The FY2027 Presidential Budget Request — a $1.5 trillion national-defence basket described by CSIS as "the highest level of funding in a single fiscal year since World War II" — is the load-bearing development that makes this the right moment to write up the defence-modernisation thesis [https://www.csis.org/analysis/unpacking-15-trillion-fy-2027-defense-budget-topline]. What makes the signal structurally different from prior cycle reads is not the topline but the plumbing: $350 billion of that basket runs through reconciliation, bypassing the appropriations channel that has slipped Sentinel, Constellation, Columbia and PAC-3 schedules in prior cycles and replacing political arbitration with a procedural lock that durably advantages the sector.
The composition reinforces the argument. CSIS flags two explicit priorities: an industrial base "on a wartime footing" and the maritime industrial base — "shipyards, mariners, logistics networks, and commercial fleets." The reconciliation money routes to PAC-3 (Lockheed Martin), Stinger, AMRAAM and SM-6 (RTX), GMLRS (Lockheed Martin) and 155mm artillery (General Dynamics), framed around inventory depletion from "Operation Epic Fury" creating "a window of vulnerability until inventories return to pre-war levels." The dollar split across line items is not disclosed in the CSIS summary and must be fetched from the actual budget exhibits next scan; what is verified is the direction of capital. Defence now consumes 63% of total discretionary spending against a 50% prior average — every other discretionary category compresses around it.
The binding constraint is physical, not digital
The most important analytical move here is separating the materials-supply bottleneck — which is binding on the 2026–2028 production-rate question — from the chips-and-compute bottleneck, which is the longer-run capability gate and is rapidly closing via commercial silicon. Relax solid-rocket-motor capacity and NdFeB heavy-rare-earth magnet supply, and the modernisation story moves on units delivered. Relax compute, and you move capability but not bill-of-materials. The reconciliation money routes to the former.
Solid-rocket motors are the single most acute system-level constraint behind every PAC-3, GMLRS, Stinger and hypersonic glide-body production rate. The market is dominated by L3Harris — which absorbed Aerojet Rocketdyne in 2023 — and a handful of Northrop Grumman divisions, with ITAR and Class 1.3 propellant-siting regulations making this a near-duopoly structurally protected from new-entrant disruption on any relevant timeline. NdFeB magnets — specifically the dysprosium and terbium gating guidance-fin actuators, F-35 motors and precision-guided-munition seekers — are the parallel constraint, with China's near-monopoly on separation as the geopolitical chokepoint. MP Materials is the US-listed pure-play, and it surfaces across both this theme and the robotics-humanoids scan from 29 May 2026, which mapped the same Dy/Tb choke across humanoid and EV motor supply chains. That cross-theme read strengthens the case to tag MP for watchlist priority.
The shipyard layer sits alongside materials as the other multi-year irreplaceable bottleneck. Huntington Ingalls is the only US facility capable of building Ford-class carriers and holds roughly half of Virginia-class and Columbia-class submarine capacity; General Dynamics Electric Boat holds the other half. No policy lever shortens the roughly decade-long throughput cycle here. ATI supplies the speciality alloys and BWX Technologies occupies the naval-nuclear-certified fabrication slot — both sit upstream of the shipyard layer behind the same multi-year capex and permitting cycle that makes substitution impossible on the relevant horizon.
Further down the stack, Howmet and Mercury Systems occupy the picks-and-shovels layer — engine forgings and structural castings for Howmet, rugged compute and radar-processing for Mercury — selling into every prime regardless of which platform wins the next programme competition. Curtiss-Wright plays the same role in naval valves and nuclear instrumentation. This is to defence what GE Vernova is to the AI capex cycle: durable revenue behind the headline technology story. GE Vernova's announcement on 26 May 2026 that its HA gas-turbine fleet had passed four million commercial operating hours [https://www.gevernova.com/news/press-releases] is an indirect but useful confirmation of the energy-and-materials picks-and-shovels frame, given the LM2500/LM6000 naval-lineage know-how underpinning allied IGT exports.
The autonomy layer — Kratos (XQ-58 Valkyrie, hypersonic targets via MACH-TB) and AeroVironment (Switchblade loitering munitions, post-BlueHalo) in the public market, with Anduril, Shield AI and Saronic as private proxies — is highest-beta to political cycle and programme restructuring. DefenseScoop on 3 June 2026 reported DISA's Mission Partner Environment exercise, the Pentagon's contested-logistics AI push and a Transcom-briefed maritime drone demonstration at MOTCO [https://defensescoop.com/2026/06/03/], signalling that unmanned-surface-vessel and manned-unmanned teaming threads are operationally tested, not merely funded. Anduril's CCA Increment 1 lead position is structurally more durable than the broader autonomy budget line, but the Replicator tranche is appropriation-renewed rather than multi-year obligated — the procedural protection that strengthens the primes does not extend here.
One political-economy signal worth flagging: the Bankrate survey from December 2025 found 53% of Americans unable to cover a $1,000 emergency [https://www.bankrate.com/banking/savings/emergency-staying-report]. A 24% real discretionary defence increase consuming 63% of total discretionary is a guns-without-butter posture that would normally face political arbitration. The reconciliation lock forecloses that arbitration — and whether it survives a full electoral cycle is the most uncertain variable in the five-year view.
What would change my mind
Three developments would break this view: FY2027 reconciliation language being stripped or materially cut in conference, which re-exposes the full programme to CR-driven schedule slip; a solid-rocket-motor delivery slip propagating into a quarterly miss disclosed in Lockheed Martin's or RTX's MD&A; or a Dy/Tb export-quota cut from China spiking NdFeB magnet input cost more than 25% before Allied refining capacity — MP Materials' Independence ramp, Lynas's heavy-rare-earth separation line and Iluka's Eneabba project — is operational.
What I'm watching next
The immediate task is pulling the actual FY2027 budget exhibits to verify the dollar split across PAC-3, GMLRS, SM-6 and 155mm line items — the CSIS topline is verified; the composition is inferring. Beyond that: MP Materials' DoD-disclosed magnet-output milestones at Independence as the leading indicator on whether the NdFeB chokepoint softens on the relevant timeline; the SECNAV shipyard workforce quarterly milestones as the leading indicator on Virginia and Columbia delivery cadence; and the Anduril IPO read-through — or absence of one — as the clearest market signal on how the autonomy layer's dependence on appropriation-cycle renewal, rather than multi-year obligation, is being priced.