Out of Ammo

Overview

The subject

Between 28 February and late July 2026, the United States fought a war against Iran in two phases. The fighting drew down four categories of munitions the Pentagon had spent the preceding decade describing as essential to any fight in the Western Pacific: Tomahawk cruise missiles, THAAD (Terminal High Altitude Area Defense) interceptors, Patriot PAC-3 Missile Segment Enhancement (MSE) interceptors, and the Precision Strike Missile (PrSM), an Army weapon designed for the anti-access/area-denial problem posed by China. PrSM made its combat debut on 1 March against Iranian naval and missile targets.

The Center for Strategic and International Studies (CSIS) estimated that the first 39-day phase alone involved strikes on more than 13,000 targets. That pace of expenditure helps explain why, by July 2026, the same analysts put the Patriot interceptor inventory below 1,000 and the THAAD stockpile at roughly 250, down from a prewar figure of 452. Those are analyst estimates built from public records, not Pentagon disclosures, and CSIS has not published a margin of error for them. Later chapters treat the depletion direction as established and the exact depth as unresolved.

The monthly replenishment rates make the scale of the problem plain. The Pentagon was receiving roughly 15 new Tomahawks and 20 new Patriot missiles per month as of August 2026, rates far too small to close the gap CSIS describes over any short horizon. CSIS projects THAAD will not be restored to prewar levels until the end of calendar year 2029, and Tomahawk not until late 2030 or early 2031 (a timeline developed further ahead). Meanwhile, a $14 billion Taiwan package containing PAC-3 interceptors and NASAMS sits paused, and Taiwan's broader Foreign Military Sales (FMS) backlog has held at $29.72 billion for months. American reconstitution and Taiwan's own backlog now draw on the same production lines.

What this study covers, and what it does not

The scope here is the inventory pipeline for four munitions: Tomahawk, THAAD, PAC-3 MSE, and PrSM. It covers what was expended in the Iran conflict and when. It covers the disclosed and estimated stockpile levels. It covers current production and delivery rates as distinct from announced capacity targets, who is competing for the output, and how far the resulting inventories fall short of what a Taiwan Strait contingency is assessed to require.

Four things are deliberately outside the scope. First, the conduct and merits of the Iran war itself: this study treats the campaign as a demand event on the munitions base and takes no position on whether it was wise. Second, classified inventory totals, which are not in the public record. Where the only available figures are analyst estimates, they are labeled as such and never upgraded. Third, operational detail of any kind, including force positions, basing, or engagement planning. Fourth, Chinese munitions inventories and production. Nobody consulted for this study has published Chinese-side figures.

How well understood this actually is

The war's chronology is documented in detail, as is the use of PrSM and Tomahawk, the announced production-expansion agreements, the FY2026 National Defense Authorization Act's $6.4 billion in enacted critical munitions procurement, the Section 804 multiyear authority granted for eight of the 13 programs the Pentagon requested, and the Taiwan package pause following the 13–15 May Trump-Xi summit.

What is not established is harder to overstate. Every current stockpile number is estimated, all of it traced to CSIS analysts working from public records rather than a disclosed Pentagon count. Every replenishment date is a projection. Whether the shortfall is severe at all remains contested. Trump and the Pentagon characterized the expenditure as "peanuts," a position reported by the Epoch Times and not independently corroborated elsewhere, and claimed that large stockpiles of unspecified lesser weapons remain (that framing is developed once, in the allocation chapter, and referenced here only as the administration's minimization stance).

Much of the sharpest analysis comes from institutions with standing interests. CSIS's Rebuilding U.S. Missile Inventory work, led by Mark Cancian and Chris Park, anchors most of the numbers in circulation. The Council on Foreign Relations' "Out of Ammo" report recommends five-year demand commitments and front-loaded industrial investment. The Heritage Foundation's wargame-derived claims are cited across congressional testimony and defense-trade press, which is part of why they carry more weight than their sourcing alone would justify. The prime contractors, Lockheed Martin and RTX, issue capacity announcements that double as sales messaging. Project On Government Oversight (POGO) supplies an oversight counterweight on lead times. The positions that follow are weighed with those interests in view.

The questions ahead

How much was actually expended, and when? That requires reconstructing the two combat phases. The first ran from Operation Epic Fury on 28 February through the Pakistan-brokered ceasefire of 8 April. The second began when the June memorandum of understanding collapsed on 7 July, after Iranian strikes on commercial vessels, and closed with the pause on 27 July. Each phase drew from a different starting inventory, and establishing what each phase drew from which stockpile is the first task.

What are current inventories, and how confident can anyone be in the numbers? The Patriot and THAAD figures are CSIS estimates. What supports them, what could contradict them, and why the Pentagon does not publish the totals all matter here.

Does replenishment mean restoring prewar inventory, or reaching an announced production capacity? These are different milestones with different dates. CSIS benchmarks against prewar levels and reaches 2029–2031. Industry and Pentagon announcements benchmark instead against future annual capacity: THAAD rising from 96 to a targeted 400 interceptors a year, PrSM to 550 units. Using the prewar-level benchmark yields a long but tractable recovery ending in 2029–2031. Using the capacity-target benchmark yields no fixed completion date at all, and a larger gap still to close.

How fast can the industrial base actually move? Production-ramp estimates run three to five years for interceptors and four to five for Tomahawk. Those timelines sit against demand from CENTCOM (U.S. Central Command) reconstitution, Ukraine, Taiwan, and Japan, all drawing on the same lines at once (a competition detailed later). POGO puts Tomahawk contract-to-delivery at 47 months, at a current production rate of about 60 missiles a year, and names manufacturing capacity as the constraint. CSIS conditions its three-plus-year Patriot estimate on Congress passing the next fiscal year's Pentagon budget, a condition some treat as near-certain and others as a live risk.

Who gets the output? CENTCOM reconstitution, Ukraine, Taiwan, and Japan are competing claimants. Every party involved, including both prime contractors, agrees current rates cannot satisfy all four at once. The fault lines run between active war and contingency deterrence, between treaty allies and non-treaty partners, and between executive discretion over sequencing and the congressional reporting mandates the Senate Armed Services Committee (SASC) has begun imposing.

How large is the Taiwan gap, and is the "eight days" figure real? The Heritage Foundation has said wargames repeatedly show the U.S. running out of critical munitions eight days into a high-intensity Taiwan conflict. It is the sharpest and most-cited figure in this whole debate. Whether it describes the current post-Iran-War baseline or an earlier, narrower scenario centered on a single weapon type (LRASM (Long Range Anti-Ship Missiles), not munitions broadly) is disputed, and the dispute is not settled here.

Production Lines: What the Four Systems Can Actually Build

The overview quantified what the Iran war took out of American stockpiles. Here, the other side of the ledger: what the industrial base that has to refill those stockpiles can actually produce, at what cost, and where the bottlenecks sit. Production-rate and cost figures across all four systems are reasonably well documented. But every number describing how many missiles currently exist in inventory is a modeled estimate rather than a disclosed count, and outlets citing the same CSIS analysis report different figures for the same measurement.

Tomahawk — a line running far under its own capacity

Tomahawk procurement numbers show a program that has spent years buying well below what its manufacturer, RTX (Raytheon), says it could build. The fiscal 2023 buy was 68 missiles, dropping to 34 in fiscal 2024. Fiscal 2025 actual delivery came to just 20, and fiscal 2026 funded 55. CSIS states the ten-year average procurement rate, spanning fiscal 2015 through fiscal 2026, was 86 missiles a year. The recent annual production rate, CSIS says, is under 200, against a stated RTX capacity goal of more than 1,000 Tomahawks a year — a gap of roughly an order of magnitude between what has actually been bought and what the line says it could build.

A December 2025 contract modification reportedly raised the nominal average rate under that specific contract line from about 3.35 to about 7.14 missiles a month, or roughly 40 to 86 missiles a year on that line alone. Even the higher figure sits far short of the four-figure capacity RTX has claimed.

Unit cost figures conflict depending on source and production lot. Wikipedia's infobox lists $2.5 million per missile for fiscal 2026. Industry and press reporting puts recent Block V buys lower, in the $1.8 million to $2.0 million range, with one fiscal 2024 figure at about $1.89 million. The December 2025 contract modification reportedly cut the notional per-unit cost on the additional order to about $1.75 million. A separate program-level Average Procurement Unit Cost figure, drawn from Selected Acquisition Report data, comes to about $1.314 million. That number is priced in base-year 1999 dollars, so it can't be set against the current-dollar figures above without adjusting for inflation, and it uses a different accounting method entirely.

Inventory figures conflict even more sharply. CSIS, per the Stimson Center's Kelly Grieco as cited by CNN, put total U.S. Tomahawk inventory at approximately 3,100 before the drawdown. A separate figure circulating elsewhere puts total missiles procured since the program's start, across all variants and including many now retired, at roughly 9,000. That is a cumulative production total rather than a current stockpile count, and it isn't comparable to the 3,100 figure without knowing how many of those 9,000 remain in service.

The prime contractor is RTX (Raytheon). The named bottleneck, per CSIS, is not a manufacturing ceiling but a purchasing pattern: low peacetime order volumes have kept the Tomahawk line running at low-rate production for years. The gap between what RTX says it can build and what the Pentagon has actually ordered reflects a choice about how much to buy in peacetime.

THAAD — the widest cost and production spread in the grid

Reporting on the THAAD interceptor is the most internally inconsistent of any system covered here. Breaking Defense and Military Watch Magazine put current production at approximately 96 missiles a year, matching the canonical current surge rate. Cronkite News, citing Payne Institute-linked reporting, puts the current rate at "about 600 a year," a figure that conflicts with the 96-a-year number reported elsewhere and even with a separate line in the same Cronkite News article, which states elsewhere that Lockheed Martin "can produce roughly 96 THAAD interceptors per year." That is a flagged internal contradiction in a single outlet's own reporting, carried here as reported rather than resolved.

What isn't contested is the direction of travel. A January 29, 2026 framework agreement and a June 2026 contract worth $35 billion commit Lockheed Martin to quadruple THAAD output, from the current baseline to 400 interceptors a year, over seven years. That is a company and Department of War commitment to a future capacity target, not a confirmed current output figure.

Unit cost also diverges. The Hill, citing Pentagon documents, reported $12.77 million per interceptor for fiscal 2025. Cronkite News and The National Interest separately report $15.5 million, up from $9.5 million in 2021, a cost trajectory that comes from a different source than the Pentagon figure and isn't reconciled with it. Other sources give a planning-estimate range of $12 million to $15 million, bracketing both numbers without confirming either.

Inventory figures show the same dispersion. Lockheed Martin's reported delivery of its 900th THAAD interceptor to all customers, in January 2025, is a cumulative production total, not a claim about current inventory. CSIS estimates the pre-war U.S. stockpile at approximately 452, as of July 2026, reported by both Visual Capitalist and CNN citing CSIS. One additional source, pizzint.watch, carries a THAAD pre-war figure that appears to duplicate a separate Patriot figure reported by the same outlet, a likely sourcing error flagged here because it couldn't be independently resolved, and not used as a THAAD data point. Post-war THAAD estimates range from roughly 234 to 278 depending on the outlet, all tracing back to the same underlying CSIS modeling and roughly consistent with the canonical figure of about 250.

Lockheed Martin is the prime contractor, with final assembly at its Camden, Arkansas facility. Named bottlenecks include radar and launch-support infrastructure losses, along with solid-rocket-motor and propulsion motor-case supply, a constraint CSIS and manufacturer statements describe as shared with the Patriot program, since both systems draw on overlapping upstream supply chains.

PAC-3 MSE — the one system with real production growth already documented

The Patriot Advanced Capability-3 Missile Segment Enhancement (PAC-3 MSE) interceptor is the one system here where an actual production increase, not just an announced target, is already on the record. Lockheed Martin produced more than 500 PAC-3 MSEs in 2024, a stated production record for the company, and reported 620 delivered in 2025. CSIS separately states current production sits "around the baseline rate of 650 interceptors per year." Those figures sit close enough together to describe the same ramp-up.

A January 2026 framework agreement and a July 2026 contract worth $53.86 billion commit Lockheed to triple that capacity, to approximately 2,000 interceptors a year, by the end of 2030.

Unit cost is contested. The most frequently cited figure, appearing in The National Interest, Cronkite News, and other reporting, is approximately $3.9 million per unit. A separate figure of $4.2 million, sourced to the fiscal 2025 Army budget request, appears in one industry-analysis source — a gap of roughly $300,000 per interceptor between the two.

Inventory estimates span the widest range of any system here. CSIS puts the pre-war U.S. Patriot inventory, covering all PAC-3 variants including MSE, at approximately 2,200 to 2,330, with roughly half to two-thirds expended during the war. Current estimates built from that baseline range from "fewer than 1,000" (Military Times, July 2026, consistent with the canonical below-1,000 figure) to "roughly 827" (pizzint.watch) to a range implying 1,060 to 1,430 missiles expended, per Fox News. CSIS's Patriot figure is not MSE-specific: it mixes PAC-3 and PAC-3 MSE counts together, so no clean MSE-only inventory number can be pulled from any of the reporting above.

Lockheed Martin is the prime contractor, with final assembly at its Camden, Arkansas facility. The named bottleneck is propulsion motor-case and solid-rocket-motor supply, the same constraint named for THAAD, a sign of how much the two interceptor programs depend on a shared upstream supply chain.

PrSM — funded quantities rising, but no fielded-inventory figure exists

The Precision Strike Missile (PrSM) is the newest system here and the one with the thinnest inventory data. A March 25, 2026 framework agreement between Lockheed Martin and the Department of War set a target production capacity of 550 missiles a year under a seven-year agreement, described as quadrupling capacity from the prior baseline. A separate outlet cited a lower target of 400 a year, tied to an earlier round of roughly $5 billion in contract awards. The two figures aren't reconciled, and it's unclear whether they describe the same ramp at different stages or genuinely different targets.

Unit cost estimates diverge by nearly a factor of two. Wikipedia's infobox states a unit cost of "less than $3.5 million" for Increment One, a figure that has not been independently corroborated. A separate industry estimate places it at approximately $2 million.

No fielded-inventory or program-of-record current-stockpile figure for PrSM appears in the public record. What exists instead is a sequence of funded-quantity figures from Army budget documents. Fiscal 2024 through fiscal 2026 funded 98, 230, and 152 missiles respectively. The fiscal 2027 budget request seeks approximately 1,134 units for about $1.9 billion, or roughly $1.7 million a unit, nearly four times the $546 million allocated in fiscal 2026. Whether a line that has never funded more than 230 units in a year can absorb a jump to 1,134 is not something the funding request itself answers. The Army's stated Acquisition Objective for PrSM Increment 1 alone is 3,986 missiles, a procurement target rather than a count of missiles built or fielded to date.

Lockheed Martin Missiles and Fire Control is the prime contractor. Production runs through a Precision Fires facility in Camden, Arkansas, the same city that hosts PAC-3 MSE final assembly, covering more than 115,000 square feet of dedicated PrSM production space. Named bottlenecks include tooling and testing-equipment lead time and component-level supply chains, distinct from the motor-case constraints shared by THAAD and PAC-3 MSE.

What the pattern across all four systems shows

Production-rate, unit-cost, and site data hold up reasonably well across Tomahawk, THAAD, PAC-3 MSE, and PrSM: anchored in corporate statements, Department of Defense (DoD) contract notices, and Selected Acquisition Report data. Even where individual figures conflict, the sourcing for each is traceable and named. But every current-inventory or program-of-record quantity figure for three of the four systems traces back to CSIS Missile Defense Project modeling, built from budget-document delivery schedules and reported combat expenditure rather than a disclosed Pentagon count. Multiple outlets citing that identical CSIS analysis report different specific numbers for the same measurement, so the state of the record is a range for how many interceptors the United States actually has on hand, not a settled point estimate.

One figure cuts across all four systems in the same direction: the gap between announced future capacity and confirmed current output. THAAD's jump from roughly 96 to a targeted 400 a year sits alongside PAC-3 MSE's tripling to 2,000 by the end of 2030 and PrSM's quadrupling to 550. RTX's stated Tomahawk capacity goal of more than 1,000 a year, against a recent rate under 200, follows the same pattern. Each is a company or Department of War statement about planned capacity, not a confirmed current delivery rate, and how fast any of them actually closes remains untested against the replenishment timelines CSIS has projected elsewhere.