Brief
CSIS's Taiwan invasion wargame series, run under the title The First Battle of the Next War, has become one of the most frequently cited public sources for claims that the United States would exhaust key precision munitions within days of a Taiwan Strait conflict. The claim traces to a specific, bounded exercise design: CSIS ran the wargame 24 times, with only a subset run under the project's designated base-case assumptions and the remainder run as excursions that varied specific parameters to test sensitivity. According to reporting on the game's structure, the base case itself was run three times, while 21 iterations were run with different assumptions, and a further single iteration modeled Taiwan fighting without any outside assistance. That structure — a three-run base case flanked by dozens of excursions probing different assumptions — is the mechanical core of everything downstream analysts, congressional staff, and journalists have said about how fast the United States would run through LRASM, JASSM-ER, and Tomahawk stocks.
The live methodological question is whether that structure can bear the analytic weight now being placed on it. A separate, related House Select Committee on the CCP simulation, facilitated by CSIS, produced the specific finding that the U.S. and Taiwan would run low on critical munitions within days of a Taiwan Strait conflict, naming LRASM, Taiwan's anti-ship cruise missiles, and JASSM-ER as most affected — and separately estimated a roughly two-year replacement timeline for Tomahawk Block V, LRASM, and JASSM at then-current production capacity. These figures now circulate as near-canonical shorthand for a documented U.S. sustainment problem, cited in congressional testimony, budget debates, and open-source defense commentary. The FY2026 DoD budget submission requested $35.7 billion in procurement and RDT&E funding for missile and munitions programs government-wide, with $30.2 billion for a defined set of selected programs — figures that policymakers explicitly connect to closing the gap the wargames identified.
What the wargame outputs cannot do, on their own methodological terms, is substitute for classified inventory accounting. CSIS's own published methodology states plainly that much of the relevant material on how a Taiwan conflict would unfold is classified and unavailable to the public, and that unclassified material is either incomplete or too narrow for policymaking — an admission built into the project's own rationale for running an unclassified wargame in the first place. That means the expenditure rates the games generate are calibrated against assumptions about adversary behavior, U.S. force employment, and munitions performance that are themselves modeled, not measured against the actual classified stockpile ledgers DoD and the services hold. Separately, GAO reviews of DoD's conventional ammunition accountability and critical-materials stockpiling have repeatedly found gaps in inventory data completeness, accuracy, and reporting to Congress — a distinct but reinforcing data-quality problem layered on top of the wargame's own modeling assumptions.
The dispute, in short, is not about whether the wargame findings are worthless — CSIS describes its combat-resolution rules as analytically derived rather than judgment-based, and the same rule set was applied consistently across iterations — but about how much epistemic weight a bounded set of base-case-plus-excursion runs, built on public-domain assumptions, can carry when the ground truth it is meant to approximate (actual current stockpile levels, replenishment rates, and classified production capacity) sits behind a classification barrier the wargame cannot see through. That gap between simulated fragility and verified fragility is precisely what force planners, appropriators, and industrial-base investors need resolved before treating the games as a load-bearing input to procurement decisions rather than a scoping exercise that identifies where classified verification is now overdue.
What It Turns On (4)
Does directional convergence across a small number of varied iterations substitute for statistical confidence, or does it merely reflect that the underlying combat-resolution rules were held constant?
If convergence across excursions is driven mainly by the fixed analytic rule set rather than by independent draws representing real-world uncertainty, the apparent robustness of the LRASM/JASSM-ER shortfall finding may be an artifact of consistent rules rather than genuine convergent evidence.
Can any unclassified model be validated against a stockpile ground truth that is itself imperfectly tracked even within DoD's own classified and administrative systems?
If DoD's internal accounting has documented completeness and accuracy gaps, then the debate over wargame validity may be moot in the sense that no outside benchmark — classified or not — currently exists against which to definitively confirm or refute the wargame's expenditure rates.
Is the appropriate use of the wargame findings a 'hard verdict' on sustainment shortfalls, or a lower-bar 'scoping signal' that flags where classified verification is warranted?
The two framings imply very different confidence levels and different downstream uses — one supports treating the games as decisive procurement evidence, the other treats them only as a prioritization tool for where classified stockpile audits should focus next.
Does the two-year production replacement estimate for Tomahawk Block V, LRASM, and JASSM rest on the same wargame-derived expenditure assumptions it is meant to validate, creating circularity?
If the replacement-timeline estimate and the depletion-rate estimate both trace back to the same simulation inputs rather than independent classified production and stockpile data, the two figures may reinforce each other analytically without adding independent corroborating evidence.
Facts & Figures (11)
The claims behind this analysis, each with its verification status — including what is contested, unverified, or could not be established.
The base-case-to-excursion structure is a recognized sensitivity-analysis design, not an ad hoc small sample.
Reporting on the CSIS game's design states the base case was run three times while 21 additional iterations varied assumptions, with combat resolved by analytically based rules applied consistently across the first and last iteration alike.
✓ DOCUMENTEDcase for
The consistency of the finding across a wide range of excursion assumptions is itself evidence of robustness, not fragility of the sample.
The separate House Select Committee on the CCP CSIS-facilitated simulation identified LRASM, Taiwan's anti-ship cruise missiles, and JASSM-ER as the most-impacted munitions, with the U.S. and Taiwan running low within days — a finding that has held up across the related public reporting on both the original 24-iteration wargame and the Select Committee exercise.
✓ DOCUMENTEDcase for
Independent corroboration from congressional and budgetary responses suggests the qualitative direction of the finding is sound even if the precise numbers are contestable.
The FY2026 DoD budget submission requested $35.7 billion in procurement and RDT&E funding for missile and munitions programs government-wide, with $30.2 billion directed to a defined set of selected missile and munitions programs, and the Senate Appropriations Committee conference summary reflected a further munitions production and R&D increase.
✓ DOCUMENTEDcase for
Public, unclassified wargaming exists precisely to fill an analytic gap that classified material cannot serve, because classified analysis cannot inform open policy debate or industrial-base investment decisions.
CSIS's published methodology states that much of the material on how a Taiwan conflict would play out is classified and unavailable to the public, and that unclassified material is either incomplete or too narrow for policymaking — the explicit rationale for building the wargame in the first place.
✓ DOCUMENTEDcase for
The analytically-based combat resolution rules reduce the risk that expenditure rates reflect subjective judgment calls by individual game controllers.
CSIS states that results were determined by analytically based rules instead of personal judgment, and that the same set of rules applied to the first iteration and the last, ensuring consistency.
✓ DOCUMENTEDcase for
A three-run base case is too small a sample to support a hard quantitative verdict on real-world sustainment margins.
CSIS ran three separate iterations of the base case, and in two of those three, China was decisively defeated with PLA forces ashore out of supplies — a result set too small to establish confidence intervals on expenditure rates.
✓ DOCUMENTEDcase against
Excursion runs test sensitivity to assumptions, not ground truth about actual U.S. inventories — conflating the two overstates what the games can tell force planners.
CSIS explored 19 iterations of the game with different assumptions more favorable to China, in which non-LRASM variants of the AGM-158 family could not target moving naval vessels — a scenario-design choice, not a measured capability constraint drawn from classified test data.
✓ DOCUMENTEDcase against
The wargame's own designers acknowledge the classified-material blind spot that limits its ability to validate real sustainment claims.
CSIS states there is little publicly available material on how such a conflict might play out, that much is classified and unavailable to the public, and that unclassified material is either incomplete or too narrow for policymaking.
✓ DOCUMENTEDcase against
Independent GAO findings on DoD's own ammunition and stockpile accounting show that even DoD's internal, non-classified administrative data has documented completeness and accuracy problems — undermining confidence that any outside model, wargame or otherwise, is being benchmarked against a clean ground truth.
GAO reviews found that some Security Risk Category I ammunition shipments had inaccurate inventory item codes, and a separate GAO review found DoD's biennial stockpile reports to Congress did not include information on all risks, with the number of critical materials in shortfall increasing by 167 percent from fiscal years 2019 to 2023 under DoD's own reporting.
✓ DOCUMENTEDcase against
Wargame design choices such as scenario framing and rule construction are not value-neutral and can shape which munitions appear as binding constraints, independent of real-world stock levels.
A peer-reviewed methodological analysis of wargaming as an IR research method concludes there are reasons to be skeptical of the representative nature of wargames, since designing a scenario representation is not an intellectual or value-neutral activity.
✓ DOCUMENTEDcase against
Logistics feasibility itself is flagged as an open question within the game's own findings, undercutting confidence in derived sustainment metrics.
Analysis of the CSIS game states it is not clear if US logistics forces have the capacity to sustain the US operations conducted during these iterations.
○ REPORTEDcase against