Who would win
a U.S.–China
war?

If the question is whether China could conquer the continental United States, or the United States could occupy China, it starts from an implausible conventional-war objective and ignores nuclear deterrence. The more useful question is whether Beijing could compel or impose control over Taiwan, and whether Taiwan, the United States and willing partners could prevent it at a cost they would sustain. A successful defense need not destroy China’s military. A Chinese political success need not involve occupying America.

Our assessment is conditional: an invasion can be defeated, but doing so cannot be assumed. A blockade could prove less visibly dramatic at first yet become an extraordinarily difficult political and economic contest. A prolonged conventional war would test production, repair, public endurance and the ability to negotiate. A large nuclear exchange would overwhelm any ordinary accounting of gains and losses. These are analytical judgments drawn from the sources, not findings of a single predictive model.[4][5][9][21]

The training and readiness problem

Modern combat is coordinated work under uncertainty. Buying a sensor, aircraft or missile does not demonstrate that a force can integrate it with other units while communications degrade and people are lost. Exercises provide evidence of preparation, but participants know more about the setting and constraints than forces would in a surprise crisis. Experience in a different war is informative without being a direct rehearsal for this one.[1][22]

Maintenance also changes the force before an opponent does. GAO’s August 2026 submarine audit reports more than 15,000 operational days lost to maintenance delays across fiscal years 2016–2025. It withholds the operational-availability figures from the public version. That supports concern about readiness; it does not let us calculate how many U.S. submarines would be available in a particular Taiwan crisis.[8]

China’s missile numbers, read carefully

The original Pentagon table distinguishes missiles from launchers and separates range classes. That matters more than a dramatic combined total. Nuclear-capable systems appear within the broader ground-based force, while many other missile types are outside this table. A count also tells us little about serviceability, allocated missions or how well a force would operate under disruption.[1]

Nor does missile reach establish assured control of an area. Military effect depends on information, decisions, countermeasures and the behavior of the opponent. A label such as “carrier killer” compresses a difficult sequence of uncertainties into a slogan. The analytical question is what a system can reliably accomplish as part of an operating force.

Recent expenditure raises a concrete question about a subsequent major conflict, but “America has run out of missiles” is too sweeping. CSIS’s July 2026 analysis distinguishes Iran-related depletion from earlier shortages and notes that important anti-ship weapons were not heavily used in that campaign. A land-attack missile cannot simply substitute for every anti-ship weapon or defensive interceptor.[7]

Money does not arrive as a finished weapon

A larger appropriation can fund more production without producing an immediate delivery. Contracts, components, skilled workers, test capacity and supply chains all impose delays. Factories can expand, but the relevant question in a short war is which weapons already exist and can be delivered. Over a longer war, sustaining production becomes a strategic capability of its own.

The same distinction applies to ships. China’s large industrial base is a meaningful structural advantage; commercial tonnage is not a direct measure of replacement warships. Combat systems, propulsion, specialized labor and naval integration create separate constraints. On the U.S. side, GAO’s April 2026 testimony describes persistent shipbuilding cost and schedule problems. Announced spending should not be counted as a fleet already delivered.[1][23]

Who would attack first? In a deliberate Chinese blockade or invasion scenario, Beijing initiates the coercive action around Taiwan by assumption. That says nothing about the likelihood of that choice. In an accidental confrontation, responsibility may be disputed and subsequent decisions could matter more than the original intent. Cyber activity, inspections and military intimidation also complicate what counts as the beginning.

The explorer below is an original, high-level synthesis of published research. It describes political and strategic mechanisms, including humanitarian consequences. It is not a forecast of specific attacks, a replay of a real plan, or a claim that public evidence can reveal the future. All six stages remain available as a readable transcript.

The aftermath could extend beyond blast damage and radiation. Peer-reviewed work models how soot from particular nuclear-war scenarios could disrupt climate and food production. The cited Nature Food study examines India–Pakistan and U.S.–Russia scenarios, not a U.S.–China exchange. Its headline famine estimates cannot responsibly be presented as the death toll of this hypothetical war.[15]

Trade disruption would reach both combatants

The economic pathways include interrupted shipping, unavailable components, changes to finance, sanctions and retaliation. An economy can lose output because firms cannot obtain a small but critical input, even when their own facilities remain intact. Companies might reroute some trade while finding other dependencies much harder to replace. Direct physical damage is only one part of the bill.[10][11]

China is deeply connected to global production and consumption. Those links create exposure for China and costs for countries trying to impose sanctions. Economic pressure could influence choices without compelling a rapid reversal. Third countries would differ in their interests and willingness to bear losses; coalition formation would be an ongoing task.[11][18]

The semiconductor prize is not a machine you can simply take

A semiconductor fabrication facility depends on reliable utilities, specialist workers, tools, software, materials and international relationships. Physical control would not guarantee continued production at previous quality and scale. War could interrupt the system even if some facilities survived. The industrial consequences would depend on how long disruption lasted and how successfully production could be substituted elsewhere.[10]

The human cost is larger than military losses

Any serious account must include civilian injury and death, displacement, interrupted medical care, lost livelihoods, psychological trauma and family separation. A conflict involving infrastructure disruption could harm people far from a battlefield. Protecting civilians and sustaining humanitarian access would affect the scale of the disaster; neither follows automatically from a side’s military success.

We do not attach a fabricated casualty range to an unspecified war. Published games can count modeled military losses under particular rules. They cannot by themselves establish a comprehensive toll across civilians, long-term health, food systems and a possible nuclear branch. The uncertainty is a reason to explain the pathways to harm clearly, rather than conceal them behind a dramatic number.[4][15]

Where reasonable analysts disagree

Disagreement often begins with assumptions rather than arithmetic. Some analyses emphasize a defender’s opportunity to frustrate a difficult amphibious operation. Others emphasize China’s proximity, the uncertainty of intervention and the fragility of distant support. Different choices about warning, civilian shipping, access, adaptation and the duration of war can produce different answers before the first simulated exchange.[4][5][9]

The public wargames used here share important limitations. Participants are not the real decision-makers. Rules simplify the world. Some capacities are classified; others do not yet exist in the form a scenario assumes. An older study of a future force can be overtaken by procurement, expenditure or policy. The three CSIS game projects are related research efforts, not three fully independent confirmations.

This report’s own explorer is simpler still: it is a set of transparent narratives and conditional interpretations. Its value is showing how assumptions change a judgment. It cannot substitute for the underlying reports, and it should not be treated as a scientific estimate of war outcomes.

What would change this assessment

More credible evidence about readiness, delivered munitions and repair performance would change the military assessment. Clearer government commitments about participation and access would change coalition assumptions. Demonstrated civil resilience would change how much time Taiwan could retain political options. A durable agreement reducing coercion and improving crisis communication would change the risk of reaching these scenarios at all.