Data: 2026-06-30T23:20:13.108410+00:00
Two parts of this are not new, and saying so up front is the point.
The concept, a risk with no historical precedent that therefore can't be priced, is Knightian uncertainty (Frank Knight, 1921). I'm using a standard term, not coining one; what's mine is the measured demonstration, not the idea.
The subject, Taiwan's energy exposure to a Chinese blockade, is a crowded 2025–26 field. The Foundation for Defense of Democracies has run tabletop exercises on China cutting Taiwan's LNG; the Center for Strategic and International Studies calls it Taiwan's "unique energy security challenge"; the Global Taiwan Institute, S&P Global, the Atlantic Council and the Australian Strategic Policy Institute have all published on the 11-day LNG stockpile, the non-reroutable nature of LNG under blockade, the nuclear phase-out worsening dependence, and supplier diversification as the standard fix. A crowded thesis means proven demand, not a closed question. But I'm citing it, not pretending to break it.
What I have that they don't is three specific things:
"Japan is the template for the next energy shock" is a cottage industry: Goldman, Bloomberg, Robeco, academics all run some version. The version worth paying for isn't the resemblance; it's the falsifiable one: freeze Japan's 2011 shock as a rule, test that rule on a second completed case, and only then project it forward to a country still exposed.
I built that test for energy. It does not hold. This post is the test in full, including the part that failed, because the failure is the most useful thing I can tell you about Taiwan.
The rule comes from Japan, 2011. Fukushima took nuclear (≈25% of generation) offline; Japan replaced it by importing more fossil fuel. Fossil generation share rose +23.3 pp (64.5% → 87.8%).
The mirror is Germany, 2022. The Russian gas cutoff removed a major energy input. If the rule transfers, fossil share should move the same direction. It didn't: Germany's fossil share fell −5.3 pp (47.9% → 42.6%). A sign flip, not a smaller version of the same move.
The price-clean check makes it worse, not better. Fuel-cost bills rose in both (Japan +51.7 indexed, Germany +15.9), but 2022 bills rose everywhere from a global price spike, so that line can't carry a structural read. The volume line, which is price-clean, also flips: Japan's fuel-import volume rose +24.1, Germany's fell −14.7.
Two of three structural metrics flip sign; the only one that agrees is the price-confounded one. By my own standard (freeze the rule, predict the mirror, show the gap), there is no directional fingerprint here. I'm reporting that rather than burying it, because a frozen claim that fails is the product, not an embarrassment.
The two events share a word ("energy shock") and nothing mechanical:
One is a supply-loss-met-by-importing-more event; the other is an import-loss-met-by-using-less event. They sit on opposite sides of the same ledger. An offset term can absorb a magnitude gap between two versions of the same mechanism. It cannot absorb a sign flip, because a sign flip means the mechanisms are different. My own rule for a valid fingerprint is "same mechanism, not just the same number." This pair fails that rule.
Now place Taiwan's modeled shock (a blockade or LNG cutoff) in the same matrix:
| Import access after shock | Adjustment margins | Completed precedent | |
|---|---|---|---|
| Japan 2011 | intact (could buy more fuel) | import more; price absorbs it | yes, the rule |
| Germany 2022 | channels open (Norway pipe, LNG, global coal) | cut demand · switch to coal · re-source gas | yes, the mirror |
| Taiwan (modeled) | cut off (sea lanes blocked) | rationing / blackouts only | none |
Taiwan has Germany's side of the ledger (it loses imports, not generation) but neither country's escape routes. Germany survived its import loss not by leaning on neighbours' electricity (it was a net power exporter in 2022) but because its supply channels stayed open: it cut gas demand by roughly a fifth, switched power generation back to coal, and re-sourced gas from Norway and LNG. Japan survived its generation loss by doing the thing a blockade prevents: importing more fuel. Taiwan under a blockade can do none of these: it can't re-source (every cargo comes by sea through the blocked water), can't lean on a grid neighbour (islanded), and isn't losing generation it can backfill. It's losing the imports that feed generation. What's left is demand destruction.
So Taiwan's exact cell, import cutoff + no interconnection, has no completed case in this study. I can borrow neither a magnitude nor even a direction. Under a true cutoff, fossil share could fall (rationing, blackouts) rather than rise; the Japan number points the wrong way, and the German mechanism (which did bend toward lower fossil use) got there through demand cuts and re-sourcing that a blockaded Taiwan can't replicate.
The absence of an analog is not a gap in the analysis. It is the result. A risk you can size against precedent is a risk you can hedge. A risk with no precedent in its own cell is one you have to watch directly and carry, which is worse, not better.
Taiwan names two triggers, and import-source concentration answers them differently:
The same vulnerability metric is decisive for one trigger and meaningless for the other. That itself is worth stating: don't let a "well-diversified suppliers" line reassure anyone about the blockade case.
Won't: a projected fossil-share number for Taiwan. The headline metric diverges and the mechanism doesn't transfer. (The "+46.8 to +56.8 pp if event" box is deleted: it was built on the metric that failed, via the mechanism that doesn't carry.)
Will: standing vulnerability, as measured fact:
This is the sell. Not the meters (published), not the backtest (free proof): this.
The backtest failed on the physical response (fossil share, volume). But the underlying shock (a sharp rise in the cost of imported energy) is common to all three, and that cost is real even when it doesn't show up the same way. Where it lands differs: in Japan it passed through to households and firms (visible in bills and consumer prices); in Germany it passed through to price too, briefly; in Taiwan it does not reach consumers: the state holds electricity prices frozen and lets the utility absorb the gap (Taipower took on the order of NT$350–400 billion in losses across 2022–2024, residential rates up only ~7% in three years while Korea, the UK and France raised theirs many times that). So Taiwan's true cost is hidden by subsidy. You cannot read its exposure off its own prices.
That is exactly what the Japan layer unmasks. Japan is the one advanced, import-dependent Asian economy where this class of shock was allowed to surface and was measured at fine grain: the 家計調査 (Family Income and Expenditure Survey) shows, month by month, which household income brackets absorbed the cost and by how much; sector fuel-cost data shows which industries took the margin hit. None of that exists for Taiwan. Its shock hasn't fired, and even its price signal is muted by design.
This is not a blockade forecast. It can't be, because no analog exists. It's the measured magnitude of the one channel that generalizes, on the one economy that recorded it.
Taiwan standing vulnerability (public, trackable, no event required):
These are direction-trackers, not event predictions. Movement is the signal. When one moves I re-read the fixed list and label it, I don't go hunting for a cause.
Procurement / supply-chain (lead): This is a tail you cannot size from precedent: there is no completed case in Taiwan's exact cell. That means you watch it directly rather than pricing it once. The meters above are the watch; the Japan cost-channel depth is what tells you how deep the damage runs if the price side fires.
Equity / utility fund (also exposed): Same conclusion, framed as positions carrying an unhedgeable supply-cutoff exposure. Taiwan's frozen prices make that exposure look smaller than it is (the bill is parked on the state utility, not the economy), so Japan's measured shock becomes a margin-exposure map for comparable Asian names whose true input-cost risk is currently masked.