Iran declared Hormuz “completely open” this morning. Brent crashed 11% to $88. Markets priced in recovery. But the chain that matters most — the one connecting every AI chip on Earth to a liquefied natural gas terminal on an island in the western Pacific — that chain’s constraint was never the strait itself. The strait just made it visible.
The Chain
Trace it backward from the thing everyone wants:
Five layers. One thread. Every AI model trained, every chip fabricated, every inference served — ultimately depends on gas-fired electricity on an island with 11 days of fuel reserves, sourced partly through a strait that was closed for six weeks and “opened” this morning under conditions that could evaporate in ten days.
The Nuclear Hole
This is the part that makes the ceiling structural, not cyclical.
In May 2025, Taiwan shut down Maanshan Unit 2 — the last operating nuclear reactor on the island. The ruling DPP had committed to a nuclear-free homeland by 2025, and they delivered. Taiwan went from 16% nuclear to zero in three years.
That capacity had to go somewhere. It went to gas.
Taiwan’s gas dependency for electricity rose from ~38% in 2020 to ~48% in 2026. Every percentage point of nuclear lost was a percentage point of import dependency gained. And unlike Japan or South Korea, Taiwan has no pipeline connections to any other country. Every molecule of gas arrives by ship.
In March 2026, a restart application was filed for Maanshan. Regulatory review, refueling, and recommissioning will take 18–24 months. The earliest possible return: late 2027, realistically 2028. That’s two summers of maximum vulnerability.
Meanwhile, TSMC’s electricity consumption doubled between 2016 and 2023. The company now uses 8–9% of Taiwan’s total grid output. By 2030, forecasts put it at 24%. Every new 2nm fab is a small city’s worth of power demand added to a grid that just lost its most reliable baseload source.
The Qatar Thread
Of Taiwan’s LNG sources, approximately 25–30% comes from Qatar. Every Qatari LNG cargo transits the Strait of Hormuz — there is no alternative route. Qatar sits inside the Persian Gulf. Its gas leaves through Hormuz or it doesn’t leave.
Australia (~30%) is safe. The US share (~10%) is growing — Cheniere’s 1.2 million ton/year supply contract begins June 2026, eventually building to 25–33% of Taiwan’s imports. But that transition takes years, and summer 2026 demand peaks before the new supply ramps.
| Source | Share | Hormuz-exposed? | Status |
|---|---|---|---|
| Qatar | ~25–30% | Yes — only route | Disrupted 6 weeks. “Open” today, conditional |
| Australia | ~30% | No | Flowing normally |
| United States | ~10% | No | Cheniere 1.2M t/yr starts Jun 2026 |
| Indonesia / Malaysia | ~10–15% | No | Stable but limited growth |
| Other / spot | ~15–20% | Mixed | Spot market volatile, premiums elevated |
The arithmetic is simple: lose 25–30% of LNG supply, on an island with 11 days of reserves, where 48% of electricity comes from gas, and you don’t have a price problem. You have a rationing problem. CPC Corp, Taiwan’s state oil company, secured supply through April. Summer is unresolved.
Why the Strait Didn’t Create This
This is the structural insight that today’s price crash obscures.
Hormuz made Taiwan’s energy vulnerability visible. But it didn’t create it. Three decisions created it:
- Nuclear phase-out — Removing 16% of baseload generation with no equivalent replacement
- Gas dependency — Filling the gap with imported LNG rather than accelerating renewables or storage
- No strategic reserves — Maintaining 11 days of LNG inventory while Japan holds 36 and South Korea holds 43
Each of these predates the Hormuz crisis. Each persists after it. Even if Iran’s opening holds, even if the mines get cleared, even if insurance resumes, even if the ceasefire extends — Taiwan’s grid still runs on imported gas, the nuclear hole still exists, and every new TSMC fab still tightens the same ceiling.
The next strait closure. The next geopolitical crisis in the Gulf. The next Qatari contract renegotiation. The next summer heatwave that spikes Taiwan’s cooling demand. Any of these hits the same constraint. And that constraint is getting tighter, not looser, because TSMC’s power consumption is growing faster than Taiwan’s ability to diversify supply.
The Acceleration Trap
Here is what makes this a supply chain problem and not merely an energy policy problem:
AI demand is unbounded — every major hyperscaler is building out as fast as TSMC can supply. TSMC’s $56 billion capex this year isn’t cautious expansion. It’s a sprint. Revenue grew 41% year-over-year in Q1. Advanced nodes are sold out to 2027.
But every fab TSMC builds on Taiwan tightens the energy ceiling. By 2030, if current trajectories hold, TSMC alone will consume nearly a quarter of Taiwan’s electricity. The gas demand that implies, the LNG import volumes required, the shipping routes that must stay open, the geopolitical stability required in the Gulf — the chain of dependencies grows longer and more fragile with every percentage point of market share TSMC captures.
TSMC’s growth is constrained not by demand, not by technology, not by capital — but by the kilowatt-hours available on an island that imports 97% of its energy and just eliminated its only fuel source that didn’t arrive by ship. The stronger AI demand gets, the harder this ceiling binds.
Arizona and Kumamoto help at the margins. TSMC’s US fabs will eventually handle some advanced production. Japan’s Kumamoto fab serves mature nodes. But the overwhelming majority of cutting-edge fabrication — 2nm, 3nm, the chips that define the AI era — stays in Taiwan. Geographic diversification is a decade-long project. The energy ceiling is binding now.
What Today’s “Opening” Actually Changes
Iran’s declaration this morning is conditional on a 10-day Lebanon ceasefire that has already seen violations. The US blockade on Iranian shipping continues. BIMCO is telling shipping companies to “consider avoiding the area.” Norwegian shipowners want clarity on mines before sending ships through.
For Qatar’s LNG cargoes specifically, even a full reopening triggers the same five-layer recovery sequence I mapped in post #10: mine clearance (3–12 weeks), insurance reinstatement (3–6 months), vessel scheduling and backlog processing, terminal readiness, and commercial trust rebuilding. An LNG carrier isn’t a tanker — cryogenic cargo on a ship worth $200M+ requires the highest insurance standards. Those standards don’t return on a press release.
The first Qatari LNG cargo to reach Taiwan after today’s announcement — assuming everything goes right — is weeks away. And Taiwan’s reserves are measured in days.
The Real Map
The semiconductor supply chain has been mapped a thousand times. Fabless → foundry → OSAT → end product. Everyone knows TSMC is the bottleneck. But the bottleneck has its own bottleneck, and that bottleneck has its own bottleneck:
Global AI demand
↓
TSMC advanced fabrication
↓
Taiwan’s power grid (48% gas, 0% nuclear)
↓
LNG tanker fleet & port infrastructure
↓
Qatar (25–30%) through Hormuz
↓
Geopolitical stability of the Persian Gulf
Every arrow is a single point of failure. No buffer exceeds two weeks at any layer. The chain is as strong as its weakest day.
Markets priced in recovery today. The chain hasn’t started recovering. And even when it does, the ceiling will still be there — lower, tighter, and binding harder with every fab TSMC breaks ground on.
Tickers to watch: TSM (direct grid exposure), LNG/GLNG (US LNG beneficiaries of Taiwan diversification), FSLR/ENPH (Taiwan renewable urgency), NVDA/AMD/AVGO/QCOM (fabless dependency on Taiwan grid stability). CPC Corp is not publicly traded. QatarEnergy is state-owned.