Military power is usually compared through inventories: ships, aircraft, tanks, missiles and drones. Inventories describe accumulated equipment, they say much less about the industrial system that must keep it operational after sustained use, disruption and loss.

The harder comparison concerns five connected capabilities:

  • producing new equipment
  • replenishing consumables
  • repairing damaged systems
  • converting deliveries into operational formations
  • adapting designs while production continues.

A force can be strong in one and weak in another, a large order book can coexist with slow delivery; rapid assembly can coexist with dependence on imported components. A successful interception can still deepen a replenishment problem.

Military endurance is the ability to regenerate usable capability over time.

Military endurance depends on the interaction between:

  • industrial output
  • existing stocks
  • maintenance
  • trained people
  • money
  • access to external supply.

Neither the most sophisticated platform nor the lowest-priced weapon is a sufficient unit of comparison.

This expanded analysis groups the evidence into resource mobilisation, the five asset domains, European national systems, wider international production models, and the economics of sustaining an exchange. The country cases describe different industrial structures rather than a league table of national superiority.

Read the industrial data before comparing the weapons

Much of the public debate moves too quickly from an announcement to an assumed military result. An appropriation, an order, a new factory, a production target, an accepted delivery and a deployable capability are separate events. A meaningful comparison must specify which event was measured.

MeasureWhat it establishesWhat it does not establish
Budget or appropriationResources authorised or spent, within an accounting definitionOrders placed or systems delivered
Firm order / backlogContracted demand, subject to the disclosed scopeImmediate output or a guaranteed delivery date
Options / framework / soft backlogPotential or conditional demandAll quantities being firmly purchased
Installed or nameplate capacityA stated production capability under defined conditionsContinuous utilisation, input availability or accepted output
Production cadenceA rate at a stage of manufactureAn annual accepted-delivery total
Accepted deliveryA customer receives an accepted systemA trained, supported and combat-ready unit
Mission availabilityEquipment can perform a specified task at a specified timeUnlimited operational tempo or replacement capacity
Replacement lag is the time required to restore a specified capability after a loss or procurement decision.
Throughput is the quantity delivered during a stated interval.

They are different: a shipyard may deliver one vessel annually while each vessel spends several years in construction. Production elasticity is a third measure—the sustainable increase in output achievable over a stated period.

For shorthand, this article uses industrial replacement rate for the regeneration of usable combat capacity. It is an analytical term, not a published global statistic. Its unit must be defined separately for each domain: accepted rounds, deployable ships, sustainable sorties or equipped formations cannot be added together.

The same discipline applies to geography. A company’s headquarters, ownership, assembly location, component origins, design authority and customers may sit in different countries. “European”, “EU-produced” and “manufactured by a European-headquartered group” are not interchangeable descriptions.

Unless a passage explicitly attributes an inference to a source, the comparative industrial implications below are the author’s analysis. Figures preserve their original dates, units and uncertainty. Forecasts are marked as forecasts; the charts use hatching for future or expected values.

The resource surge: spending is the beginning of the conversion chain

The Stockholm International Peace Research Institute (SIPRI) estimates world military expenditure at USD 2.887 trillion in 2025, with the United States at USD 954 billion, China at an estimated USD 336 billion and Russia at an estimated USD 190 billion.1 These figures describe resources committed under SIPRI’s methodology, they cannot be read as proportional amounts of equipment or war-sustaining output.

2026-11-13T11:42:58.232588 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 0 20 40 60 80 100 120 Military expenditure (current USD billion) Germany France Italy Poland Spain Netherlands Sweden Romania Finland Czechia 114 68 48.1 46.8 40.2 28.9 16.5 9.7 8.1 7.1 Selected EU states: military spending, 2025 Source: SIPRI, Trends in World Military Expenditure 2025, Table 1. Spending indicates demand and resources; it does not measure factory output.
Current USD billion; selected countries on the same SIPRI basis. Spending measures resources, not national factory output. 1

Inside the EU, different fiscal scales coexist with different industrial specialisations. Germany’s spending base is larger than those of the other countries shown, but that does not mean it can independently replace every capability faster. A smaller country can control a scarce component, a production process or a maintenance capability that a larger customer needs.

The European Defence Agency (EDA) latest consistent series places EU expenditure at €349 billion in 2024, €418 billion in 2025 and a projected €454 billion in 2026.2

2026-11-13T11:42:58.423671 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 2024 2025 2026 projected 0 100 200 300 400 500 EUR billion, constant 2025 prices 349 418 454 EU spending: reported and projected Source: EDA, Defence Data 2025–2026, Figure 1. The 2024 value is rebased; do not splice it with an older current-price series.
Constant 2025 EUR billion. EDA’s 2025–2026 publication vintage; the 2026 value is projected. 2

For 2025, EDA reports €115 billion in equipment procurement. Collaborative procurement represented 24% of equipment spending, including arrangements with third parties.2 This is not a measure of EU-only procurement, European manufacturing content or national supply independence; the report also flags incomplete and uneven reporting.

The definition can change the apparent policy result. NATO reported that all Allies met its 2% spending guideline in 20253; SIPRI’s own accounting placed 23 of the 32 NATO members at or above 2%.1

Company revenue is another intermediate measure. SIPRI’s Top 100 arms-producing and military-services companies generated USD 679 billion in arms revenue in 2024.4 That total includes services and revenue recognition across long programmes. It is evidence of industrial economic activity, not a count of completed weapons.

The strategic question is conversion: how much of the new demand becomes accepted output, on what schedule, and with what continuing dependence on scarce inputs? Spending can support that conversion, but it can also be absorbed by inflation, programme complexity, infrastructure work and the cost of rebuilding skills.

Five domains, five regeneration problems

DomainWhat must be regeneratedIndustrial constraint to trackBetter operating measure
NavalShips, crews, sensors, magazines and repair accessYards, design maturity, propulsion, skilled labourDeployable, armed and supported ships
AircraftAirframes plus engines, software, crews and supportAccepted configurations, parts, maintenance and trainingSustainable mission-capable sortie generation
Land vehiclesNew builds, upgrades, recovered and repaired equipmentRepair depth, stored-stock condition, componentsEquipped formations sustained after losses
MunitionsUsable, compatible and certified roundsEnergetics, bodies, propulsion, seekers and acceptanceStocks, deliveries and expenditure by type
Uncrewed systemsOperational platforms and effective design revisionsComponents, quality, software and operator preparationUseful systems fielded per adaptation cycle

Ships concentrate sensors, launch capacity, mobility and endurance in capital assets that are difficult to reproduce quickly. A navy can lose operational capacity through maintenance delays, damage, shortages of ammunition or unavailable crews without losing a hull. Construction output and fleet availability therefore need separate accounts.

U.S. Government Accountability Office (U.S. GAO)'s April 2026 assessment reported a one-submarine-per-year construction pace for Virginia-class work as of June 2025, against a two-per-year goal. The Navy nevertheless accepted two Virginia-class submarines in 2025, both over three years late.5 The apparent contradiction disappears once work pace and completed deliveries are separated.

The same assessment recorded the November 2025 termination of work on four of six contracted Constellation-class frigates, with the other two under construction and review at the assessment date.5 This is an example of a programme failing to convert expenditure and contractual commitments into the intended fleet capability.

Commercial shipbuilding provides a useful industrial backdrop. The UN Trade and Development (UNCTAD) attributes 54.6% of 2024 delivered merchant-ship gross tonnage to China, 28% to South Korea and 12.6% to Japan.6 Those shares concern completed commercial vessels, not warship production or repair capacity.

2026-11-13T11:42:58.583444 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 0 10 20 30 40 50 60 Share of delivered merchant-ship gross tonnage (%) China South Korea Japan Other countries 54.6% 28% 12.6% 4.8% Merchant shipbuilding: delivery shares, 2024 Source: UNCTAD, Review of Maritime Transport 2025, Table II.6. These shares do not measure naval construction, repair capacity or combat power.
Share of gross tonnage delivered for propelled seagoing merchant vessels of at least 100 GT. Commercial deliveries do not measure warship production. 6

The industrial implication is broader than a U.S.–China comparison. Japan and South Korea also hold substantial commercial construction capability. Access to partner production, repair and specialist supply could matter alongside domestic expansion; it remains conditional on compatibility, contracting, transport, available yard slots and the demands of a conflict affecting those partners.

A Center for Strategic and International Studies (CSIS)’s research describes the overlap between China’s commercial and naval yards.7 Shared infrastructure and suppliers can support military construction, but commercial tonnage cannot be converted mechanically into submarine output. The often-cited 230-times comparison discussed by CSIS concerns shipbuilding capacity, rather than observed ratios of warship deliveries or combat power.8

Cost evidence is useful when the accounting basis is common

US Congressional Budget Office (CBO)’s January 2025 shipbuilding study supplies a more defensible cost comparison than a global list of unspecified ship prices. Table 5 estimates average construction costs for purchases under the Navy’s 2025–2054 plan, in billions of constant 2024 dollars.9

Ship classNavy estimateCBO estimate
Ford-class carrierUSD 15.3bnUSD 16.5bn
Columbia-class SSBNUSD 7.9bnUSD 9.4bn
Virginia-class SSN with payload moduleUSD 4.3bnUSD 5.1bn
Arleigh Burke Flight III destroyerUSD 2.7bnUSD 2.7bn

These are historical plan averages, not 2026 market quotations. The table excludes research and development and is not a lifecycle-cost comparison; CBO’s class-specific footnotes define the averaging boundaries.9 The difference between estimators is itself informative: apparent affordability depends on assumptions about future construction performance.

A carrier’s air wing, a submarine’s deterrence mission and an escort’s air-defence function are not equivalent outputs. Cost per tonne or an invented effectiveness score cannot establish which is strategically “better value”. A useful decision compares the cost of providing a named mission, including support, alternatives and the consequences of being unable to perform it.

Aircraft: acceptance timing can conceal the production problem

Lockheed Martin’s annual series records F-35 deliveries of 142 in 2021, 141 in 2022, 98 in 2023, 110 in 2024 and 191 in 2025.10 The series is a customer-delivery measure, not a direct count of airframes newly manufactured during each year.

2026-11-13T11:42:58.744999 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 2021 2022 2023 2024 2025 0 50 100 150 200 Aircraft delivered in calendar year 142 141 98 110 191 F-35 customer deliveries, 2021–2025 Source: Lockheed Martin, Q4 2025 Corporate Overview, slide 12. Acceptance timing and backlog release affect annual deliveries.
Company-reported deliveries; includes the effects of acceptance timing. This is not a series of newly manufactured airframes alone. 10

GAO found that all 110 aircraft delivered in 2024 were late, by an average of 238 days, with Technology Refresh 3 hardware and software a principal driver.11 A later delivery surge can include acceptance of aircraft held back earlier. It should not automatically be extrapolated as a new sustainable annual factory rate.

Dassault reported 26 Rafale deliveries in 2025 and a backlog of 220 aircraft at year-end.12 The backlog supports production continuity, but dividing it by one year’s deliveries does not produce a delivery schedule: customers, configurations and future ramps differ.

For China, The Royal United Services Institute (RUSI) discussed late-2025 estimates of around 120 J-20A/J-20S and 100 J-16 aircraft produced annually.13 These are reported open-source estimates. Placing them beside audited or company-reported deliveries does not make the accounting bases identical.

The regeneration problem extends into engines, weapons, software, airfields, spares and personnel: a force can receive aircraft faster than it can prepare additional operational units. Conversely, better maintenance or parts availability can increase usable capacity before another airframe is delivered, industrial analysis should track both routes.

Land vehicles: repair and refurbishment are production’s partners

Land systems reach service through new manufacture, upgrades, repair, reactivation and external transfer. These routes have different costs, lead times and limits. Combining them in a single “tanks produced” number conceals whether a force is expanding manufacturing or drawing down an inherited stock.

UK Defence Equipment and Support describes Challenger 3 as an upgrade of 148 existing Challenger 2 tanks.14 It is a substantial modernisation programme, but not evidence of a new-build tank line delivering 148 additional hulls.

RUSI’s February 2024 assessment attributed approximately 80% of Russian tank and other armoured-vehicle output in its assessment to refurbishment and modernisation of stocks.15 That is a dated estimate, not a current composition figure. Its later-stock-depletion forecasts must not be treated as observations that reserves were exhausted on schedule.

Programme cancellations also change the industrial picture. The U.S. Army’s June 2025 announcement terminated M10 Booker low-rate initial production and ruled out the previously planned full-rate stage.16 A planned line belongs in a forecast until its continued funding and production status are established.

The practical account should separate vehicles destroyed, repairable, undergoing repair, upgraded and returned to units. A drone strike may produce a permanent loss, a temporary mission kill or damage requiring repair. Acquisition price alone does not measure the industrial or operational cost of that event.

Recovery vehicles, transporters, spare assemblies and depot capacity are consequently part of the combat-generation system. Increasing the number of frontline vehicles without those supporting capabilities can create an inventory that is difficult to keep mobile and protected.

Munitions: production capacity is not delivery capacity

A factory’s stated capacity tells us what it is designed to produce, not how many finished missiles it delivers in a given year. Announced expansions and new orders tell us something about future capacity and demand. Neither tells us how quickly stocks can be replenished today.

That distinction matters when comparing missiles with the other systems in this article. A munition is expended when used, so the relevant measure for replenishment is the flow of finished deliveries, not the size of the order book. Where delivery figures are not public, I distinguish announced capacity from observed output rather than treating one as a proxy for the other.

Supporting evidence: deliveries and capacity

Lockheed Martin reported 620 PAC-3 MSE deliveries in 2025. Its January 2026 framework envisaged increasing annual capacity from approximately 600 to 2,000 under a seven-year agreement.17 These are two different observations: an achieved delivery total and a future capacity ambition.

The U.S. Army’s April 2026 announcement for a new Parsons, Kansas facility described full operational capacity of 12,000 M795 projectiles monthly, within a wider 100,000-round monthly expansion goal.18 An opened facility does not establish that it is already operating continuously at nameplate capacity.

MBDA reported an accelerated Aster delivery in July 2025 and a commitment to more than halve 2026 production lead time relative to 2022. Its statement on doubling new-missile production between 2023 and 2025 concerned the group as a whole.19 Neither statement should be converted into an unspecified Aster inventory figure.

The European Parliament’s September 2025 briefing described a projected two-million-round annual European ammunition capacity by end-2025.20 That projection should retain its status until a comparable accepted-output series is available. It also overlaps the output and capacity of companies discussed in the national cases below.

For endurance, the relevant unit is a usable, compatible round at the point where it is needed. Aggregate missile or shell totals conceal variant, launcher, certification and logistics differences. Replenishing the wrong mix does not restore the same operational capability.

Uncrewed systems: distributed assembly, concentrated inputs

“Drone production” spans small tactical aircraft, large reusable UAVs, loitering munitions, surface craft, underwater vehicles and ground robots. Their industrial economics differ. A short assembly task for a small aircraft is not the complete replacement cycle of a large uncrewed platform.

RUSI’s November 2025 research identifies dependence on Chinese-origin motors, sensors, electronics and upstream materials, alongside the advantages of flexible Ukrainian designs.21 Domestic assembly can coexist with concentrated foreign input dependence.

That distinction changes the sovereignty question. The relevant test is whether alternative inputs are available, qualified, affordable and compatible with production at the required scale. Replacing a component is not only a purchasing decision: a design may need to be retested and its operating support revised.

The industrial advantage of adaptation lies in the entire feedback loop: recognising a changed requirement, revising the design, procuring the inputs, verifying quality, training users and fielding a useful revision. More assembly sites do not necessarily create that complete loop.

Europe is a network of national systems

A useful European comparison starts with the capability a country supplies to the wider network. Some states sustain broad platform portfolios; others provide specialised electronics, munitions, propulsion, manufacturing capacity or support. A procurement customer can simultaneously be an importer, an assembly location and a component supplier.

Country / clusterIndustrial emphasis in these casesWhat the evidence must distinguish
FranceCombat aircraft, land systems, missiles and programme continuityReported cadence, export demand and future expectations
GermanyLand systems, ammunition and industrial expansionGerman output versus international group capacity
ItalyNaval construction, aerospace and systems integrationDefence versus civil work; hard versus soft backlog
SpainNaval programmes, electronics and integrationCompany/segment revenue versus national physical output
PolandAcquisition, ammunition expansion and localisationImports, future domestic capacity and qualified production
SwedenAircraft, sensors, underwater and ground-combat systemsLarge orders versus rates of delivery
FinlandVehicles, sustainment and distributed manufacturingFirm orders, options and partner-country production
CzechiaAmmunition and land-system supply networksGroup ownership versus plant geography and contracting
NetherlandsSelected technologies and industrial partnershipsPriority-setting versus delivered production capacity
Romania / Baltic partnersLocations in expanding cross-border production networksPlanned plants, commissioning and actual output

This map is an analytical synthesis of the cases that follow, not an exhaustive national capability catalogue. “EU” means EU member states; the UK and Norway are addressed separately. The source evidence concerns selected firms and programmes rather than a complete census of every national factory.

France: continuity across national and export demand

France’s 2025 arms-export report gives a specific production-cadence example: CAESAR output rose from two systems monthly before the war to six in 2024, while eight monthly was expected for 2025.22 The future wording matters; the source does not confirm twelve months of deliveries at the expected rate.

2026-11-13T11:42:58.900617 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ Pre-war baseline 2024 reported 2025 expected 0 2 4 6 8 10 CAESAR systems per month 2 6 8 CAESAR: reported and expected cadence Source: French arms-export report to Parliament 2025, printed p. 17. A monthly cadence must not be multiplied into assumed annual accepted deliveries.
Two before the war and six in 2024 were reported; eight per month was expected in the 2025 report. The expected value is hatched. 22

The French case illustrates why exports can contribute to industrial endurance. A long production run can help preserve specialist skills and supplier continuity between national procurement cycles. It also creates a scheduling problem when several customers need accelerated delivery at once. A full order book is useful industrial visibility, but it is not spare capacity waiting to serve an unexpected domestic requirement.

KNDS’s 2025 results separate €2.5 billion of German land-system revenue, €1.3 billion of French land-system revenue and €612 million in its ammunition segment.23 These are segment financial observations, not three independent measures of national capacity. Shared group programmes and suppliers make simple country addition unreliable.

Germany: scale-up through an international production footprint

Rheinmetall’s September 2026 presentation shows group annual 155mm capacity of 70,000 in 2022, with targets of approximately 1.1 million in 2027 and 1.5 million in 2030. It also shows separate propellant-capacity targets.24 The figures concern an international corporate network, not German-only accepted output.

2026-11-13T11:42:59.079699 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 2022 2027 target 2030 target 0 250 500 750 1000 1250 1500 1750 155mm ammunition (thousand/year) 70 1,100 1,500 2022 2027 target 2030 target 0 5 10 15 20 Propellant (thousand tonnes/year) 5 12 20 Rheinmetall: group capacity and targets Source: Rheinmetall Investor Presentation, slide 34. International group footprint. Different units shown separately; these quantities do not add up to complete rounds.
Separate units for ammunition and propellant. International corporate footprint; 2027 and 2030 are targets, marked with hatching. 24

Geographic expansion can diversify production and draw on different workforces. It can also leave several plants dependent on common upstream inputs. The useful measure is not simply how many sites appear on a map, but how many independently qualified production paths can deliver a complete product under disruption.

Kiel’s May 2026 procurement study distinguishes first and final expected delivery dates. For the German sample, average reported initial deliveries were around two to three years after ordering, while full fulfilment was around four to five years. Only 151 of 396 recorded items had both dates.25 This is evidence from a disclosed sample with substantial missing data, not a universal German delivery lag.

The distinction is strategically important even without a national score. Initial equipment deliveries can begin improving a force while complete formations and support arrangements remain several years away. Tracking the first delivery alone overstates the speed of regeneration.

Italy: naval capacity and integration require separate accounts

Fincantieri reported 24 group vessels delivered in 2025 and a year-end backlog of €41.095 billion. Its larger €63.195 billion “total backlog” includes soft backlog.26 These group figures span commercial and defence activities; none is a count of Italian warships available for wartime replacement.

Leonardo’s final 2025 report records €19.503 billion in revenue and €46.624 billion in backlog.27 It provides evidence of systems-integration demand and programme activity across an international group. It does not make the entire total an Italian domestic weapons-output figure.

The analytical value is the combination of platform and subsystem capabilities. Naval construction depends on combat systems and electronics as well as the hull. Expanding one without the other can produce a partially complete platform or a longer integration queue. Civil work can preserve yards and skills, but it also occupies facilities and supplier capacity.

Spain: a different mix of naval programmes and electronics

Navantia’s individual-company 2025 accounts report approximately €1.729 billion in revenue, including activity associated with the F-110 and S-80 programmes.28 Indra reported a defence backlog of €11.336 billion at year-end, while explaining that large modernisation programmes and changes in group scope affect comparisons.29

These are distinct roles in the industrial chain. Naval programme execution and electronics integration can both constrain delivery, but a revenue or backlog total cannot reveal which stage is limiting a particular ship. The right follow-up measures are accepted platforms, subsystem milestones, test completion and time to operational integration.

Spain should therefore be examined as a set of programme and supplier capabilities, rather than placed beneath a generic European capacity label. The same principle applies to other countries whose importance is concentrated in selected systems or production stages.

Poland: procurement scale and domestic production are different transitions

In September 2025, Poland’s government announced that a PGZ–BAE Systems agreement was intended to bring annual shell production to around 130,000 within two years, with PLN1.3 billion allocated to Dezamet investment.30 It was a ramp-up announcement, not evidence of that output already being achieved.

Poland illustrates two parallel industrial tasks: acquiring capability promptly and building the means to sustain or manufacture it domestically. Imports can accelerate inventory growth before local production is qualified. Localisation can increase later resilience, but an announced transfer of technology does not establish completed tooling, trained staff or an independent component supply.

WB Group reported the groundbreaking for its joint guided-missile factory with Hanwha Aerospace in Gorzów Wielkopolski on 4 November 2026.31 Construction is an intermediate milestone. It should not be entered into an output table as an operating line.

Comparisons should track the transition explicitly: imported delivery, local support, assembly, component production and independent manufacturing are different steps. They may overlap for years within the same programme.

Sweden: a broad specialist portfolio with a growing backlog

Saab reported SEK79.1 billion in 2025 sales, SEK168.5 billion in order bookings and SEK274.5 billion in year-end backlog. It described deliveries and demand across aircraft, surveillance, underwater and ground-combat activities.32 These indicators establish scale and demand, while leaving system-specific throughput to be measured separately.

The Swedish case shows why a small home market need not imply a narrow industrial contribution. Specialist products can serve several customers and connect to wider national systems. The constraint is that scarce engineering and production resources must be allocated across programmes whose demand can increase simultaneously.

The most useful additional data would therefore be programme-level delivery cadence, supplier expansion and acceptance performance. A group backlog divided by annual sales is a financial coverage indicator, not the number of years a new customer must wait for a particular weapon.

Finland: common designs, options and distributed manufacture

Patria reported that seven countries participated in CAVS by the end of 2025 and that orders including options approached 2,000 vehicles. Its annual review also described expansion in Finland and Latvia and a German agreement exceeding €2 billion including options.33 Partner membership, options and firm accepted deliveries must remain distinct.

The industrial attraction of a common design is the opportunity to share development, support and learning across a larger customer base. Distributed manufacturing can increase access to facilities and personnel. It can also complicate configuration control if national changes erode the common baseline.

A useful programme comparison would record common components, local variations, qualified production sites and the ability to exchange spares. Programme size alone does not establish those forms of interchangeability.

Czechia: ammunition networks are not contained by headquarters

CSG’s 2025 issuer statement described a €15 billion backlog, automated 155mm filling lines in Slovakia, licensed large-calibre production in Ukraine and expanded manufacturing at Excalibur Army.34 These activities illustrate the difference between a Czech-headquartered group and production occurring entirely inside Czechia.

The relevant national contribution can include contracting, finance, component manufacture, final processing or integration. Counting a group’s entire portfolio as one country’s physical production would hide that network and could double count the contribution of partner locations.

For ammunition in particular, the country where a round is finished may differ from the origins of its body, energetic materials or other components. A useful resilience map follows those dependencies without mistaking final assembly for full input independence.

The Netherlands, Romania and the Baltic production network

The Netherlands’ 2025–2029 industrial strategy prioritises quantum, smart materials, space, intelligent systems and sensors.35 This is a selective technology-and-partnership approach, rather than a policy of making every military asset domestically.

Elsewhere, the corporate and programme evidence places Romania and Baltic partners inside expanding manufacturing networks: Rheinmetall’s presentation identifies country locations and planned expansion24, while Patria describes production investment in Latvia.33 These entries must be followed through commissioning and qualified deliveries before being treated as realised additions to European output.

The distinction suggests a more useful European policy question: which critical stages need additional independent supply paths, and which benefit most from common production runs? Multiplying final-assembly sites is not necessarily the best answer if all sites depend on the same scarce upstream input.

Across the EU boundary: the UK and Norway

The UK’s September 2026 Defence Nuclear Enterprise update reported £12 billion of spending in FY2025/26, a workforce above 47,000 and demand expected to reach around 65,000 by 2030.36 These figures reveal the scale of a specialised industrial undertaking; future workforce demand is not evidence of workers already available.

Nuclear-submarine programmes illustrate why continuity matters. Skills, safety assurance, propulsion, construction and support must remain coordinated over long periods. Interruptions cannot necessarily be reversed by placing a larger short-term order. Industrial resilience includes preserving the ability to execute the next programme.

Norway-based Nammo reported 2025 revenue of NOK14.41 billion, order stock of NOK54.979 billion and 86.6% on-time delivery to military customers.37 Its facilities and expansion span several European countries and the United States. The reported delivery-reliability measure adds a dimension that backlog alone misses, although it does not reveal quantities for every munition.

Europe’s industrial system consequently extends beyond EU membership. Procurement, ownership, manufacturing and operational interoperability follow different boundaries. A realistic regional assessment should show the connections while keeping the statistical populations separate.

Global production models: different strengths, different constraints

United States: programme depth and the time needed to convert it

GAO’s July 2026 weapons assessment examined 104 of the Department’s costliest programmes and found an overall average capability-delivery horizon above 12 years.38 That is an acquisition-programme measure, not the time required to manufacture every individual aircraft or missile.

The U.S. evidence reviewed here spans three different clocks: long naval construction programmes, aircraft deliveries affected by configuration acceptance, and multi-year missile-capacity expansion. A single claim that the country is either “industrialising” or “deindustrialised” conceals those differences.

The operational question is which constraints can be relieved within the relevant horizon. Extra accepted munitions, repairs, parts and software changes may improve current capacity sooner than a new platform programme. Longer-term investment is still necessary where replacement requires infrastructure or specialist labour that cannot be created quickly.

China: commercial scale does not eliminate specialised constraints

The commercial shipbuilding and drone-component evidence supports a structural case for Chinese manufacturing scale. It does not justify converting every industrial advantage into an identical military surge factor. Specialist nuclear construction, sensors, integration and operational support still require their own measures.

Commercial activity can support a workforce and supplier ecosystem independently of defence orders. That creates a different starting point from a factory dependent on one intermittent national programme. But the availability of commercial capacity during a major conflict remains an analytical question, not an observed wartime outcome.

For China, the strongest public comparisons in this article are therefore the defined merchant-delivery series, disclosed industrial structures and labelled open-source estimates. They provide evidence of scale without claiming access to an audited national wartime-production plan.

Russia: persistence must be analysed through firms and regeneration routes

CSIS’s July 2026 study examined five firms—Uralvagonzavod, Almaz-Antey, KRET, Kalashnikov and Ushkuynik—covering armoured vehicles, air defence, electronic warfare, weapons and drones. It found expansion alongside labour shortages, imported-component dependence and sanctions-related costs.39 This is a firm-based study, not an audited census of all Russian output.

The stronger interpretation is selective persistence: a state can sustain certain production categories while facing constraints elsewhere. That does not demonstrate unlimited quality, uniform self-sufficiency or a constant rate of expansion across all systems.

Regeneration routes remain central. Stored vehicles can support near-term returns to service while consuming a finite inheritance. New manufacture creates a different supply path. The composition of total deliveries is therefore as important as their headline quantity.

Ukraine: industrial capacity, funding and delivered equipment

Ukraine’s National Security and Defense Council described projected 2026 defence-industrial capacity of USD 55 billion.40 The monetary valuation is a capacity estimate, not realised annual production or a funded procurement total.

For 2025, the Ministry of Defence reported USD 6.1 billion in foreign industry funding, including nearly USD 1.8 billion through the Danish model and over USD 4.3 billion through direct partner procurement. It also described USD 1.1 billion from frozen-asset proceeds.41 The latter is a source of finance that can overlap procurement mechanisms; adding every category as an independent amount would double count.

This creates a different capacity problem from a country whose factories are fully funded but physically constrained. Where usable production capacity is underfunded, additional contracts may matter before new construction does. However, dividing foreign funding by a national capacity valuation would not produce a valid utilisation rate: domestic spending, scope and accounting definitions differ.

The same ministry’s March 2025 announcement described more than 1.5 million drones procured in 2024 and a plan to purchase approximately 4.5 million FPVs in 2025.42 A November 2026 update reported over 1.5 million drones and other equipment items delivered through DOT-Chain since its July 2025 launch.43 Those are different populations and periods, not contradictory annual production totals.

The UK’s September 2025 Octopus announcement set a target of thousands of Ukrainian-designed interceptor drones monthly from UK manufacture.44 It illustrates a partnership route to expansion, while remaining a target until corresponding accepted-delivery evidence is supplied.

Ukraine’s industrial model also places adaptation and procurement unusually close together. The important strategic lesson is the ability to turn a changing operational requirement into funded, accepted and usable equipment. That is a system capability, rather than simply a low assembly cost.

Iran and Israel: input dependence versus export contracting

Ukraine’s intelligence service has published foreign-component findings for an upgraded Iranian-made Shahed-136.45 The attributed evidence does not support a blanket claim that Iranian drone manufacture is independent of foreign inputs. A nationally assembled system can retain external supply dependence.

Israel’s Ministry of Defense reported USD 19.2 billion in defence export agreements signed in 2025.46 The measure is contracting, not shipments delivered in the year. IAI’s Mini Harpy product page supplies a separate kind of evidence: manufacturer-stated system capabilities, rather than annual output or cost competitiveness.47

The comparison demonstrates why national “models” should remain specific. Component sourcing, design and integration, export demand and large-scale assembly describe different industrial strengths. None alone establishes the longest possible wartime endurance.

South Korea and Japan: partners in the production system

Korean–Polish industrial cooperation illustrates the movement from export acquisition towards local production and support. The November 2026 construction milestone noted above is relevant to that transition, while leaving commissioning and accepted output to be demonstrated.

Japan’s 2026 defence white-paper digest reports 164 Stable Manufacturing Plans approved in FY2025, including 124 initiatives involving SMEs.48 Those are approvals to strengthen the industrial base, not a count of extra weapons already delivered.

Together with Japan’s and South Korea’s commercial shipbuilding shares, these cases widen the comparison beyond the traditional U.S.–Europe axis. Industrial partnerships can bring capacity and diversification. They do not automatically make every product interchangeable or remove the time required for qualification, integration and support.

Compare the industrial claims without adding incompatible quantities

The following table brings selected observations together specifically to show the differences in their status. It is an evidence dashboard, not a total production count.

System / geographyQuantity and periodStatusScope limit
F-35, global programme191 aircraft, 202510Company-reported deliveriesNot newly manufactured airframes alone
Rafale, global programme26 aircraft, 202512Company-reported deliveriesNot combat-ready squadron formation
PAC-3 MSE620 missiles, 202517Company-reported deliveriesFuture 2,000-capacity ambition is separate
CAESAR6/month in 2024; 8 expected in 202522Reported cadence / expectationNot an accepted annual-delivery series
Parsons M795 facility12,000/month at full capacity18Facility capacity statementNot observed April 2026 utilisation
Rheinmetall 155mm~1.1m/year in 202724Group capacity targetInternational footprint; overlaps European totals
Polish shell expansion~130,000/year within two years of Sep. 202530Government ramp-up targetNot achieved September 2025 output
European ammunition2m/year projected for end-202520Capacity projectionCannot be added to producer-level capacity
CAVS vehiclesNearly 2,000 orders including options by end-202533Order-and-option portfolioNot firm accepted deliveries

The apparent precision of a number does not establish its maturity. A measured delivery total and a future capacity target should be visually distinct even if both are expressed as units per year. “Unknown” is more informative than filling a missing national production estimate with an unsupported score.

Cost exchange, stocks and the economics of endurance

Low-cost systems can impose disproportionate defensive expenditure. But a weapon-price ratio is not the same as strategic advantage. Defence protects people, infrastructure, operational freedom and other capabilities whose value may exceed the marginal cost of an engagement.

Take a deliberately hypothetical example: ten attack drones at USD 30,000 each and ten interceptors at USD 3 million each, with one interceptor expended per drone. Attacker hardware expenditure is USD 300,000 and defender interceptor expenditure USD 30 million, producing a 100:1 defender-to-attacker spending ratio. These are assumptions, not verified platform prices or an observed engagement.

The result excludes supporting systems, failed attack attempts, the value of damage avoided, alternative defensive methods and each side’s ability to replenish. A force may rationally accept an unfavourable hardware-price ratio to protect a critical asset. The endurance problem emerges when repeated expenditure reduces future protection faster than replacement restores it.

QuestionData needed
Was the engagement economically justified?Marginal defensive cost, expected damage prevented, available alternatives
Can protection continue?Compatible stocks, reserve requirements, deliveries, expenditure and reload capacity
Was a platform permanently lost?Damage classification, repair cost, recovery and time out of service
Can another weapon substitute?Mission coverage, compatibility, support and operational limits

For one compatible munition type, a transparent account is:

Closing usable stock = opening usable stock + accepted deliveries + inbound transfers − expenditure − other withdrawals − outbound transfers.

The boundary and period must be consistent. A transfer between domestic units is not a new national delivery. Stock removed through expiry or damage must be recorded. For reusable systems, the account also needs transitions into maintenance or repair and subsequent returns to service.

2026-11-13T11:42:59.291431 image/svg+xml Matplotlib v3.9.4, https://matplotlib.org/ 0 5 10 15 20 25 30 Days 0 200 400 600 800 1000 Usable compatible rounds remaining Expenditure 80/day · depletion 25.0 days Expenditure 100/day · depletion 16.7 days Expenditure 120/day · depletion 12.5 days Endurance depends on net drawdown Illustrative: opening stock 1,000; accepted replenishment 40/day. Author calculation. Constant rates; no reserve floor or other withdrawals. Not a forecast.
Author calculation: 1,000 opening usable rounds and 40 accepted replacements per day. Constant rates, no reserve floor or other withdrawals; this is a hypothetical accounting illustration.

With 1,000 opening rounds and 40 accepted replacements daily, expenditure of 80, 100 or 120 daily produces net drawdown of 40, 60 or 80. With no reserve floor or other changes, depletion occurs after 25, 16.7 or 12.5 days. These are constant-rate accounting illustrations, not forecasts.

This model clarifies why initial stock and industrial throughput are complementary. Stocks bridge time until additional output arrives. Output sustains operations beyond that bridge. A large future factory does not resolve a near-term shortage unless existing stocks or external supplies cover the interval.

Adaptation is valuable only when it survives contact with the sensing system

An inexpensive modification can be introduced much faster than a replacement platform. Its effectiveness still depends on the system it is intended to influence. Naval camouflage supplies a compact example of that distinction.

In March 2024, H. I. Sutton compared visible-light and infrared footage associated with the attack on Sergei Kotov.49 The black-painted bow and stern were apparent in visible imagery but not in the infrared footage. His interpretation supported visual deception as plausible and infrared concealment as much less likely.

The observation does not establish a quantified reduction in attack success, a particular radar-absorbent material or a validated machine-recognition rate. Its wider lesson is that rapid modification and effective adaptation are separate outcomes. An industrial feedback loop needs evidence that a revision still performs its intended function.

What a credible industrial-endurance dashboard would track

The most useful next step is a common reporting structure, rather than another composite national ranking. The following dashboard is a proposed analytical framework, not a set of measured results.

Thematic clusterCore measuresDecision it informs
Demand and financeFunded orders; payment timing; firm backlog; options separatelyWhether demand is stable enough to support investment
Industrial conversionQualified capacity; bottleneck stage; accepted throughput; rejectsWhether more spending becomes usable output
Supply resilienceCritical supplier concentration; alternative qualification; input buffersWhether production survives disruption
Operational regenerationAvailability; repair turnaround; trained crews; compatible stocksWhether delivered equipment restores a force
AdaptationRequirement-to-fielding time; revision acceptance; performance after changeWhether learning reaches usable production
Regional integrationCommon configurations; shared support; plant geography; transfer constraintsWhether partner capacity is actually accessible

A new production site should move through explicit milestones: investment announced, construction started, equipment installed, line qualified, output accepted and sustainable rate demonstrated. Reports should retain the history when a date changes instead of replacing an old target with a new one.

The financial side matters as well. Buyers need enough visibility to justify capacity investment, while producers need working capital and payment arrangements that sustain the ramp. National preferences can preserve strategic capabilities, but excessive variation can fragment otherwise common production and support.

The trade-off is not resolved by choosing either complete national autonomy or unrestricted dependence. Some capabilities require sovereign control; others can be supplied more effectively through qualified partner networks. The practical question is which dependencies are acceptable for a specified mission, time horizon and disruption scenario.

The evidence points towards a more precise understanding of military power. Naval endurance includes repair and availability; air power includes acceptance and sustainment; land regeneration combines manufacture with recovery; munitions require a balanced chain and compatible stocks; uncrewed systems depend on components and effective adaptation.

The strongest industrial system is the one that can continue delivering the required usable capability under sustained pressure. Its advantage must be demonstrated through conversion, regeneration and access—not inferred from a budget, an impressive factory opening or the price of one weapon.

Research scope and downloadable reports

Research updated on 13 November 2026. Sources include government agencies, official programme reporting, company disclosures and original research. Company and government claims remain attributed; estimates and forecasts retain their status. Country coverage is selective, with additional depth on EU industrial differences. No access to classified stocks or production data is assumed.

The notes identify the evidence beside the relevant text. The report library below links to the original documents where a separate PDF or report download was available. Publication year and data year differ: a report published in 2026 may describe 2025 activity, while a future capacity target may appear in either.

Reference / reportPublisher / dateOriginal document
1. Trends in World Military Expenditure, 2025SIPRI
April 2026
Download PDF
2. Defence Data 2025–2026European Defence Agency
16 July 2026
Download PDF
3. Secretary General’s Annual Report 2025: spending announcementNATO
26 March 2026
Download PDF
4. The SIPRI Top 100 Arms-producing and Military Services Companies, 2024SIPRI
December 2025
Download PDF
5. Navy and Coast Guard Shipbuilding: A Disciplined, Strategy-Driven Approach Is Needed to Achieve Ambitious GoalsU.S. GAO
22 April 2026
Download PDF
6. Review of Maritime Transport 2025UN Trade and Development (UNCTAD)
2025
Download PDF
9. An Analysis of the Navy’s 2025 Shipbuilding PlanU.S. Congressional Budget Office
6 January 2025
Download PDF
10. Corporate Overview, Fourth Quarter 2025Lockheed Martin
As of 31 December 2025
Download PDF
11. F-35 Joint Strike Fighter: Actions Needed to Address Late Deliveries and Improve Future DevelopmentU.S. GAO
3 September 2025
Download PDF
12. Deliveries, order intakes and backlog as of December 31, 2025Dassault Aviation
7 January 2026
Download PDF
17. PAC-3 MSE production framework agreementLockheed Martin
6 January 2026
Download PDF
20. Scaling up Arms Production Capacity in the EUEuropean Parliamentary Research Service
September 2025
Download PDF
21. Drones: Decoupling Supply Chains from ChinaRUSI · Robert Tollast
November 2025
Download PDF
22. Rapport au Parlement sur les exportations d’armement 2025French Ministry of the Armed Forces
2025
Download PDF
24. Investor Presentation: Closing GapsRheinmetall
September 2026
Download PDF
25. Leading in spending, lagging in innovationKiel Institute · Wolff, Binder and Morgan
May 2026
Download PDF
26. 2025 Financial Results ApprovedFincantieri
25 March 2026
Download PDF
27. Integrated Annual Report 2025Leonardo
2026
Download PDF
28. Navantia 2025 annual accounts and auditor’s reportNavantia
2026
Download PDF
29. Results for the financial year 2025Indra Group
25 February 2026
Download PDF
32. Annual Report 2025Saab
March 2026
Download PDF
33. Patria’s year 2025Patria
2026
Download PDF
35. Defence Strategy for Industry and Innovation 2025–2029Netherlands Ministry of Defence
4 April 2025
Download PDF
37. 2025 finance and operating performanceNammo
2026
Download PDF
38. Weapon Systems Annual Assessment, GAO-26-108457U.S. Government Accountability Office
2 July 2026
Download PDF
39. Rebuilding to Reconquer: Can Russia’s Defense Firms Deliver?CSIS
14 July 2026
Download PDF
48. Defense of Japan 2026: DigestJapan Ministry of Defense
2026
Download PDF

Disclaimer

I wrote this article with AI assistance. I used AI to support the research, analyse the material I gathered, refine the English, and create charts and tables.

All data comes from publicly available sources. The interpretations and opinions expressed are my own and do not represent the views of any current or former employer.