Scarcity Portfolio: Long Copper, Power & Crashed Memory

I run a thematic long book built on one macro premise: physical scarcity is repricing faster than equities acknowledge. Four engines, ten positions, all catalyst-dated. Sharing the full logic and my actual allocation.

ENGINE 1 — COPPER: THE CLEANEST SCARCITY TRADE (FCX x15, SCCO x5)
LME copper cancelled warrants (metal already earmarked for withdrawal, so not really available) hit 62.4% of total stock — versus 4-18% for every other base metal. Live stock is down 71.8% from the April peak. Chile keeps losing supply (Caserones offline until mid/late August, Antofagasta H1 output -9.5%), and Morgan Stanley now forecasts a 600,000-tonne refined deficit for 2026 — the largest in over 20 years. Deficit estimates have widened three times in a row: 150kt → 520kt → 600kt. Every indicator points one direction.

ENGINE 2 — POWER SCARCITY (VST x7, CEG x4, TLN x3, GEV x1)
PJM (the grid operator for 13 US states / 65M people) formally came up 6.8 GW short of its reliability target; a backstop auction runs Sept 30–Oct 21 — a dated catalyst for generators. Datacenters drove 38% ($6.3B) of the $16.4B capacity auction cost. Meanwhile ~40% of new US datacenter capacity is projected to go off-grid by 2030 (Google discloses interconnection delays up to 12 years), which is the GEV leg: someone has to build those gas turbines. Europe just joined the theme — Hungary's Paks nuclear plant dropped to ~25% output on record-low Danube cooling water.

ENGINE 3 — THE MEMORY BIFURCATION (MU x1, MSFT x2, AMZN x4)
The AI-chip complex lost over $1 trillion in market cap in the week of July 28; MU trades ~37% below its June high. Yet the physical market tightened in the same week: Micron guides next quarter to $50B revenue at ~86% gross margin, HBM (high-bandwidth memory used in AI accelerators) is fully booked through 2027, Samsung guides the shortage through 2028, and Q3 DRAM contract prices are forecast +13-18%. Price crashed; product is sold out. I'm buying the gap, sized small because a third China-capacity headline (CXMT) can extend the drawdown. MSFT and AMZN are the demand underwrite — both held/raised capex (~$175B and $200B for 2026) and were rewarded for it, unlike GOOGL/META.

ENGINE 4 — RARE EARTHS (MP x21)
China shipped ZERO gallium/dysprosium/terbium/yttrium to Japan in June — a full-month embargo posture against a G7 economy. NdPr alloy is +21.4% in a month. Two hard dates ahead: Nov 10 and Nov 27, when China's export-control suspensions expire. MP is the ex-China supply answer, and my largest share-count position on purpose.

CATALYST CALENDAR
- Aug 19: 50% US tariffs on Canadian autos/alcohol/dairy take effect (first Section 338 use since 1949)
- Aug 26: NVDA earnings — the referendum on the whole AI-chip crash
- Sept 30–Oct 21: PJM backstop auction (VST/CEG/TLN catalyst)
- Nov 10 / Nov 27: China export-control cliffs (MP catalyst)

RISKS
US-Iran de-escalation holding would ease the inflation/scarcity backdrop broadly; a CXMT headline can hit MU another leg down; power-scarcity politics (price caps, moratoriums) are a two-sided risk for generators; MP is priced for confrontation — a genuine China trade détente hurts it.

Positions disclosed above are my actual holdings. Not financial advice — do your own research.