When the world's at stake,
go beyond the headlines.

National security. For insiders. By insiders.

National security. For insiders. By insiders.

Join War on the Rocks and gain access to content trusted by policymakers, military leaders, and strategic thinkers worldwide.

Cogs of War
Cogs of War
Cogs of War

Factories First: Winning the Drone War Before It Starts

July 17, 2025
Factories First: Winning the Drone War Before It Starts
Cogs of War
Cogs of War
Cogs of War
Factories First: Winning the Drone War Before It Starts

Factories First: Winning the Drone War Before It Starts

Martin C. Feldmann and Gene Keselman
July 17, 2025

Wars are won by factories before they are won on the battlefield. Recognizing that the United States lacks the manufacturing depth for the coming drone age, Secretary of Defense Pete Hegseth issued his “Unleashing U.S. Military Drone Dominance” memo, pledging to “bolster the nascent U.S. drone manufacturing base” and delegating buying power to frontline units. That promise, however, will take far more than procurement tweaks and approving “hundreds of American products for purchase” — it demands a national-level, wartime-scale industrial mobilization.

Beijing shows why. Chinese civilian manufacturers have the capacity to retool in under a year to turn out one billion weaponized drones annually — without slowing the rest of China’s economy. By our calculations, that would require less than 1 percent of its assembly capacity, less than 5 percent of its battery output, and a fraction of its printed-circuit-board capacity. If the United States is to deter a capability of that magnitude, industrial policy should focus on enabling the mass production of autonomous systems. Part of the effort should be the establishment of new enterprises and the development of new capabilities. At least as important, however, will be the efficient utilization of the existing national industrial base.

Start with plastics: The United States has a strong plastics industry and produces about 130 billion pounds per year. It is also strong in the injection molding industry — Milacron, one of the world’s leading molding machine makers, sits in Cincinnati, not Shenzhen. A modern drone airframe, designed as two bonded injection-molded shells and weighing approximately a pound, would consume less than 1 percent of that capacity even at a billion units per year.

What about motors? First-person view or small loitering drones run on brushless direct-current motors consisting of several small, machined, mostly steel and aluminum components, strong magnets, and a stator holding the copper windings. Tier 2 automotive suppliers in the United States are already familiar with machining more than a billion such precision parts per year. Companies such as Haering Precision USA can use their expertise in operating high-rate computer numeric control machines that can turn out five million motor subcomponents per machine with almost no operator intervention. Alliance Winding Equipment, Odawara Automation, and U.S. magnet startups like Noveon can help close the rest of the loop. The authors believe it is completely feasible to achieve a purely domestic brushless direct-current motor line — cradle to crate — in 12 months at roughly $30 per motor with the right investment and demand signal from the U.S. government.

Batteries are an Achilles’ heel. Hitting a billion drones would require roughly 0.25 terawatt-hours of lithium-ion capacity. The United States is projected to have 0.8 terawatt-hours online next year. An effort to produce a billion drones would be a significant burden on the entire supply chain and substantially reduce our ability to produce electric cars, but these numbers also indicate that it is possible. While not all announced capacity will come online, long-term offtake contracts — not new technological innovations — are what our battery cell makers need.

Printed circuit boards are even tougher. America has fallen from 30 percent of global board production to about 4 percent. Yet firms like TTM Technologies and Summit Interconnect still run sizable lines that could be expanded quickly once they see a multi-year backlog.

For cameras, the type of drones necessary for future wars just need cheap complementary metal-oxide-semiconductor modules. Chinese suppliers will sell that for under $10 apiece in single-unit lots. The United States can meet similar pricing and volumes by tapping the same car camera lines that already equip millions of vehicles every year. Firms such as Magna Electronics Technology in Michigan, which already turn out millions of modules for driver-assistance systems, could divert a portion of that existing capacity to drone production. Such plants could commission additional production lines as demand ramps up.

One area of strength — assembly — plays to America’s industrial muscle. A fully optimized first-person view drone consists of roughly ten parts (including the motor subsystem), a rounding error next to the 30,000 components in a passenger car of which the United States produces roughly 10 million a year, snapping together 300 billion parts in the process. Building one billion drones, therefore, equates to approximately 1 percent of today’s automotive assembly capacity. While automotive assembly lines are just one use case, this example illustrates that at least on this front, the United States can draw on existing domestic know-how in the form of a trained workforce.

To win at scale and unlock enduring economic gains, the United States should pair sheer production capacity with software-driven technological flexibility. A modular code stack turns swarms of small autonomous systems into “hardware-enabled, software-defined” platforms whose function changes with a simple firmware update. One week it’s a loitering munition, the next a crop-sprayer, warehouse picker, or oil rig inspector. The same low-cost microcontrollers and AI libraries that animate aerial drones can also drive factory robots, subsea monitors, and medical assistants. Opening this architecture to both public and private sector developers will spur real-time solutions across industries worth hundreds of billions of dollars and generating more growth and tax revenue on top of the initial defense investment.

Stack all these numbers atop one another, and the picture clarifies. It is possible to hit a million-drone per year pilot line in 12 months, scale to ten million in year three, and if there is a commitment to parallel capacity builds, reach 100 million units by year five. That trajectory does not close the entire gap with China, but it gives Washington a credible deterrent and a bridge to the billion-drone annual target the moment Congress decides the stakes warrant it.

Some may say that matching scale is not the optimal counter to drone swarms. Rather, it’s better to “innovate” towards the problem with new technologies like directed energy, jammers, or exquisite interceptors. All of those have roles, but none work economically. Shooting down a $500 quadcopter with a million-dollar missile is bad arithmetic. Even if we accelerated laser kill chains to one shot per second, defending a six-mile (or approximately 10-kilometer) front against a million-drone swarm would demand roughly 278 one-megawatt emitters — about 800 megawatt after photonic losses — nearly the output of a grid-scale nuclear reactor. The only antidote to a cheap swarm is an equally cheap counter-swarm, one that can attrit the enemy faster than they can replenish.

Industrial policy in the United States works best when it focuses on demand signals, not central planning. The Pentagon memo frames drone supremacy as “a process race as much as a technological race,” aligning perfectly with this call for demand-signal government programming. The Franklin D. Roosevelt administration did not design the B-24 Liberator — it guaranteed Ford that if Willow Run built one an hour, the government would buy every last bomber. The same clarity is needed today.

A drone swarm factory is not sunk cost — it is a moonshot that would lead to commercial robotics boom in agriculture, logistics, infrastructure and much more. Every dollar the United States spends signals a future dual-use market. Just as NASA’s Apollo guidance computers seeded Silicon Valley, a drone swarm program of record would do the same for next-generation power electronics, battery chemistries, advanced plastics, and AI-enabled edge devices.

In his memo, Hegseth stated: “Emergent technologies require new funding lines. To address the urgent need for drones, investment methods outlined in Executive Order 14307 are being investigated.” In line with this, Congress should authorize a five-year, $25–$30 billion procurement — about what the Navy spent on just three Zumwalt-class destroyers — to buy Group 1 drones and their subcomponents at pre-agreed price caps. The Defense Department’s message to industry should be unequivocal: Stand up the line and we will clear the loading dock.

They should also incentivize performers to partner with Tier 1 and Tier 2 auto suppliers for frame molding, precision metal parts, camera modules, and sensors. These firms already understand Six Sigma quality at million-unit volumes — what they lack is a reason to pivot. Guaranteed drone orders supply that reason. Once domestic output clears 100 million units, the United States should allocate export packages to European and Indo-Pacific allies willing to mirror our manufacturing lines and processes. Distributed production complicates enemy targeting and reinforces collective deterrence.

 

Martin C. Feldmann is a manufacturing expert and entrepreneur. He is the founder of Manhattan Technologies and previously the founder & CEO of VulcanForms Inc., an advanced manufacturing unicorn company headquartered in Massachusetts.

Gene R. Keselman is a lecturer at MIT School of Management, the executive director of MIT Mission Innovation Experimental, and managing director of MIT’s venture studio, Proto Ventures. He has also spent 25 years as an officer in the U.S. Air Force.

Become an Insider

Subscribe to Cogs of War for sharp analysis and grounded insights from technologists, builders, and policymakers.