The defense-industrial base — empty magazines, China's grip on the reload, and the defense-tech wave

Web-verified 2026-06-11. Structured + sources: geopolitics-defense-industrial-base.json. Overlay — evidence-graded, excluded from the proofs. Connects to defense_web, macro-critical-minerals-equities, geopolitics-chip-chokepoint-war, geopolitics-contested-resource-states, macro-ai-power-grid-bottleneck.

This is where the abstract "materials chokepoint" becomes concrete national power. The corpus proves the rare-earth/antimony chokepoints; this block connects them to what they actually gate — weapons — and to the capital wave digitizing the fight.

The arsenal is depleted

• Ukraine (since 2022): the US shipped ~⅓ of its Javelins, ~¼ of its Stingers, >2 million 155mm rounds, and thousands of GMLRS; the Pentagon's own assessments describe a "near-term risk" of running out, and a 2025 shipment was briefly halted over stockpile concerns (fact).
• The 2026 Iran war: the US reportedly burned ~45% of its Precision Strike Missiles, ~half its THAAD interceptors, ~half its Patriot PAC-3, ~30% of Tomahawks, and >20% of long-range JASSMs (fact). Two wars emptied the magazine faster than industry can refill it — the physical-production mirror of the AI power bottleneck.

The production base can't reload fast

The flagship metric: 155mm output is ~40,000 rounds/month, with the 100,000/month goal slipping to ~mid-2026 (>1M shells for 2026 total). The Army has put >$4.2B into the organic industrial base (a new Parsons, Kansas M795 plant at ~12,000/mo; Hanwha/UNION adding US capacity). But building shell lines, propellant, and energetics is a multi-year industrial problem, not a budget line — the same shape as transformers and gas turbines (macro-ai-power-grid-bottleneck).

China's materials gate the weapons

The dependency is blunt: "If China said we're not going to give you rare earths, that means no F-35s, no missiles." Every F-35 needs samarium-cobalt magnets; THAAD and PAC-3 interceptors rely on REE inputs; antimony (ammunition/hardening) is ~90% China — the Aug/Dec-2024 ban spiked prices +134% (fact). The Pentagon drew a 2027 line (no Chinese-origin materials in US defense gear after 2026 — widely judged extraordinarily hard for heavy REE) and is racing to secure supply (the MP/Perpetua/UAMY/Ucore response, macro-critical-minerals-equities).

The vicious loop: the wars that deplete the arsenal — including the Iran war (geopolitics-contested-resource-states) — require materials to rebuild that China controls and has shown it will weaponize (geopolitics-chip-chokepoint-war). As Fortune put it: "America shot its arsenal empty in two wars and now needs Beijing's permission to reload."

The defense-tech capital wave (the AI core's adjacency)

• Concentration: Anduril + Palantir + SpaceX reportedly account for ~88% of significant defense-tech investment and contracts — the same concentration as the AI core, with overlapping principals (Thiel/Founders Fund, Musk).
• Anduril: >$2B 2025 revenue (+110% YoY), a $20B Army counter-drone contract, a 5M-sq-ft Ohio drone factory (Arsenal-1), and a $5B Series H at a $61B valuation.
• The money: a16z's "American Dynamism" practice (~$1.2B Fund 2, part of a ~$15B a16z raise) pours VC into defense/aerospace/energy/supply-chains — the same marquee AI VCs now funding defense the way they fund the AI core.
• Narrative-beta split: as with the AI/quantum/critical-minerals cohorts, separate the real, state-anchored builders from the froth. Reality checks: Anduril's Altius drones crashed in USAF tests; CNBC noted the Iran war was "defense tech's chance to shine" but few systems were ready; and Anduril's own Trae Stephens warns defense tech is headed for a shakeout (fact).

Why it matters

The physical-chokepoint thesis — that the AI build can't buy past Taiwan / ASML / HBM / HALEU / power — extends one-to-one to hard power: the defense build can't buy past China's materials + a slow industrial base. The US is fighting wars that empty a magazine its own base can't refill on the timeline, using materials an adversary controls — while the same AI-core capital funds a concentrated, partly-frothy defense-tech wave to digitize it. Materials are not an AI input or a defense input; they are the shared binding constraint of both, and one country holds them.

← Research index · structured data: geopolitics-defense-industrial-base.json · geopolitics-defense-industrial-base.md