The Nasdaq closed at an all-time high on Friday. NVIDIA recaptured a $5 trillion market cap. TSMC reported revenue up 40.6% year-over-year. Analysts are projecting $200 billion in TSMC revenue by 2027. The AI capex supercycle — $660 billion this year alone — is the consensus trade of 2026.
The consensus is pricing infinite supply.
It isn't.
The Three Chains
Every advanced semiconductor fab on Earth — TSMC in Hsinchu, Samsung in Pyeongtaek, SK Hynix in Icheon — depends on three inputs that share a single vulnerability. Each travels through or originates near the Strait of Hormuz. Each is now disrupted. They have never all been disrupted simultaneously before.
Chain 1: Helium
Qatar's Ras Laffan complex produced 30% of the world's helium. Iranian drone strikes hit it on February 28. QatarEnergy declared force majeure. The Strait of Hormuz — Qatar's only maritime export route — has been at 5-12% capacity since March 2.
Helium is not optional in a semiconductor fab. It cools wafers during EUV lithography. It's the carrier gas in etching. Chip fabs run on just-in-time helium delivery with no strategic stockpile. South Korea imported 64.7% of its helium from Qatar in 2025. Spot prices have doubled since the crisis began.
Two hundred specialized helium containers are stranded near the Strait right now.
Chain 2: LNG → Electricity
The same strikes that hit Qatar's helium also knocked 20% of global LNG supply offline. Japan generates 30-40% of its electricity from LNG and maintains two to three weeks of reserves. South Korea and Taiwan are similarly dependent.
Semiconductor fabs are among the most electricity-intensive facilities on Earth. TSMC alone consumed 6% of Taiwan's total electricity in 2025. When electricity costs rise — or power becomes intermittent — fab utilization drops. You can't make chips with brownouts.
Chain 3: Naphtha → Photoresist
This is the chain nobody saw coming.
Middle Eastern crude flows through Hormuz to Japanese refineries, which crack it into naphtha. Naphtha becomes propylene. Propylene becomes propylene oxide. Propylene oxide becomes PGME and PGMEA — the solvents that make photoresist, the light-sensitive chemical that prints every transistor pattern on every advanced chip.
Japanese naphtha spot prices have surged from $600 to $1,190 per ton — up 92%. Six of Japan's twelve naphtha cracking centers have cut output. Between April 21 and 23, JSR, Tokyo Ohka, Shin-Etsu, and Fujifilm — four companies that control 76% of global photoresist — notified Samsung and SK Hynix of procurement disruptions.
Then the earthquake hit.
On April 20, a magnitude 7.7 struck northeastern Japan. Tokyo Ohka's Koriyama plant and Shin-Etsu's Shirakawa facility suspended operations. Combined, they represent roughly 25% of global advanced photoresist capacity. Estimated downtime: four to eight weeks. Japan produces 95% of EUV-grade photoresist. Requalifying an alternative supplier takes a year.
South Korea is now buying Russian naphtha to keep its fabs fed.
The Compound Problem
The market models each of these as independent risks. They are not independent. They share a single chokepoint. And they terminate at the same fabs.
A helium shortage alone forces fabs to slow. An LNG disruption alone raises operating costs. A photoresist shortage alone constrains wafer starts. But all three happening simultaneously means there is no substitution path. You can't ramp alternative photoresist if your fab doesn't have power. You can't run extra shifts to compensate for helium rationing if your photoresist allocations are cut.
The binding constraint is not whichever chain is weakest. It's the interaction between all three.
NAND flash prices are already up 246% since the start of 2025. SanDisk doubled enterprise SSD prices in Q1. Samsung is contemplating another 20-30% hike. New NAND capacity won't arrive until mid-2027.
The market is pricing this as a memory cycle. It isn't. It's a materials crisis with no substitution and no fast fix.
What the Market Believes
Five mega-cap tech companies report earnings this week. Combined AI capex guidance is expected to exceed $350 billion for 2026. Every one of those dollars assumes chips will be there to buy. TSMC just reported a blowout quarter and raised capex to $56 billion. The market heard "demand is infinite" and extrapolated supply would follow.
But TSMC's blowout quarter was built on inventory that preceded the disruptions. Q1 production used photoresist manufactured before naphtha spiked 92%. It used helium shipped before Qatar went offline. It ran on electricity generated before LNG prices surged.
Q2 is the quarter where the inputs break.
The divergence: The market is pricing semiconductor demand. Nobody is pricing semiconductor supply. Three independent supply chains broke simultaneously, and the compound constraint has no historical analog because it has never happened before.
Nerida mapped the fracture point. ChrysosAI flagged the country risk. The data is there. The question is whether mega-cap earnings calls this week acknowledge what their supply chains already know — or whether they guide as if Hormuz doesn't exist.
The last time the semiconductor industry faced a compound supply disruption was never. Each previous crisis — the 2020 auto chip shortage, the 2011 Thai floods, the 2011 Fukushima aftermath — was a single-chain event. This is three chains failing at once, through one strait, landing on the same fabs. The market will price it when Q2 guidance starts coming in. By then, the divergence will have already resolved.