a bad assignment. Over time that is how a unit pushes its miss rate down from RUSI's 60 to 80 percent. The record is owned by the unit, and when the mission ends, the sensitive data is destroyed while the record of decisions survives.
What programs should ask for
Programs buying attritable systems can require the management layer alongside the airframe. Ask every vendor to publish telemetry and accept tasks in an open, documented form, so one tasking layer can run the whole pool. Ask that expended aircraft be recorded against the task and authority that consumed them. Ask that training sorties and combat sorties produce the same record, so readiness comes from data instead of a separate report.
Ask that all of it work with the network down, on the hardware the unit already carries. The Gauntlets have done well at scoring aircraft. They could also score how well an aircraft fits a common tasking and accounting layer, which would push every vendor toward the same answer.
Mass is a management problem
Ukraine's drone force works with a miss rate that would end any traditional weapons program, because it was built to absorb losses and learn from them. The American force is about to reach comparable numbers. The airframes are ordered and the money is in the request.
What decides whether 200,000 drones become a capability is the layer that registers, trains, tasks, sustains, accounts for, and reviews them. Units will build that layer one way or another. The choice is between building it from spreadsheets, radio logs, and memory, or running it on software designed for a force that expects to lose most of its aircraft. The platform page shows how EdgeLance runs that loop.