The Real Bill from the Hormuz Closure Threat — 9 AI Infrastructure Choke-Point Stocks

TL;DR: Iran rattled the Hormuz closure card again, and helium, copper, and natural gas all moved in the same week. The real winners aren't Nvidia — they're the nine "choke-point" companies sitting on top of it: Vistra, Eaton, Vertiv (power/cooling), Micron, Amkor, Broadcom, Marvell (silicon), Southern Copper, Corning (materials). The pattern is identical to 1849, when the people selling jeans got rich, not the gold panners.

Iran put the Strait of Hormuz closure back on the table this week. The first thing my mind jumped to wasn't oil prices — it was helium.

Qatar supplies more than 30% of the world's helium. Where does that helium go? Semiconductor fab cooling, optical fiber manufacturing, MRI, and most critically right now, AI chip packaging. If Hormuz closes, Qatari LNG and helium move together — and so does every downstream input.

Why this round of Hormuz threats is different

Hormuz threats aren't new. They've been recycled for two decades. But the 2026 round has two things going for it that prior rounds didn't.

First, the AI infrastructure capex cycle is approaching a peak. Global data center capex blew past $400B in 2025 alone, with every new build pulling on the same parts and the same raw materials simultaneously. One supply-side shock at this timing and prices spike hard.

Second, there are essentially no substitutes for helium, copper, or specialty minerals. If Hormuz tightens, Korean fabs pay up for helium while US fabs ride domestic reserves. That asymmetry rewrites the stock list.

The 1849 pattern, again

This is the oldest pattern in markets. When a constrained resource gets squeezed harder, the people holding it take everything.

During the California Gold Rush, tens of thousands rushed west. Most went broke. The man who built a fortune wasn't panning for gold — he was selling denim and shovels to those who were.

The 90s internet was the same. Cisco was the right call. But buying at the absolute peak meant 20 years just to break even.

Right now everyone is buying Nvidia. I get it. But the real money has never come from the obvious play. It comes from the company sitting on the physical constraint that the obvious play can't run without.

After mapping out the AI build-out for the past week, I count exactly nine of those constraints in 2026.

The 9 choke points, 9 stocks — quick scan

Constraint	Company	Key number
Baseload power	Vistra Corp	3,800MW contracted nuclear (2x nearest competitor)
Transformers/switchgear	Eaton	128–144 week lead times, 11-year backlog
Liquid cooling	Vertiv	Handles 132kW per rack, $15B backlog
HBM memory	Micron	21% of global HBM supply, sold out through 2026
Advanced packaging	Amkor	#2 CoWoS source after TSMC, packaging revenue tripling
Custom AI silicon	Broadcom	$73B AI backlog, 60–70% of custom chip market
Optical transceivers	Marvell	Chip inside ~63M 800G+ transceivers shipping in 2026
Copper	Southern Copper	51.1M ton reserves, $0.42/lb production cost
Optical fiber	Corning	$6B Meta deal, but supply only grows 11–19%

The two asymmetric beneficiaries from Hormuz itself

The table reads like a regular AI shopping list. The Hormuz variable bends two of these names asymmetrically in their favor.

Micron: SK Hynix and Samsung run Korean fabs with high Qatari helium dependency. Micron runs US fabs with domestic helium access. If Korean input costs climb, Micron picks up one of two outcomes — share gain (because Korean producers cut output) or market-wide price expansion. Either result expands its margins.

Vistra: When natural gas prices rise, gas-plant margins compress. Nuclear has fixed cost structure, so the more gas climbs, the wider Vistra's relative margins open. The 20-year contracts with Meta and AWS are tied to market rates, so cost increases pass straight to revenue.

I'd argue these two stocks aren't just AI plays — they're a hedge against the Hormuz scenario itself.

Do you need all 9? — my view

No. And honestly, "buy all 9" was never the point of the exercise.

The real value of this list is breaking out of the "I bought Nvidia, I have AI exposure" mindset. Nvidia depends on HBM in its chips, CoWoS to package those chips, optical transceivers to connect them, electricity to run them, and copper to build the facility — every single one is a precondition for Nvidia even functioning.

In Topic 3 I'll walk through how I'd split these nine into three layers and size positions. For now, just having all nine in your head while reading the Hormuz news is enough.

What to watch

Whether or not Hormuz actually closes, the rising frequency of the threat itself matters more than any single closure event. Markets react more to "this risk is becoming chronic" perception shifts than to one-off events.

The three triggers I'm watching closest over the next 12–18 months: Qatari helium export volume, the rate of Korean semiconductor input cost inflation, and the average grid-connection wait for new US data center projects. If all three deteriorate in the same quarter, operating leverage at these nine names lifts together.

This is for educational purposes, not investment advice. But I'd suggest re-reading Hormuz news as an AI supply chain story — not just an oil story.