The critical-minerals picture has sharpened materially this cycle: what the 13 June baseline treated as a single bottleneck (materials-supply/_agent/2026-06-13.md) now resolves into three mechanistically distinct sub-bottlenecks, each promoted to candidate status by converging evidence from adjacent scans. That resolution matters now because the energy-demand signal it feeds has become unambiguous — PJM's 2025/26 and 2026/27 base residual auctions both cleared at the FERC cap of $329.17/MW-day (energy-grid/_agent/2026-06-15.md), and NERC's 2025 LTRA is calling for a 69% uplift in US summer peak capacity over ten years against transformer lead times that have widened from six-to-nine months in 2015 to eighteen-to-twenty-four months today.
The regime bifurcation established in the June baseline remains the organising frame. Light-tier materials — lithium, Class-2 nickel, cobalt — are structurally oversupplied through Chinese and Indonesian capacity; spot lithium carbonate sits at roughly $10–12k/t against its Q4 2022 peak of $80k/t, and Albemarle (ALB) is the right calibration entry here precisely because it illustrates why not all picks-and-shovels earn the rent. The heavy and strategic tier is a different world entirely. China's April 2025 MOFCOM Decree 2025-18 added seven medium-and-heavy rare earths — samarium, gadolinium, terbium, dysprosium, lutetium, scandium, yttrium — to the export-licence regime (2026-06-13 primary chain), and the licence-grant rate to G7 buyers has been running at multi-month delays through 2025. Prior Chinese controls on gallium, germanium, antimony, and graphite (August 2023 through February 2025 dual-use controls, per chips-compute/_agent/2026-06-16.md) complete the picture: the export-control ratchet runs from compute-node restrictions at the digital layer to feedstock controls at the physical layer, and the two regimes are now explicitly coupled.
The US sovereign response has established a precedent I think the market is still underweighting. The July 2025 DoD preferred-equity investment in MP Materials (MP) — anchoring an NdPr Oxide price floor of approximately $110/kg alongside multi-decade offtake, with Apple providing the demand-side closure via the Independence, TX magnet line — is structurally analogous to WWII Reconstruction Finance Corporation intervention (2026-06-13). It reprices the reservation-value analysis for every US-domestic strategic-tier processor: the question is no longer whether sovereign support is available but which projects are next in the DPA Title III queue. Perpetua Resources (PPTA) and its Stibnite project — the only US-domestic antimony pure-play with a credible construction pathway — sits at the head of that queue, with applicability to flame-retardants, ammunition primers, and ball-bearing alloys making it the defence-modernisation crossover pick (defense-modernization/_agent/2026-06-04.md).
Three sub-bottlenecks are promoted to seed-YAML candidate status this cycle, each earning the label because it has now been independently surfaced by multiple adjacent-theme scans.
Heavy-rare-earth separation is the binding chemical step — solvent-extraction circuits qualified for Dy/Tb at production scale — sitting behind a five-to-eight-year permitting-and-ramp timeline. China holds more than 90% of global capacity. Lynas (LYC.AX) Kalgoorlie heavy-RE plant and Iluka (ILU.AX) Eneabba refinery are the only non-China commercial-scale capacity slated for 2026–2027 commissioning. The same Dy/Tb chokepoint that binds humanoid traction motors (robotics-humanoids/_agent/2026-05-29.md) also binds wind direct-drive generators, naval propulsion, and MRI magnets — the bottleneck is broader-spectrum than any single theme's framing warrants.
Grain-oriented electrical steel (GOES) is the transformer-content binding that directly transmits the PJM-at-cap signal into the physical layer. Cleveland-Cliffs (CLF) Butler Works is the US sole-domestic producer, making it a single-site failure risk against the NERC LTRA–implied transformer demand curve (energy-grid/_agent/2026-06-15.md). CLF appears as a shared cross-theme pick in both the materials-supply and energy-grid scans, which is exactly the kind of convergence that earns a name additional weight in the portfolio.
Antimony refining via the Stibnite pathway is the third promotion. Perpetua (PPTA) has no US-domestic operational-scale competitor; the DPA Title III–eligible designation closes the sovereign-backstop loop in the same way the MP precedent did for NdPr, and the defence-modernisation scan (2026-06-04) independently identified it as a sole-source risk.
Copper sits across all three sub-bottlenecks as the general transmission medium — grid copper demand compounds against Codelco, Las Bambas, and Cobre Panama production disappointments through 2025. Freeport-McMoRan (FCX) is the primary-copper name in the frame; this cycle adds Southern Copper (SCCO) to complete the duopoly representation. At the digital upstream, the gallium/germanium refining sub-tier — a by-product of zinc-aluminium smelting, not an incrementally scalable standalone process — sits at Layer 3 of the value-chain map and binds AI power electronics and defence simultaneously per chips-compute/_agent/2026-06-16.md. The C-26 rent-migration framing runs concurrently through the physical upstream (primary refining, by-product strategic-refining, component manufacturing) and the digital upstream (ASML, TSMC CoWoS, HBM oligopoly) — these are parallel expressions of the same dynamic, not competing framings.
On capital flows, sovereign money is the marginal funder: the Druckenmiller basket (C-25-capital-reallocation-regime) routes public-market capital into Layers 1–2 explicitly via Alcoa (AA), Barrick, and uranium names including Cameco (CCJ). The implied extension is COPX for copper and REMX for the rare-earth complex — REMX holds MP, LYC, ILU, and Chinese names and is not currently on the watchlist, which is an omission worth correcting given how cleanly it maps to the value-chain layer where rent is migrating.
What would change my mind
Three concrete observations would break this view. First, MP Materials' Q3 2026 10-Q magnet-shipment run-rate misses its Apple-plus-DoD offtake guide by more than 20%, signalling that the Layer 2 separation ramp is slipping rather than executing. Second, the PJM 2027/28 base residual auction clears below $200/MW-day, indicating that the demand-side signal has not transmitted into materials pull at the pace NERC's LTRA implies. Third, China lifts heavy-RE export licensing to G7 buyers in a form visible in the Federal Register, signalling a regime-change reversal that strands Western separation capex before it scales.
What I'm watching next
The highest-priority deferred observable is MP Materials' Q2 2026 10-Q (August 2026 filing window) for magnet-shipment run-rate against the DoD and Apple offtakes. Alongside that: Perpetua Stibnite construction milestones and DPA Title III award status; the USGS 2026 critical-minerals list refresh, which is statutorily due this year and expected to add bismuth, tellurium, and lutetium as candidate commodities; the Lynas Kalgoorlie and Iluka Eneabba commissioning reports; and IRA Section 45X survival through the FY2027 reconciliation tranche, which is the single policy variable most capable of permanently flattening the domestic-processing cost curve.