We Are Building the Wrong Factories
Why Silicon Valley's Defense Obsession is Missing the Real Path to American Reindustrialization
Introduction: The Illusion of a Defense Industrial Base
As China rolls tanks, missiles, and fighter jets through Tiananmen Square last week to mark the 80th anniversary of the end of WWII, the United States pivots for the first time in modern history to a military strategy that places defending the homeland above projecting power abroad. The shift is significant, but the real wake-up call is this: the U.S. military is Made in China.
From electronics and propulsion to the parts inside Tomahawks and JASSMs, Chinese firms are embedded across every layer of U.S. weapons production.
In recent years this dependency has been at the center of debate, with policymakers and experts calling to strengthen the U.S. military industrial base. Focusing on industries tied to national security is understandable, but doing so in isolation misunderstands the true source of industrial strength and resilience.
There is no such thing as a defense industrial base separate from the rest of manufacturing. There is only the industrial base. Defense production rests on the same foundations that support cars, household appliances, and consumer electronics.
How different are the processors in our phones from the processors in a missile? What separates an electric sedan from an electric Humvee? Tooling, metallurgy, sensors, battery cells, and semiconductor wafers form the backbone of all production. When those foundations are strong and scaled, they diffuse across every industry, reduce costs of goods sold, and transform Silicon Valley’s breakthroughs into the homes of millions.
This is not a treatise against defense spending, far from it. I’m instead pointing out the artificial limitations this “trickle-down innovation” theory imposes on American dynamism and industrial strategy.
What is needed instead is to bring the shared foundations of production back onshore, where they can create the diffusion and depth that make scale possible. That is the most viable way to rebuild a resilient industrial base, and by extension, the right way to secure defense capacity.
To understand the weakness in practice, look at a basic military necessity: the 155 mm artillery shell.
How a 155 mm Artillery Shell Exposes the System
The surge in demand from the Ukraine-Russia war exposed how brittle U.S. munitions capacity is. We are not even talking about exotic missiles, but simple steel-body shells. Congress’s own analysts flag “production issues for less sophisticated munitions (e.g., 155 mm artillery rounds),” putting 155 mm in the same breath as precision weapons because the industrial base can’t keep up with the surging demand1.
A 155 mm shell consists of four major components: the steel casing, the explosive fill, the fuze, and the propellant. The process of making a shell shows how every part of the supply chain must work in sync for production to reach the quantity needed.
First, the Scranton Army Ammunition Plant in Pennsylvania takes raw steel billets, often shipped in from mills in Ohio, and then forges, machines, and finishes them into painted shell bodies. Propellant charges come from other Army facilities in Virginia or Tennessee2. Final assembly then happens in Iowa, where the semi-finished bodies are loaded with explosives and fitted with fuzes at the Iowa Army Ammunition Plant.
The critical bottleneck is the casing. Nearly all U.S. shell bodies are forged at the Scranton Army Ammunition Plant. That plant’s maximum output was 24,000 units per month3, while Ukrainian battlefield demand has reached as high as 600,000 shells per month. There is no alternative source of casing production at scale. Once Scranton hits its ceiling, the entire supply chain stalls.
To address the shortfall, the Department of Defense opened a new production facility in Mesquite, Texas. The plant was intended to scale casing production quickly, but it has already run into problems. The Army cited delays in sourcing production equipment, reliance on international suppliers, and difficulties qualifying parts. The Army originally aimed to reach 100,000 shells per month by October 2026. That target has now been delayed to mid-2027.
To this end, for all its budget, the United States’ ability to wage war (and supply allies) with 155mm artillery shells is capped at 40,000 rounds per month.
The Structural Weakness in U.S. Manufacturing
The 155 mm shell is not an isolated case. It reflects a broader pattern across American manufacturing. Yes, the United States still manufactures, but the way we manufacture has fundamentally changed.
The 155 mm shell should have been the easiest thing to mass-produce. It is the very definition of a low-tech, high-volume item. Yet the U.S. cannot produce at scale because the industrial base has become a high-mix, low-volume (HMLV) system optimized for precision and specialization, not throughput. Plants like Scranton and Iowa operate as niche facilities dedicated only to military munitions, producing limited batches without the steady scale that comes from serving broader consumer markets. For all our nostalgia of WWII-era defense manufacturing prowess, we tend to forget those volumes were sustained by running consumer goods alongside specialized military parts, not by serving government buyers alone.
This shift has consequences. When you optimize for specialization over volume, you lose the benefits of repetition, tooling amortization, and production leveling, which is the ability to keep output steady and predictable. Capacity becomes fragile. Every supplier down the value chain must absorb fixed overhead without any assurance of future orders.
Over time, this HMLV model has hollowed out the sub-tier supplier base. The real fragility in the U.S. manufacturing system sits below the top layer, namely tier 2 & 3 suppliers. In the case of the 155 mm shell, this means the factories that provide the fuze components and the shell bodies.
Between 2002 and 2018, the United States lost 4,874 machine shops, a decline of 20%. Over the same period, industrial mold companies shrank 42.7%, with nearly 1,241 firms disappearing, and nearly 6,000 workers lost. Going further back, the tool-and-die maker workforce collapsed from 162,000 in 1998 to 89,600 in 2010, a nearly 45% decline4.
This is the downstream effect of a production strategy that favors high specialization over scale. When the supplier base thins, you cannot surge. The worst mistake in supply chain planning is to build capacity only in response to crises. This is what has happened in the case of 155 mm shells.
The Arsenal of Democracy, Then
This is not how America used to build. The clearest demonstration of the right sequencing came during World War II. The United States was able to turn Detroit into the “Arsenal of Democracy” because a vast commercial base already existed. Automotive factories employed hundreds of thousands of skilled workers, supplier networks stretched across the Midwest, and production lines were already tuned for scale. When Ford’s Willow Run plant shifted from cars to B-24 bombers, it was able to roll one off the line every hour.
This sequencing has guided the industrial strategies of nations like Japan and Korea.
What Japan and Korea Achieved Through Commercial-First Strategy
Japan and Korea both began with mass-produced consumer goods such as household appliances, televisions, and VCRs in the 70s and 90s, and later personal computers in the 2000s. These markets created enormous demand for DRAM chips and forced companies to compete in an unforgiving environment. Margins were thin, the competition was global, and survival depended on driving costs down and yields up. In this setting, firms learned to perfect process control, quality discipline, and manufacturing efficiency.
Scale did more than create efficiency. It also generated innovation. In the 1980s, Japanese firms like Nikon and Canon pioneered lithography steppers and advanced wafer polishing to keep pace with higher yields5. Korean firms later introduced breakthroughs in high-density DRAM design and packaging, such as Samsung’s stacked capacitor DRAM cell, which became the industry standard6. The pressure to serve massive consumer demand also drove advances in production automation, statistical process control, and just-in-time supply chains that spilled over into other sectors.
As the competition intensified, firms in both countries moved up the supply chain. The same skills that allowed them to win in commodity DRAM translated directly into producing more advanced chips at smaller nodes with greater precision. The experience of producing for mass consumer markets became the foundation for moving into advanced semiconductors.
Korean firms like Samsung and SK Hynix now dominate the memory market, holding a majority share of global DRAM and NAND production, and they remain at the forefront of memory innovation. SK Hynix, for example, is a leader in High Bandwidth Memory (HBM), a new form of 3D-stacked DRAM considered a key bottleneck for modern AI and high-performance computing workloads. Japan remains indispensable through its dominance in semiconductor materials, specialty chemicals, and equipment. Two Japanese companies, Shin-Etsu and SUMCO, collectively hold a majority of the global market for silicon wafers7.
China’s military-civil fusion (MCF) strategy follows the same sequence. As I have written in my previous post , the MCF strategy anchors defense capability on top of a broad commercial base rather than attempting to build defense capacity first.
Inside China’s Hybrid Military-Industrial Model
The United States built the world’s first fifth-generation stealth fighter. The F-22 Raptor entered service in 2005 after more than a decade of development and was widely considered unmatched in air superiority. But production was abruptly halted in 2011, with fewer than 200 jets delivered. High costs, shifting priorities, and a lack of export approval …
The iPhone Thought Experiment
These historical lessons raise a provocative question: what would happen if the U.S. has rebuilt its own commercial foundation today? To explore that, let’s run a thought experiment. Imagine if we had not outsourced our process knowledge for consumer microelectronics, and manufactured the iPhone end-to-end in the U.S. today (“Designed in California, Assembled in Ohio”), how would it fundamentally change the cost structure and availability of key defense hardware?
Optics. Take the iPhone’s camera module. Every phone contains multiple high-precision lenses and CMOS sensors that require micron-level alignment and ultra-cleanroom assembly. The same production lines and optical calibration systems are used to manufacture surveillance-grade drone optics, industrial automation cameras, or even night vision modules. Sony’s CMOS fabs in Japan already anchor the global supply of sensors for both consumer electronics and specialized imaging markets8.
Sensors & telemetry. The inertial measurement unit (IMU) in each iPhone, used for motion tracking, is a low-cost MEMS sensor built at nanometer tolerances. The fabrication techniques are directly applicable to smart munitions, guidance systems, and aerospace navigation. But MEMS foundries in the U.S. are sparse, often tied to defense-only programs, and lack the yield discipline that comes from commercial volume9. Apple ships hundreds of millions of IMUs each year. That production scale drives process refinement unavailable to defense-only fabs.
Even the vibration motor and actuators, though trivial in consumer use, that capability directly maps onto precision motor assembly used in targeting systems, robotic arms, and miniature propulsion mechanisms.
Battery cells. The li-on battery module and power management ICs inside each iPhone reflect another lost opportunity. Ruggedized versions of these same components power small UAVs, ground systems, and portable radar kits. Commercial production has driven major advances in energy density, thermal stability, and packaging – all of which the Pentagon now seeks to replicate through standalone efforts.
Each of these components share vendors, tooling, and forming, machining processes. When Apple builds 200 million phones a year, it sustains hundreds of tier 2 and 3 suppliers who refine and perfect specialized capabilities at scale that can be spilled over into strategic sectors. If those capabilities are localized, the DoD and defense tech startups could tap into them at marginal cost, lowering the bar for new entrants to build more innovative products at scale.
The point is not to manufacture consumer phones for their own sake. It is to restore the production architecture that makes advanced components affordable, repeatable, and easily transitionable.
The Vertical Integration Trap
Now let’s return to reality. The firms which once represented our OEM, Tier 2, and Tier 3 suppliers in the 1990s and 2000s suffered once production for consistent consumer demand was offshored. Many were unable to shift to irregular defense contracts and either closed or were acquired by larger entities. That consolidation led to the prime contractors falling from 51 in the 1990s to just five today, known as the “Big Five”10.
The result is an even wider “Valley of Death”, the gap between the end of an Small Business Innovation Research (SBIR) Phase II prototype award and the successful transition to Phase III commercialization, where the absence of a robust supplier base makes scaling into production increasingly difficult11. Faced with that gap, many defense tech startups see little choice but to adopt a vertical integration strategy.
Here lies the irony. What may begin as a survival strategy turns into the same structural trap that hollowed out the system in the first place. Popular new entrants - the “neoprimes” like Anduril, Shield AI, Saronic, and Relativity Space, among others - are following the same path as the legacy primes they seek to disrupt: building the entire supply chain systems in-house. While this accelerates early delivery, it starves the very tier 2 and tier 3 supplier base that forms the backbone of resilient industrial capacity.
By replicating the prime contractor model rather than diffusing capabilities across a wider network, the result is faster initial scaling but a weaker ecosystem overall, which further perpetuates the problem.
It is worth pausing to ask whether the up-and-coming defense tech companies are on the path to becoming the new primes in everything but name. And with so much capital and talent now pouring into the defense tech boom, we should also ask whether these resources are being allocated in the most efficient way to strengthen the broader industrial base.
A Path Forward
This leaves us at a crossroads. If today’s “neoprimes” repeat the mistakes of yesterday’s primes, we will end up with more logos but the same brittle structure. The alternative is to chart a different path forward.
Imagine a world where the U.S. has already rebuilt that broad industrial base. Defense startups would no longer need to vertically integrate or reinvent supply chains from scratch, because they could tap into a diverse, deep pool of domestic suppliers. That environment would accelerate defense innovation by lowering barriers to entry and enabling many more startups to compete and thrive.
The U.S. is the best in the world at pushing the frontier, at going from 0 to 1. The problem is that we are terrible at scaling from 1 to 100 - or at least, we have forgotten how to. Countries like China, Japan, and Korea know how to turn breakthroughs into mass production, to make innovation repeatable, affordable, and widely diffused.
This doesn't mean abandoning current defense investments. It means asking how we can be better at efficiently allocating resources, capital, and talent to build the broader industrial base first. In my view, a more productive path forward begins going back to first principles:
What are the core bottlenecks that make our current manufacturing hard to scale? What foundational capabilities do we need to bring back onshore before anything else can grow?
Those questions lead me to a set of frameworks I’ll be unpacking in future essays:
Mapping the manufacturing value chain and identifying bottlenecks at every stage from design, tooling, and prototyping through testing, mass production, and quality assurance reveals where time and resources are most concentrated today and which steps have become the critical choke points holding back scale.
Crossing those bottlenecks with market breadth, prioritizing structural components and digital infrastructure that cut across multiple industries. Tooling, castings, precision machining form the physical backbone of everything from automotives to household appliances to aerospace. Chips, sensors, and connectivity modules form the electronics backbone of consumer electronics, communications, industrial automation, and AI. Batteries and energy storage systems underpin EV cars, household electronics, defense systems, and the energy grid.
Examining why software progress in manufacturing has fallen behind, with commercial CAD remaining proprietary and monopolized. This closed model drives up licensing costs, restricts interoperability between tools, slows automation, and blocks smaller firms from innovating. Is a different path possible if the government convenes experts across the field to develop a robust open-source kernel that could lower costs, improve compatibility, and accelerate AI application across the entire industrial base?
Reassessing workflow documentation, where weak standards have slowed automation and interoperability.
Understanding America’s low adoption rate of industrial robotic arms (and no, I’m not talking about the humanoid form!), and how do we incentivize small and mid-sized business, which make up the majority of the U.S. manufacturing industry, through new business models such as a Robot-as-a-Service business model to lower the barriers to adoption.
How can we better connect startups working on the most cutting edge problems with the industry knowledge and testbeds they need to scale? Unlike their peers in tech and finance, most industrial Fortune 500 companies lack corporate venture arms and choose instead to acquire at later stages. If these firms invested earlier, they could provide capital, data, and integration opportunities that make startups’ solutions more relevant to industry needs, while also creating credible exit paths that strengthen the venture pipeline. In doing so, they would build their own capacity to absorb new technologies and accelerate innovation, while sending demand signals that draw more entrepreneurs and more investors into the space early on. The result is a flywheel effect for industrial innovation that compounds over time.
Exploring startup opportunities and investments along these lines will likely yield far greater impact than building yet another drone company for the DoD. By channeling entrepreneurial energy into the critical but overlooked parts of the manufacturing ecosystem, we can create the scale and resilience that defense innovation ultimately depends on.
Acknowledgements: Big thank you to Ryan Cunningham, Kelly Vedi, Mike Riggs, Adam,Benedict Springbett for all your feedback and discussions.
Congressional Research Service, The U.S. Defense Industrial Base: Background and Issues for Congress, Updated September 23, 2024.
Heid, Marc. "Defense Production for Ukraine: Background and Issues for Congress." Congressional Research Service, R48182, August 28, 2024.
Pennsylvania ammo plant boosts production of key artillery shell in Ukraine’s fight against Russia. Associated Press. August 28, 2024.
BLS data compiled by Michael Collins, Is U.S. Manufacturing Losing Its Toolbox?, IndustryWeek, August 16, 2019.
Atsuhiko Kato, Chronology of Lithography: Milestones, Version 0.9 (LithoGuru, May 2007), https://www.lithoguru.com/scientist/litho_history/Kato_Litho_History.pdf
Dae‑Je Chin and Tae‑Young Chung, Stack Capacitor DRAM Cell Having Increased Capacitor Area, U.S. Patent 5,378,908 A (filed December 13, 1988; granted January 3, 1995), assigned to Samsung Electronics Co. Ltd.
Sujai Shivakumar, Japan Seeks to Revitalize Its Semiconductor Industry (Washington, DC: Center for Strategic and International Studies, August 1 2023).
Sony Semiconductor Solutions Group. “Group Information.” Accessed September 4, 2025. https://www.sony-semicon.com/en/company/group/index.html.
Rethinking the MEMS supply chain and foundry business model. CMM Magazine, Doug Sparks, October 27, 2022.
U.S. Department of Defense, State of Competition within the Defense Industrial Base, February 2022.
U.S. Defense Innovation Board, Terraforming the Valley of Death: Making the Defense Market Navigable for Startups, July 17, 2023.
Excellent insight on what’s needed for broad-based success! America needs to build and foster an ecosystem where reindustrialization can thrive. Union Technologies (https://union.tech) appears to be tackling the munitions challenge, but we need more companies addressing these fundamental issues.
A lot of complaints in the cost of US military systems. One reason they’re so expensive is because we build them end-to-end with US labor. Components might be sourced from overseas but even that requires approvals. Why is an iPhone cheap to build and a US military system is not? Maybe it is the difference in who’s building them.