|
Part I
The Mechanism
|
Helium.
The financial press moved on three months ago. Samsung and SK Hynix signed long-term supply deals with Linde and Air Products back in April. CNBC ran the chyron: "Helium Crisis Easing." Bloomberg quoted an analyst saying the "acute phase" was over. If you were watching the headlines, you'd think the problem was solved.
It wasn't solved. It was redistributed.
Qatar's Ras Laffan complex — the single largest helium production site on Earth, responsible for roughly 33% of global supply — has been under force majeure since March. The helium comes out as a byproduct of LNG processing. No LNG, no helium. You can't call up Qatar and order just the helium. The Strait of Hormuz remains effectively closed, traffic down 95% since Iran declared it shut on March 2. And on April 14, Russia slapped export controls on helium with no clear end date. Between Qatar offline and Russia locked down, approximately 40% of global accessible supply vanished in six weeks.
What Samsung and SK Hynix actually did was outbid everyone else for the remaining supply. The deals didn't create new helium. They rearranged who gets rationed last. Hospitals running MRI machines, research labs, fiber optic manufacturers — they all moved down the priority list. The fabs moved up.
That's not a resolution. That's triage.
|
Part II
The Diagram
|
Story off. Numbers on.
Global helium production in 2025 was approximately 190 million cubic meters. Three countries — the United States, Qatar, and Russia — accounted for roughly 85% of that output. US at 42%. Qatar at 33%. Russia at 9.5%. That's not a supply chain. That's a three-legged stool, and two legs just snapped.
Helium is not like copper or lithium. You cannot mine it on demand. It accumulates over billions of years from radioactive decay deep underground, trapped in geological formations alongside natural gas. When you extract the gas, you can capture the helium as a byproduct — if you've built the purification infrastructure. If you haven't, or if the gas plant shuts down, the helium either stays underground or vents to atmosphere. Once vented, it's gone. Helium is the only element that permanently escapes Earth's gravity. It rises through the troposphere, passes the stratosphere, and leaves the planet. There is no recovering it from the sky.
The semiconductor industry consumes roughly 24% of global helium, projected to hit 30% by 2030. Every advanced chip requires helium at multiple fabrication stages: wafer backside cooling during lithography and etching, purging the optical path inside EUV scanners, leak detection in vacuum systems, carrier gas in plasma-enhanced deposition. A 2nm High-NA EUV wafer requires 1.2x more helium than a 3nm wafer. The more advanced the node, the more helium per chip.
TSMC alone consumes roughly 500,000 cubic feet per year across its most advanced fabs. Leading fabs run recycling systems that recapture 90–95% of helium used in certain processes. But recycling isn't perpetual motion. You lose 5–10% per cycle through leaks and process contamination. Fresh supply is non-negotiable.
South Korea is the single most exposed node. It sourced 64.7% of its helium imports from Qatar in 2025. Samsung and SK Hynix together produce more than two-thirds of the world's DRAM and nearly half of all NAND — chips flowing into every AI server, every smartphone, every data center. Fitch Ratings flagged South Korea as "among the most vulnerable" to helium supply disruption.
The long-term US supply deals bought time. They didn't buy independence.
|
Part III
The Weak Link
|
Here's the part nobody is modeling.
Every major economy is building new semiconductor fabs right now. TSMC in Arizona. Samsung in Taylor, Texas. Intel in Ohio. SK Hynix in Indiana. These are CHIPS Act projects, billions of dollars in subsidies, ribbon-cutting photo ops on the front page. Every single one of them needs helium. More helium, in fact — because they're all targeting advanced nodes where per-wafer helium consumption is higher than anything currently in production.
The construction timelines for these fabs run 2026–2028. The helium supply recovery timeline? Unknown. Qatar's Ras Laffan repairs are estimated at 3–5 years for the damaged trains. Gazprom's Amur plant — Russia's main helium source — is export-controlled with no clear end date. New pure-play projects like Helix Exploration's Rudyard field in Montana just started first production in February 2026. It's the first helium producer in Montana's history. It is a rounding error against global demand.
The US government used to have a buffer for exactly this scenario. The Federal Helium Reserve in Amarillo, Texas — built in the 1920s, once holding billions of cubic feet. Congress mandated its sale. The BLM completed the $460 million disposal to Messer in 2024. The strategic cushion that took decades to accumulate was auctioned off two years before the worst helium supply shock in history.
Samsung has deployed its in-house Helium Reuse System on select lines, with a full rollout projected to cut consumption by about 18.6% per year. That's real engineering. It's also a band-aid on an arterial bleed. You can optimize consumption all you want. At some point, you need molecules coming in the door. And the door is narrower than it's been in decades.
I've watched enough commodity squeezes to know what "managed allocation" looks like from the outside versus the inside. From the outside: "supply secured, crisis easing." From the inside: every purchase order is a negotiation, every contract renewal comes with a surcharge, and the guy who was your backup supplier just told you he's fully allocated through Q2 2027.
|
Part IV
The Chain Reaction
|
Follow the sequence.
Qatar stays offline through winter — which is the base case, since repair estimates run 3–5 years for the damaged trains. Russia stays export-controlled. US domestic production runs at capacity but can't cover the gap. Linde, Air Liquide, and Air Products continue allocation protocols, prioritizing long-term contract customers while capturing spot premiums on incremental volumes.
First crack: memory output. Samsung and SK Hynix hold roughly six months of helium inventory. That buys them until approximately Q1 2027 at current consumption rates. After that, any supply shortfall translates directly into reduced wafer starts. DRAM and NAND output drops. Not because demand fell. Because a noble gas that accounts for less than 1% of fab operating cost became unavailable.
Second crack: HBM and AI infrastructure. High-Bandwidth Memory is the most helium-intensive memory product in production — advanced packaging, multiple die stacks, tighter process tolerances. If total wafer throughput drops 10%, HBM doesn't escape the math. You can prioritize all you want. The molecules don't care about your product mix.
Third crack: the AI buildout timeline. Every hyperscaler — Microsoft, Google, Amazon, Meta — has committed billions to data center expansion predicated on GPU and HBM availability. If HBM gets constrained, GPU assembly slows. If GPU assembly slows, data center commissioning slips. The entire AI capex cycle has a hidden dependency on a gas that most investors have literally never thought about.
Where does capital position? Not in the broad semiconductor ETFs — those are diluted with fabless designers who don't touch helium and equipment makers whose revenue lags the constraint by two quarters. The edge, if there is one, sits in two places.
First: the industrial gas companies with locked-in helium supply and pricing power during allocation. Linde and Air Products are printing margin on every molecule. Their contract books just became the most valuable assets in the supply chain — not because the gas is expensive, but because they're the only ones who can deliver it.
Second: small-cap helium-leveraged names with actual production or near-term commissioning — Helix Exploration at Rudyard, a pure-play, and US Energy Corp in Montana, an integrated energy company with helium operations — whose helium thesis just got validated by a geopolitical shock they didn't cause and can't control. These are early-revenue, speculative names. If Qatar comes back online or the conflict resolves, they face real downside. Price the risk accordingly.
The consensus is that helium is a solved problem because two Korean chipmakers signed supply contracts. The physical market says it's a 40% supply deficit with no new major capacity for 3–5 years and a demand curve that only steepens from here. In my experience, when the contract lawyers say "secured" and the loading dock says "allocated," you follow the loading dock.
