The Defence Economy

• SIPRI (27 April 2026): $2,887B, up 2.9% in real terms, the 11th straight annual rise and 2.5% of world GDP, the highest since 2009.
• US spending fell 7.5% to $954B (about 33% of the total); excluding the US, spending rose 9.2%.
• Europe rose 14% to $864B; Germany rose 24% to $114B (2.3% of GDP); China $336B (+7.4%); Russia $190B; Ukraine $84.1B (+20%).

• The Congressional Budget Office (15 Sep 2026, as reported by Defense News) put the cost of US combat operations against Iran at about $38B through 1 August, and $2B to $3B for each further month if intensity stays low. A Pentagon inspector general estimate cited alongside it was $33.4B through 29 June.
• Replacing the munitions fired through 1 August would cost about $21.7B: $13.1B for interceptors, $7.3B for land-attack cruise missiles and $1.2B for other munitions.
• CBO estimates the conflict used between one half and two thirds of the US inventory of Patriot, THAAD, SM-3 and SM-6 interceptors since June 2025, and that rebuilding stock would take at least five years even if production rises. The Pentagon does not publish inventories, so this is an estimate and carries wide uncertainty.

This is the clearest example in the deck of the capacity point. The money to buy interceptors exists, but a THAAD or SM-3 is not made faster by a larger cheque.

• The White House asked for about $1.5 trillion: roughly $1.15T of base budget plus $350B that depends on a reconciliation bill. Golden Dome is the most exposed item, with $17.1B of its $17.9B request tied to reconciliation (CSIS, 18 Aug 2026).
• The House Appropriations Committee approved a defence bill of about $1.072T (H.R. 9495) in June. The House passed its NDAA (H.R. 8800) 216 to 212 on 26 July, authorising about $1.15T. The Senate NDAA (S. 4784, about $1.14T) failed a cloture vote 50 to 46 on 14 July.
• Both chambers passed stopgap funding at FY2026 levels in July and August (to 4 and 11 December respectively), which blocks new starts and production-rate increases unless exceptions are written in. The House adopted a budget resolution with $60B for defence, not $350B.
• In June the White House requested an $87.6B supplemental, $67.1B of it for the Department of Defense and mostly tied to the Iran war. It had not passed as of the CSIS update.

Not confirmed: what Congress has done since mid-August. Check Congress.gov for the current state of the stopgap, the NDAA and the supplemental before relying on any of this.

• Lockheed Martin (Q3, 22 Oct 2026, SEC filing): sales up 11% to $20.7B; backlog a record $292.0B (from $193.6B at end 2025), including $52.5B for the PAC-3 MSE interceptor; 43 F-35 deliveries in the quarter.
• RTX (Q3, 20 Oct 2026, as reported): sales about $25.0B, up 11%; backlog a record $340B ($176B commercial, $164B defence); over $81B of awards including about $37B of multiyear Tomahawk and AMRAAM contracts.
• General Dynamics (Q3, SEC exhibit): revenue $14.6B, up about 13%; operating margin 10.5%; backlog a record $220.3B ($270.7B including estimated potential value) after about $98.8B of orders in the quarter, a company book-to-bill of 6.8 to 1. Marine Systems backlog is $146.9B. Reported orders include nine Virginia-class and five Columbia-class submarines.
• Northrop Grumman (Q3 10-Q and release, Oct 2026): backlog $109.6B, book-to-bill 1.5x, sales up 7% to $11.1B, guidance raised.
• Boeing Defense, Space & Security (Q3, 27 Oct 2026): revenue $7.9B, up 15%; operating margin 6.2%, helped by a $252M gain on a business sale; backlog a record $100.5B, 23% from outside the US. Boeing was selected for the Navy's F/A-XX fighter, which is not yet in backlog.

A caution on the General Dynamics number: the unfunded part of submarine backlog depends on future appropriations, so a record backlog is a record of orders, not of cash already approved.

Lockheed Martin says it is still negotiating F-35 Lot 20 and later production contracts, and that its costs on the Lot 20 advance contract began to exceed current authorization and funding in Q3 2026 (Lockheed Martin Q3 2026 release, SEC filing).

A Department of Defense Inspector General report (July 2026, via Defense News) found the Army made about 36,000 155mm rounds a month as of March 2026, against a goal of 100,000 by October 2025. Metal parts production is the limiting factor: the Mesquite, Texas plant could not produce parts to specification. Wilkes-Barre, Pennsylvania reportedly makes about 21,000 a month, and a new Parsons, Kansas load-assemble-pack facility opened in April 2026 (planned 12,000 a month).

• The Pentagon's own estimate rose to $185B in March 2026, $10B more than before, mostly to speed up space-based capabilities.
• The Congressional Budget Office put the 20-year cost at about $1.2 trillion including operations, with space-based interceptors the largest driver. The two figures cover different periods and scopes.
• In April 2026 the Space Force announced awards worth up to $3.2B to twelve companies for space-based interceptors, with a demonstration targeted for 2028. Lockheed Martin, RTX and Northrop Grumman were added as prime contractors on the programme in March.

These figures come from trade-press reports of official numbers, with coverage running to about May 2026. Later changes are not reflected.

Anduril announced a $5B Series H led by Thrive Capital and a16z in May 2026, and said its revenue doubled to about $2.2B in 2025. Later press reports mention a higher tender or new-round valuation (about $90B to $120B), but these were not confirmed by a priced round and are not used here.

Palantir reported US government revenue of $953M in the third quarter of 2026, up 96% from a year earlier and 18% from the prior quarter, and total revenue of $2.31B, up 96%. For the first time its US commercial revenue ($984M) exceeded US government revenue. Palantir raised 2026 revenue guidance to $8.53B (Q3 earnings release, SEC filing).

• NATO's July 2026 estimates, reported ahead of the Ankara summit, show five allies above 3.5% of GDP on core defence in 2026: Lithuania (5.33%), Estonia (5.11%), Latvia (4.92%), Poland (4.68%) and Greece (3.65%).
• Norway, Sweden and the United States are projected above 3% but below 3.5%. All allies except Slovenia are expected to reach 2%.
• These are secondary reports of NATO estimates, which use constant prices. The 2035 goal agreed at The Hague in 2025 is 3.5% for core defence plus 1.5% for related spending.

Rheinmetall reported first-half 2026 sales up about 39% and an order backlog of roughly EUR 80B, but cut its 2026 sales guidance to EUR 13.7B to 14.2B from EUR 14.0B to 14.5B, with the margin target of about 19% kept (secondary reports of the August 2026 results; Q3 not yet checked). The EU's SAFE scheme offers up to EUR 150B of defence loans; reports of the tentative split put Poland first at about EUR 43.7B, though sources disagree on how much is drawn.

Ukraine's stated goal for 2026 was about 7 million military drones, against a reported 4 million or more in 2025 (reported targets, not audited output). Reports in the first half of 2026 describe interceptor drones rising from about 1,500 a day in January and the military receiving twice as many in four months as in all of 2025; of 1,000 weapon systems approved in the first half, 892 were made in Ukraine. Engines are reported as a bottleneck for faster jet-drone interceptors.

Not refreshed: Lockheed, RTX and Northrop margins and FY2024 revenue rankings, DoD budget category shares, US share of GDP, overseas bases, warhead counts, Ukraine aid totals, China ship counts and the F-35 lifetime cost, Rheinmetall and other European Q3 results, and any congressional action after mid-August 2026.

Up 2.9% in real terms on 2024, the 11th consecutive annual rise (SIPRI, April 2026). Roughly 2.5% of global GDP now goes to arms, the highest share since 2009, and the fastest growth is in Europe (up 14% to $864 billion).

1. 1993The 'Last Supper': DoD tells contractors to merge
2. 1995Lockheed Martin: Lockheed + Martin Marietta
3. 1997Boeing + MDD: one commercial-defence giant
4. 2018Northrop + Orbital: vertical integration accelerates
5. 2020RTX is born: Raytheon + United Technologies
6. 2024New entrants: Anduril, SpaceX challenge the club

• Lockheed Martin: The purest-play defence prime. F-35, missiles, satellites. ~$71B revenue (2024).
• RTX (Raytheon): Missiles and radars (Patriot, missiles), plus commercial aero. ~$81B.
• Northrop Grumman: Bombers, space, autonomy. Builds the B-21 Raider. ~$41B.
• General Dynamics: Submarines, tanks, IT (Virginia-class subs, M1 Abrams). ~$48B.
• Boeing (Defense): Fighters, tankers, rotorcraft (F-15, KC-46). Troubled parent.

1. Win a multi-year program
2. Lock in decades of revenue
3. Fund lobbying & relationships
4. Help shape the next requirement
5. Out-position rivals on the next bid

At the centre of the loop sits incumbency.

1. Requirement: a service names a capability gap (the JCIDS process).
2. Budgeting: funded across a 5-year plan; Congress must appropriate.
3. Competition: an RFP is issued; contractors bid; a source is selected.
4. Development: engineering & manufacturing; prototypes and testing.
5. Production: low-rate, then full-rate, then decades of sustainment.

Between a working prototype and a funded program lies a valley of death that kills most defence startups.

The technology works. The problem is surviving the years until anyone will buy it at scale.

1. Prototypea working demo (garage prototype)
2. SBIR / pilotsmall study contracts
3. The valleyno bridge to scale
4. Full-rate productionfunded, at scale (program of record)

Acquisition plus ~66 years of sustainment for the US fleet, per GAO (2024). One program, larger than most nations' entire economies, is the clearest illustration of how cost compounds across a weapon's decades-long life.

• Requirements creep: the pursuit of a 100% solution delays the 80% one.
• Color of money: appropriations are rigid; funds can't flex to need.
• Oversight & compliance: FAR rules, audits and reviews add years.
• Incumbency bias: it is always 'safer' to buy the known supplier.
• Long horizons: a decade from specification to fielded system.
• Bid protests: losing bidders can legally freeze an award for months.

The next war may be won less by the best platform than by the best software running on cheap, expendable hardware.

That is a different industry, one that looks more like Silicon Valley than like an aircraft plant.

A two-axis map plots speed and software focus against incumbency and scale. The insurgents (Shield AI, Anduril, SpaceX, Palantir) cluster top-left as fast challengers; Palantir and SpaceX sit toward the scaled and modern edge. The primes (Northrop, RTX, Lockheed, General Dynamics) sit bottom-right as legacy primes.

• 05 Command & decision: AI-assisted targeting and planning (Maven, AIP)
• 04 Autonomy & software: the 'brains' (Lattice, Hivemind)
• 03 Data & networks: sensors and comms, the connective tissue
• 02 Platforms: jets, ships, drones, missiles, the 'iron'
• 01 Industrial base: factories, supply chains and materials

• Anduril: Autonomy and Lattice OS; sells products, not cost-plus services.
• SpaceX: Reusable launch and Starshield; collapsed the cost of orbit.
• Palantir: Software and data fusion (the Maven Smart System, AIP).
• Shield AI: AI pilots (Hivemind) that fly drones without GPS or comms.
• Saronic: Cheap autonomous surface vessels, mass at sea.
• Skydio & others: Small drones, built at commercial scale and speed.

A $500 drone can now destroy a multi-million-dollar tank, and that single fact inverts the economics of war.

When the cheap thing reliably kills the expensive thing, mass beats quality.

1. FPV drone~$500, expendable (attritable & cheap end)
2. Loitering munition~$50k, one-way
3. Cruise missile~$1-2M each
4. Fighter jet~$80-110M (exquisite & scarce end)

• Mass matters again: cheap, expendable systems can overwhelm expensive ones.
• Magazines run dry: peacetime stockpiles empty in weeks of real war.
• Software wins the duel: electronic warfare and counter-EW decide who sees whom.
• Iteration speed is a weapon: drone designs now turn over in weeks, not years.
• The base is the bottleneck: you cannot surge what your factories cannot build.

• Barriers to entry: clearances, plants and certifications take years and billions.
• A single customer: the government sets the rules, the price and the demand.
• Incumbency bias: buyers default to the supplier they already trust.
• The valley of death: no revenue bridge from prototype to production.
• Switching costs: decades of sustainment lock a platform in once chosen.

The return of great-power competition is the single force reshaping every defence budget on Earth.

Deterrence is back as the organising principle, and industry is being asked to prove it can deliver.

1. Does production actually scale? 155mm, missiles and drones: output, not announcements.
2. Can a disruptor reach true scale? The first insurgent to win a major program of record.
3. Does acquisition reform stick? Faster paths from prototype to fielding.
4. How fast does China build? The shipbuilding and missile gap is the pacing metric.
5. Where does the software margin land? Who captures the top of the stack, prime or insurgent?

The West spent thirty years buying the finest weapons ever built and forgetting how to build them quickly. The next war will be won not by whoever owns the best machine, but by whoever can make the most of them.

1. Spending is surging: a genuine arms era, the fastest build-up since the 1980s.
2. Five firms still rule: a protected oligopoly, efficient in peace, brittle in war.
3. Procurement is the constraint: cost-plus and the valley of death throttle speed.
4. Software is the high ground: value is moving from the platform to the code.
5. Capacity beats quality: Ukraine proved mass and industrial depth decide wars.

• SIPRI: Military Expenditure Database and Yearbook (2024-2025).
• US Department of Defense: Comptroller / National Defense Budget 'Green Book' (FY2024).
• GAO: Weapon Systems Annual Assessment; F-35 sustainment reports (2024).
• Defense News: Top 100 defense companies (2024).
• Company 10-K filings: Lockheed Martin, RTX, Northrop Grumman, General Dynamics, Boeing (FY2024).
• CSIS: 'Empty Bins' and defense-industrial base analyses (2023-2025).
• Sacra; PitchBook: Anduril and defense-tech venture data (2024).
• Figures marked 'estimate' are compilations of the above.