Morning Analysis — 2026-06-04

Dek

Oil cracks on a Hormuz truce — but copper, uranium, rare earths and steel all slide with it, a tell that the AI-buildout trade is de-risking its money, not its demand.

The Big Shift

The big overnight macro move is Hormuz de-escalation: the UK and France are finalizing a mine-clearing mission within days of a US–Iran deal to reopen the strait, and WTI crude fell 3.9% to $92.41 (¹). The tell isn't the oil drop — it's that copper, uranium, rare earth and steel ETFs all fell *together* while gold rose 1.4%. When fundamentally bullish metals sell off in lockstep with a fading war premium and gold bids up, that's a financing-layer repricing, not demand erosion. It points to the AI-infrastructure bottleneck shifting from *building* to *funding* — which leaves demand-floor theses intact and the most leveraged links most exposed.

Analysis

Geopolitics stays a tail risk, not a base case. The truce is real but fragile — Trump says the blockade "may last all summer," and the second-order moves haven't unwound: coal sits at a 14-year high, aluminum at a four-year high, and Henry Hub gas rose another 2.4% to $3.33 (²). For our desk this doesn't touch a tracked thesis directly, but firm gas is a quiet tailwind for the gas-and-nuclear power names (VST, CEG, GEV) that feed the data-center buildout.

Materials: the contradiction is the signal. Goldman *cut* its copper supply forecast by 350kt and lifted its price target above $13,700/ton — textbook bullish — yet copper miners (COPX) closed flat and the broader metals complex fell (³). Bullish fundamentals failing to lift the tape confirms the read: this is money de-risking, not a demand call. The copper-deficit thesis (FCX/SCCO) is confirmed but priced.

The one thesis that actually moved is MP, and it moved the wrong way. USA Rare Earth finalized — not announced, *finalized* — definitive agreements unlocking $1.6B from Commerce, taking it to ~$3.5B total capital, plus a $1.2B South Carolina magnet plant (⁴). That's a US competitor crossing from "planned" to "funded" on integrated mine-to-magnet — the core of MP's "only integrated Western supplier" pillar. The headline is stale and recurring, so surprise is low, but the *state-change* (funded) is the accumulating disconfirmer on our lowest-conviction major. Conviction down.

Compute and power: the demand floor holds, and the bottleneck is now cash. Goldman's $5.3T hyperscaler capex call and the $21B/3GW AirTrunk–Maharashtra deal keep the demand floor where it was (⁵). Two structural threads keep strengthening: behind-the-meter power as the default plan (Google funding a 100MW virtual power plant in PJM, Waymo's retired batteries becoming grid storage in California and Texas — ⁶), and the equipment layer re-rating as a block (VRT filed an 8-K today; ETN/PWR/GEV alongside). Both say the same thing: power and financing are the twin chokepoints, and the picks-and-shovels demand sits *below* the financing wobble.

The sharp end is the crypto-to-AI pivot. Former BTC miners turned AI hosts (WULF, IREN) are financing the rebuild on fresh debt — IREN filed a material-debt 8-K on 6/01. If today's de-risking is the start of a financing repricing rather than a one-day move, this is the link that breaks first, because it's funded on the most leverage and the least cash flow.

What Would Prove Us Wrong

• Demand erosion, not de-risking. If the metals slide continues *and* any hyperscaler trims forward capex guidance or cancels a named project this week, the "demand floor holds" read fails — hitting the equipment-block (VRT/ETN/PWR/GEV) and on-site-power theses, not just the miners.
• MP fully breaks. USA Rare Earth (or another US entrant) announcing *qualified, volume* NdFeB magnet shipments to a Western OEM — not capacity, deliveries — would shatter MP's integrated-supplier pillar outright, beyond today's funding milestone.
• Behind-the-meter reverses. A rollback of a state grid-cost rule (e.g. Pennsylvania's) or a flagship hyperscaler campus reverting to grid interconnection instead of on-site generation would undercut the strengthening behind-the-meter-power thesis.

Thesis Impact

MP | Conviction: DOWN | Surprise: LOW | USA Rare Earth *finalized* definitive agreements unlocking $1.6B from Commerce (now ~$3.5B total capital) plus a $1.2B South Carolina magnet plant — a US competitor crossing from "planned" to "funded" on integrated mine-to-magnet. CONTRADICTS MP pillar 1 ("only integrated Western REE→magnet supplier"). Recurring/headline-stale so surprise is LOW, but the state-change (funded, not just announced) is the accumulating disconfirmer worth not missing on our lowest-conviction major. | ⁴

No other thesis-moving signal. Everything else material is recurring or already in the prior: Goldman copper supply cut + copper +0.8% (FCX/SCCO — CONFIRMS deficit, priced); TMI/Crane FERC capacity-rights waiver (CEG pillar 3 — CONFIRMS, recurring); Goldman $5.3T capex + AirTrunk 3GW Maharashtra (ETN/VRT/Own-the-Bottleneck — CONFIRMS demand floor, expected). The one genuinely new macro event — Hormuz de-escalation (UK/France mine-clearing, US-Iran reopen, WTI −3.9%) — doesn't touch a tracked thesis.

Inflection Radar

[dismissive] Coal Plant Obsolescence | DOE commentary suggests older, politically maintained assets are functionally obsolete, signaling a potential regulatory/asset write-down cycle in grid planning. | Touches: NEW | ⁷

[emergent] AI for Nuclear Physics | Applying advanced ML (Inverse Critical Experiment Design) to validate next-gen reactor designs, moving AI from general compute to core infrastructure science. | Touches: NEW | ⁸

[emergent] Strategic Material Scarcity | CFR highlights the critical national vulnerability in securing strategic materials, suggesting policy/procurement bottlenecks are forming before major geopolitical events. | Touches: NEW | ⁹

[dismissive] DOE Rebate Constraint Shift | DOE narrowing appliance rebates to focus only on upgrading *existing* electric equipment, effectively capping the scope of immediate, large-scale electrification stimulus. | Touches: NEW | ¹⁰

Watch Signal:

[emergent] Voice AI in Under-served Markets | Founders leaving major tech/finance firms to build specialized voice AI stacks for non-English, low-resource markets, signaling a shift in AI deployment focus. | Touches: NEW | ¹¹

QA & Caveats

• The main call is low surprise, despite the funding change, as the core event is already headline news.
• The contradiction with Pillar 1 is noted, but the evidence provided does not offer a strong counter-argument to the funding reality.

Sources

1. UK and France Finalize Postwar Hormuz Mine-Clearing Mission bloomberg.com
2. Hormuz shock drives coal to 14-year high northernminer.com
3. Copper price: Goldman, Citi make bullish calls on supply woes mining.com
4. USA Rare Earth finalizes $1.6B with Commerce Dept northernminer.com
5. Government of Maharashtra, India, signs LoI with AirTrunk for $21bn data center deal datacenterdynamics.com
6. Waymo’s retired robotaxi batteries are heading back to work electrek.co
7. Indiana coal plant that Trump forced to stay open is not operating. DOE says it is key for reliability, but repairs mean reddit.com
8. Inverse Critical Experiment Design via Gradient Optimization and a Multigroup Attention-Based Neural Network Architectur arxiv.org
9. The Stockpile Gap: How America Can Secure the Strategic Materials It Needs to Win - Council on Foreign Relations news.google.com
10. DOE issues guidance ending gas-to-electric appliance rebates utilitydive.com
11. These two founders left Goldman and Meta to build voice AI for markets everyone else overlooked techcrunch.com