Mobilize

The capacity to act

Reindustrialization is the ability to build, repair, replenish, and adapt at scale, on our own timeline. It has nothing to do with nostalgia for old factories. It means suppliers who can surge, workforces trained before demand spikes, production lines that can be retooled when a supplier disappears, and energy to run the new plants. A ribbon-cutting is a milestone. Durable output, and the ability to change course under pressure, is the achievement.

Supply-chain security sits at the center of it. Refining of key energy minerals is concentrated in a few countries (see Intelligence Runs on Atoms), and a nation that can’t process its own critical inputs, or source them from trusted partners, carries a dependency that can be turned into leverage against it at any moment.

No single firm can fix that alone, which is why federal mobilization matters. For the first time in a long while, Washington is reaching for tools sized to the problem. In July 2025 the Pentagon bought a $400 million equity position in the rare-earth producer MP Materials, set a ten-year price floor for its magnet feedstock, and guaranteed that the full output of a new U.S. magnet plant will have buyers for a decade.¹ In February 2026, the Export-Import Bank approved a loan of up to $10 billion to launch Project Vault, a U.S. Strategic Critical Minerals Reserve built to carry domestic manufacturers through supply shocks.²

You can argue about the instruments; I’d rather argue about results. Equity stakes, price floors, guaranteed offtake, and stockpiles are demand signals. They make private capital willing to build the refinery the market alone won’t justify. The test is how much real capacity exists afterward, not how much was announced.

Defense is the forcing function

Defense makes all of this concrete, because defense can’t fake it. A weapon you can’t manufacture, sustain, or replace is a museum piece.

The National Defense Industrial Strategy, released in January 2024, put the priorities on paper: resilient supply chains, workforce readiness, flexible acquisition.³ In November 2025 the department pushed further, redesignating its Defense Acquisition System as the “Warfighting Acquisition System” and directing that program offices be reorganized under portfolio acquisition executives accountable for speed and outcomes.⁴ Renaming things is cheap and reorganizations are easy to announce, so the proof will be in delivery times. The direction is right. Military power depends on an industrial system that can produce, sustain, and replace what it fields, and that system has to exist before the crisis, not after.

Two mechanisms inside the Department of War, the secondary title the department has used since September 2025,⁵ deserve close attention.

The first is the CTO. The Under Secretary for Research and Engineering serves as the department’s chief technology officer, overseeing technology development and transition, prototyping, experimentation, and testing, along with DARPA itself.⁶ That office sits at the most important junction in defense innovation, the “valley of death” between a promising prototype and a fielded capability. Judge it by one metric: how fast good technology reaches the people who need it.

The second is DARPA’s competition model: define a hard problem, set measurable conditions for winning, open the field to anyone, and let results speak. In the August 2025 AI Cyber Challenge final, competing AI systems found 54 of the synthetic vulnerabilities planted across 63 challenges and patched 43 of them. Along the way, they turned up 18 real vulnerabilities in open-source software that nobody had planted.⁷ It’s a bounded result under test conditions, and proof the model works.

That model should spread far beyond defense, because it’s how you bring in the small, clean-slate builders from Don’t Retrofit. Define the problem, let anyone prove themselves, measure the result, and then do the hard part: carry what works into sustained production.

This matters to me personally. This country gave me every opportunity I’ve had, and I believe people who can build have an obligation to put some of that ability toward national capacity and defense. Not because conflict is inevitable, but because a nation that can credibly build is far less likely to be tested.

The government is not ready

Federal and local governments never caught up with the last technology cycle, the web. They are not ready for the AI wave.

The evidence is specific. In July 2025, the Government Accountability Office examined eleven of the most critical federal legacy systems, some of them decades old.

In April 2026, GAO reviewed thirteen AI acquisitions at four agencies and found officials struggling to reach the technical experts needed to evaluate proposals, and struggling to understand what AI actually costs.⁹

Washington knows it. The Office of Management and Budget’s April 2025 AI memo ordered agencies to cut bureaucratic barriers to adoption and to name a Chief AI Officer.¹⁰ That’s the right instinct, but a title isn’t a capability, and a memo can’t rewrite a decades-old system. Local governments face their own versions: permitting offices still pushing paper, procurement rules written for a slower world, too few technical staff to judge what they’re being sold.

Slowness is only part of the danger. AI lets companies and individuals generate proposals, applications, and filings at a volume no existing process was built to absorb. A government that can’t keep up becomes either a bottleneck that freezes everything or a rubber stamp that approves what it doesn’t understand, and neither is acceptable.

Some of the slowness is deliberate. Due process, public accountability, environmental review, and the duty to keep essential services running all slow things down for real reasons. The goal isn’t to strip out those protections. It is to deliver them faster and better: good decisions made promptly, with clear records, real technical expertise, and a way for people to challenge mistakes. A ready government has:

• Permitting that can tell a manageable risk from an unresolved one, and decide quickly.
• Procurement that can evaluate unfamiliar suppliers, run fast trials, and scale what works.
• Technical talent inside the institution, with the authority to use it.
• Clear responsibility when software takes part in decisions, including how errors get fixed.
• Modern systems, because the web-era work has to be finished before the AI-era work has something to stand on.

None of that happens quickly. Hiring, retraining, rewriting rules, and earning public trust take years, which is why the work has to start now, before automated activity overwhelms the processes we have.

National strength in this era comes down to industrial capacity, secure supply chains, a defense base that can produce and sustain, and institutions fast enough to govern it all. Power on that scale raises a question of its own: where to point it.