In an era of drone warfare, it is past time for the U.S. or other NATO nations to internalize drone research, development, and production.
In the early years of the American Revolution, George Washington commissioned the Springfield Armory to manufacture cartridges and gun carriages for the war. The armory later went on to design and manufacture a large line of weapons and ammunition, from the Model 1795 musket, to the Model 1903, to the M1 Garand and, finally, the M14 rifle. Such armories were not unique to the United States. The British Royal Small Arms Factory (better known as the Enfield Armory) has almost as long and prestigious a history, including the creation of the famous Lee-Enfield rifle. France, too, had the Manufacture d’armes de Saint-Étienne, which produced designs from the French Revolution up until the FAMAS rifle in 1978.
Today, however, in-sourced weapons development is increasingly rare. Although the U.S. retains some indigenous capability with the Picatinny Arsenal, manufacturing facilities are run explicitly by outsourced defense contractors, which often engage in rent-seeking behavior.
It is time for the U.S. or another major Western country—such as a member of NATO or Five Eyes, South Korea, Japan, or Taiwan—to bring back the tradition of in-sourced weapons development. Below, I consider the “Springfield Dronery”—dedicated to developing, advancing, and producing small drones in a tightly integrated, low-cost, explicitly government-led operation.
The Role of Small Drones in Warfare
Both Ukraine and Russia have innovated greatly in small drones over the past three years of the conflict. Ukraine initially used crude grenade-dropping drones; it now produces over a million small drones every year, each one armed with roughly a grenade’s worth of explosives.
From a tactical standpoint, such drones have effectively supplanted the role of field artillery with organic reconnaissance: Instead of trying to place a 50 kilogram shell within 50 meters of the target—the job of an artillery piece—the drone’s job is to place an explosive (such as a 250 gram anti-personnel payload or a 4 kilogram tank-busting explosive penetrator) within a few meters of the target. Ideally the drone should cost the same or less: A 155 millimeter artillery shell costs $3,000 to $5,000, meaning that the ideal small drone would cost about $1,000. This pricing is typical for Ukrainian FPV drones—but, as explained below, anything involving U.S. military contractors can be 10 to 50 times more expensive; the Switchblade 300, for example, is reported to cost over $50,000 each for the same sized payload.
Additionally, current FPV drones are limited in their efficiency by the control model: An operator needs to guide the drone to the target. This requires both a reliable data link and the attention of a well-trained soldier.
I’ve long argued that autonomy mitigates these limitations, allowing the drone to seek targets in absence of reliable communication and removing the need for a one-to-one ratio of skilled operator to in-flight drone; instead of launching one drone at a time, the attacker can launch dozens or hundreds. The problem becomes one of placing a 5-100 kilogram package within 10-50 kilometers of a target. There are numerous ways to solve this delivery problem, ranging from ISO containers, narco-subs, larger carrier-drones, high-altitude balloons, or simply placing a dozen drone launchers in the bed of a Hilux Champ.
From there, it is basically a numbers game: The more drones, the better. The aim of the Dronery is to enable an autonomous army of a million drones, an aerial force tailored to ensure that any invading soldier faces perhaps a dozen automated weapons—all focused on stopping the invasion, where any given strike can field dozens or even hundreds of drones to overwhelm possible defenses.
Research and Development Challenges
The challenges of developing autonomous drones combine hardware and software problems, requiring close collaboration between those designing the drone’s computer and those programming the resulting system. These challenges are best overcome with a small, well-coordinated team.
The first challenge is developing low-cost machine vision that is accurate enough to reasonably identify targets, avoid collisions, and navigate in a GPS-denied environment. The second challenge is networking—enabling the drones to be able to coordinate effectively even in the presence of substantial interference. And the third challenge is ensuring that these designs can be produced for a very low cost.
Take a dozen highly skilled engineers, each with a broad background in and deep understanding of a particular subspecialty. You need a hardware designer or two to develop the circuit boards, a control theory expert for the low-level autopilot, one or two machine-vision experts to enable the drone to understand its environment, an explosives expert, and a network designer for the mesh network for interdrone communication to allow the drones to communicate directly with each other and to relay messages back to the controllers.
This engineering team needs to be tightly coupled with the prototype and production manufacturing: designing not to develop the best possible design, but to create the best manufacturable design—one that uses low-cost components and that can be assembled swiftly and reliably.
The Small Drone Supply Chains and Assembly
One remarkable feature of small drones is that the primary supply chain—for everything except the actual warhead—is divorced from the traditional military supply chain. Even the military component is a commodity item, as it is typically a grenade-sized explosive and often just an off-the-shelf 40 millimeter grenade or rocket-propelled grenade warhead. As such, there are many suppliers of components that compete aggressively on price and quality.
Small drones are powered by $3 accelerometers, $10 microcontrollers, less-than-$20 cameras, and $50 computers available from Digikey or similar suppliers. The bare circuit boards, although custom-designed, are only a couple of dollars each when ordered from a typical fabrication facility, of which there are hundreds around the world. Assembling the circuit boards is best done in-house: A $150,000 investment allows prototype production while a $1 million investment enables producing thousands of drone boards a day.
Drone chassis are similarly straightforward, composed of either expanded polypropylene foam for low-cost fixed-wing designs or computer numerical control (CNC) cut carbon-fiber sheets for quadcopters, with threaded inserts pressed in manually for screws. The batteries are common lithium polymer pouch cells available from a wide variety of manufacturers, and the motors are also commodity products with numerous vendors.
Consequently, the Dronery should in-source final assembly, utilizing a pick-and-place line for assembling circuit boards, CNC cutters for carbon-fiber frames, and a focus on designs that are optimized to minimize labor during assembly. Such a setup would be both very flexible and low cost, minimizing marginal cost per drone to the cost of components and a few minutes of a skilled employee’s time to bolt the components together and plug in the connectors. Everything upstream of this is a “low-end commodity”—items where there are plenty of vendors and a large amount of interchangeability, allowing bulk purchases on the civilian market.
In-house design and assembly would not only minimize costs—it would maximize flexibility. A new prototype would be designed; boards and components would arrive within a week; and a dozen would be flying out the door a day later. This would enable an iterative design flow by ensuring that the critical steps can be done in-house. When a particular design is successful, production of that design could be ramped up, from a dozen a day to thousands.
The Role of Defense Contractors
Perhaps the biggest (and possibly only) advantage of a defense contractor over a government-led dronery is the ability to hire quickly and then pay competitive salaries. The overall federal pay schedule, even for a high-cost area like Washington, D.C., is low compared to the private sector. In the D.C. area, a GS-9 job opening (with a starting salary of $70,000 for a master’s-degree-level position) wouldn’t even get a resume from a competent engineer, let alone an engineer with substantial experience.
Governments have long recognized this problem and, instead, simply launder personnel through existing contractors. This can take the form either of contractors working on premises or through rotator programs such as the Intergovernmental Personnel Act, or the local equivalent.
In the small drone space, however, Western military contractors exist largely for rent-seeking—taking advantage of a monopoly on a particular product in order to generate profit for the contractor, rather than savings for the government.
Take, for example, the cost of software. Building a piece of robust software can be expensive, but the cost of producing one copy or a billion copies is effectively the same. Every time a contractor—which was already reimbursed for developing the program in the first place—charges even $.01 a copy, it is rent-seeking.
A good example of possible rent-seeking is Anduril’s recent contract with the Marines, worth a reported $200 million, to integrate its “Anvil” interceptor into the Marine’s MADIS anti-air system. The Marines expect to field less than 200 MADIS systems by 2029. So, although volume is not specified, it is reasonable to assume this contract will involve deploying perhaps a couple thousand drones.
The Anvil design is a 10 pound quadcopter with a fairly sophisticated computer and sensor suite that should cost $2,000 a copy in quantity, in an airframe that should cost perhaps $500. Which means all the costs beyond this—perhaps another $100,000 a drone—are for the sole supplier of Anvil drones either to recover the research and development (R&D) costs or to extract rents from the government. There is no way to distinguish between these two possibilities: reasonable compensation for the R&D costs or rent-seeking behavior.
If the government does wish to utilize defense contractors, it is critical to separate out R&D from production. The R&D contracts must specify that all rights become property of the government, allowing the government to select as many producers as desired. This prevents contractors from being able to rent-seek to the government’s detriment.
The Cost of the Dronery
By government standards, the Dronery would be an inexpensive project. The R&D costs are likely to amount to a couple hundred thousand dollars of prototype equipment (to enable the in-house development of prototypes) and the cost of 10 highly skilled engineers and developers.
Let us assume that the cost per engineer is $400,000 a year, based on the need for an industry competitive salary, overhead expenses, and whatever costs are incurred by laundering the hiring through contractors capable of paying competitive salaries. This still results in an annual expense of under $5 million a year for the entire R&D program.
In return, the government gains a unique capability: the ability to produce hundreds of drones a day at cost. If each grenade-carrying drone costs $1,000, and each larger, tank-killing drone costs $5,000, fielding a million-drone fleet would cost just a little over a billion dollars. To give a sense of scale, a fleet of a hundred thousand drones would cost less than a single F-35B.
The physical plant required is also minimal. A modest industrial building is more than sufficient to house both the R&D staff and the production lines. Drone production uses relatively small machines. Even a high-end, fully automated board assembly line is just 20 meters long, and each CNC cutter for the carbon-fiber sheets requires only a few square meters of floor space. Over time, further commodity items (notably the motors and cameras) could then be in-sourced, but only if they appear vulnerable to potential supply issues.
And the risk is low: If the Dronery succeeds, the capability would be unprecedented. If the Dronery fails, however, the money at risk is negligible by military standards—just a few million dollars for personnel and equipment. The majority of the Dronery’s costs lie in producing the actual drones, not in developing them.
Who Should Do It?
Although I would love to see the United States adopt this model, the current U.S. military procurement system makes this improbable, as rent-seeking contractors would likely work to ensure that the modest annual R&D spending necessary would never occur. On top of this, the current administration’s policies would make it difficult to attract the talent necessary for this endeavor.
Instead, a smaller NATO nation—one that faces a potential invasion and could therefore benefit from an army of a hundred thousand or a million drones—is more likely to attempt this model. Developing and building a million-drone army for the cost of a few fighter jets would make any potential invader think twice. Finland or the Baltic states (Lithuania, Latvia, or Estonia) seem like obvious candidates; however, Canada or Denmark might consider this model as well. All of these countries would benefit from the ability to deploy massive fleets of small drones.
Such a Dronery could explicitly support Ukraine’s effort, providing tens of thousands of “free” weapons to Ukraine in return for honest evaluations and opportunities for optimization. And given the low cost of production, the Dronery could then sell at cost-plus to other Western allies.
The U.S. and other Western nations have long relied on internal weapons manufacturing to support defense operations. As the world enters a new era of warfare, it is time to reconsider the merits of the in-house model.
– Nicholas Weaver is a senior staff researcher focusing on computer security at the International Computer Science Institute in Berkeley, California, and Chief Mad Scientist/CEO/Janitor of Skerry Technologies, a developer of low cost autonomous drones. All opinions are his own. Published courtesy of Lawfare.
