Evidence current to 2026-08-12 — 2 days ago Every present-tense claim below is present tense as of 2026-08-12, not as of reading. Built from docs/hormuz/data/countries.geojson.
Hormuz Energy · Infrastructure Dependency Atlas

Where a Hormuz shock leaves the least room for error

The Philippines declared a national energy emergency with forty-five days of fuel in reserve. Thailand, at ninety-five, has ordered civil servants to take the stairs. Vietnam publishes a margin of twenty to forty-five days and is buying Russian crude. Those are the countries that publish a number at all — and across the twelve, the ones whose condition reads worst are the ones with no number to read.
12
countries assessed
7
publish a reserve figure
5
publish none at all
2
of the seven count something else
EVIDENCE WINDOW OPENS 2026-01-01 RECEIPTS 31 COUNTRIES 12
observed  ·  derived  ·  interpreted  ·  refused  — marks appear beside claims; defined under Method.
Published margins · observed
Two countries publish a margin entirely under sixty days. One straddles the line.
Days of national reserve, for the five countries whose published figure counts that quantity. Where the source gives a range, the range is drawn; the tick marks the single value the corpus stores.
Vietnam
20–45 d
source range; corpus stores 30
India
25–74 d
source range; corpus stores 50
Philippines
45 d
Thailand
95 d
Japan
254–260 d
source range; corpus stores 257
0 60 120 180 240
Published — but counting something else
These have a number. It is not the number above, so it is not drawn against that axis.
South Korea210 d Comprehensive data; 210 days strategic oil but only 9 days working LNG inventory
Singapore245 d Excellent data; 245 days reserves (includes foreign-owned storage); 95% gas generation is the exposure
WINDOW 2026-01-01 → 2026-08-12 n 7 of 12 publish a figure — the other 5 are below UNIT days

Reserve depth measures how long a country can absorb a shortfall, not whether it can distribute what it has. The Philippines declared an emergency at 45 days; Thailand, at 95, has not. Those are not two points on one scale.

▌Interpreted
The other five

The missing bar is the bar

Five of the twelve publish no reserve figure. This is what the corpus records them doing instead — their own current_status, verbatim.
Sri Lanka
no reserve figure
Current status · as of 2026-08-12Fuel rationing tightened into a QR-code system capping weekly purchases at 15 litres for cars and 60 for buses. Fuel prices raised about 25% — the second increase within a week. Four-day working week, with Wednesdays a public holiday for most sectors. Government appeals for fuel and electricity conservation.
Bangladesh
no reserve figure
Current status · as of 2026-08-12Fuel rationing in force and universities closed. Military deployed for anti-hoarding enforcement earlier in the crisis. LNG costs continue to force load-shedding, and fertiliser plants have been shut to conserve gas.
Pakistan
no reserve figure
Current status · as of 2026-08-12Moved to a four-day working week. Universities closed and fertiliser plants shut earlier in the crisis. Government continues to argue that roughly 74% domestic power generation cushions the LNG exposure, while imported fuel remains critical for transport.
Taiwan
no reserve figure
Current status · as of 2026-03-26No public emergency declarations. Decommissioned nuclear fleet (restarting planned). 70% of crude from Middle East. Reserve data not publicly available.
Cuba
⊘ not a hormuz case
Current status · as of 2026-03-26Repeated national grid collapses March 2026. Russia shipping fuel as humanitarian aid. US cut off Venezuelan oil supply to Cuba. Generation shortfall 1,300-1,700 MW during peak (half of demand).
OBSERVED status text and the absence of a reserve figure, per-country receipts WINDOW 2026-01-01 → 2026-08-12
Military at the fuel depots, universities shut, queues reintroduced — and no figure to rank any of it by.

Cuba disclaims membership in its own entry, and is kept because the corpus keeps it — marked refused rather than quietly dropped. An atlas that silently drops its awkward row is not showing you the corpus.

▌Interpreted
Secondary evidence · interpreted
The classification, mapped. Nothing in this shape is measured.
Fill encodes the corpus's stress_band — an authored assessment class with no derivation rule (see Method). The colours are categorical, not a scale: the classes describe different kinds of thing, and nothing in the corpus ranks them. Hatching marks the countries with no published reserve figure.
Philippines Sri Lanka Bangladesh Thailand South Korea Vietnam India Japan Singapore Pakistan Taiwan ⊘ CUBA — 220° AWAY “Not a Hormuz case.” ASSESSMENT CLASS — CATEGORICAL Already stressed Acute if disruption persists Price shock first Resilient but expensive no reserve figure
n 11 in frame · Cuba inset
INTERPRETED every fill. Shade encodes an authored band; neither it nor land area encodes a measured quantity. SUBSTRATE national boundaries are reference geography, not evidence — they carry no claim (source in Method). OBSERVED only the hatch — whether a reserve figure exists.
Qualification

What each published figure counts

The receipts for the split above. The corpus stores one integer per country; each country's own confidence note says what that integer is.
CountryStoredWhat the corpus says about that figure — verbatimTreated as
Philippines45 d Declared national energy emergency March 2026; public reserve data available on the axis
Thailand95 d Good reserve data; active government conservation measures documented on the axis
South Korea210 d Comprehensive data; 210 days strategic oil but only 9 days working LNG inventory off the axis
Vietnam30 d Generation mix clear; reserve data weak (20-45 days range); actively pursuing Russian alternatives range
India50 d Extensive official data; reserve estimates vary (25-74 days); diversification actively underway range
Japan257 d Best data coverage of any country; 254-260 days of reserves; comprehensive IEA/EIA profiles range
Singapore245 d Excellent data; 245 days reserves (includes foreign-owned storage); 95% gas generation is the exposure off the axis
DERIVED the flag. A figure is flagged when its own confidence note contains an explicit day-range, or names a basis narrower than the label — “strategic oil”, “working LNG”, “foreign-owned”, “estimates vary”, “data weak”. Applied mechanically, no per-row judgment. 5 on the axis (3 as ranges), 2 off it.

South Korea's 210 days is strategic oil; its working LNG inventory is nine. None of this is new research — every word was already in the corpus, in a field the published map renders as a footnote.

▌Interpreted · not evidence
The inputs that do exist put the unmeasured countries in the worst corner.
Share of oil arriving through the strait against fossil share of generation — the two dimensions published for nearly everyone. Dashed rings have no reserve figure.
≥75% imported · ≥70% fossil Philippines Sri Lanka Bangladesh Thailand South Korea Vietnam India Japan Singapore Pakistan Taiwan 0% Share of oil imports through the strait → 100% Fossil share of generation → reserve figure published no reserve figure
n 11 of 12 — Cuba has no recorded import share OBSERVED both axes
DENOM oil imports by volume; generation by share of output DERIVED the shaded quadrant, drawn for reading only
Country evidence

The twelve

Each claim as the corpus states it, against what is measured behind it. Every band is an authored judgment.
CountryInterpreted bandConf.OilLNGReserveFossil
Sri Lanka Already stressed medium 85%nonenone45%
Already rationing fuel with no public reserve data — Hormuz is an accelerant on existing fragility. Hydro 37%, solar 14%, coal/oil 45% If it persists: Minimal. Small domestic hydro base helps but insufficient. Emergency purchases at extreme premiums. Capacity to ration: Already rationing. Prior 2022 economic crisis built institutional muscle for austerity measures but population tolerance is low.Active crisis well-documented; reserve data limited; LNG import data sparse
Bangladesh Already stressed medium 95%99%none89%
99% of LNG from Qatar/UAE, military deployed for fuel hoarding, universities shut — no reserve data public. Gas 47%, coal 30%, oil 12%, renewables 4% If it persists: Near zero. 99% of LNG from Qatar and UAE. No pipeline alternatives. Domestic gas production declining. Capacity to ration: Low. Military enforcement suggests political stress. History of grid instability and load-shedding even before crisis.Active crisis documented; reserve data unavailable; 99% of LNG from Qatar/UAE
Philippines Already stressed high 92%none45 d76%
92% of oil from the Gulf, 45 days of reserves, energy emergency already declared. Coal 53%, gas 18%, renewables 12%, oil 5% If it persists: Limited. No pipeline alternatives. Attempting emergency spot purchases. Domestic coal dominant but imported. Capacity to ration: Low. Limited state capacity to manage prolonged shortage. History of grid stress during typhoon season.
Pakistan Acute if disruption persists low 95%99%none49%
Government says 74% domestic power sources, but 99% of LNG from Gulf, no public reserve data, universities already shut. Gas 25%, hydro 19%, coal 14%, nuclear 13%, renewables 13%, oil 10% If it persists: More diversified generation than it appears (hydro 19%, nuclear 13%). But 99% of LNG from Qatar/UAE. Domestic gas declining. No strategic reserve data public. Capacity to ration: Low. History of political instability and grid management challenges. Already experiencing rolling blackouts in many regions pre-crisis.Reserve data unavailable; conflicting reports on domestic vs import dependency; 99% of LNG from Gulf
Taiwan Acute if disruption persists low 70%nonenone85%
85% fossil generation, no public reserve data, decommissioned nuclear — the data gap is the red flag. Gas 42%, coal 39%, nuclear 4%, renewables 12% If it persists: Limited. Decommissioned nuclear (restart planned). Coal stockpiles provide some buffer. But 42% gas generation with unclear LNG source diversification. Capacity to ration: High state capacity. But the absence of public reserve data is itself a signal — either reserves are thin or disclosure is politically sensitive.Generation mix clear; reserve data essentially unavailable publicly; no crisis reporting yet
Vietnam Acute if disruption persists medium 90%49%30 d61%
Coal and hydro provide a generation buffer, but 90% of oil is from the Gulf and reserves are thin at 20-45 days. Coal 52%, hydro 31%, gas 6%, solar/wind 8% If it persists: Pursuing Russian alternatives. Domestic coal and hydro provide buffer. LNG dependency still small. Emergency procurement of 4M barrels non-Middle-Eastern oil (~6 days). Capacity to ration: Moderate. Coal and hydro provide structural resilience. The vulnerability is oil for transport and industry, not electricity generation.
Thailand Acute if disruption persists high 80%28%95 d80%
95 days of reserves and active conservation measures, but 55% gas-fired generation is the structural exposure. Gas 55%, coal 20%, renewables 15%, oil 5% If it persists: Limited. Some pipeline gas from Myanmar. Qatar supplies 28% of LNG (second-largest buyer). Exploring alternatives. Capacity to ration: Moderate. Government acting early with conservation measures. 95 days provides buffer but gas-heavy generation mix is the vulnerability.
South Korea Acute if disruption persists high 70%18%210 d74%
210 days of strategic oil reserves but only 9 days of working LNG inventory — the mismatch is the risk. Coal 42%, gas 27%, nuclear 18%, renewables 8% If it persists: Some. Nuclear provides 18% baseload. Coal stockpiles exist. But working LNG terminal inventory is only ~9 days despite 210-day strategic oil reserve. Capacity to ration: High state capacity. Institutional ability to enforce conservation. The gap between 210-day oil reserves and 9-day LNG working stock is the real vulnerability.
India Price shock first, physical shortage later high 45%53%50 d77%
Coal-dominant grid limits electricity exposure, but 88% oil import dependent and 90% of LPG through Hormuz. Coal 73%, renewables 12%, gas 4%, nuclear 3%, hydro 8% If it persists: Best positioned of the vulnerable countries. Coal-dominant generation reduces electricity exposure. Russian oil deals active. But 88% oil import dependent overall. Capacity to ration: High state capacity. Large domestic coal base insulates electricity. Primary risk is transport fuel and LPG for cooking, not grid collapse.
Japan Price shock first, physical shortage later high 77%7%257 d70%
257 days of reserves and strong state capacity — pain is price, not blackouts, unless disruption extends past 6 months. Coal 32%, gas 33%, nuclear 10%, renewables 20%, oil 5% If it persists: Best in Asia. Massive strategic reserves. Nuclear restart program. Diversified LNG sources. But volume dependency on Gulf crude is structurally high. Capacity to ration: Very high state capacity. Institutional experience from Fukushima energy conservation. Can sustain months of elevated prices before physical shortage.
Singapore Resilient but expensive high 70%45%245 d95%
95% gas generation and 45% of LNG from Qatar — money buys time, not immunity. Gas 95%, solar 2%, other 3% If it persists: Large storage capacity (245 days, though much is foreign-owned). Refining hub status provides some flexibility. But 95% gas generation is extreme single-fuel exposure. Capacity to ration: Very high state capacity and fiscal reserves. Can subsidize and purchase at premium for extended period. The question is whether global LNG supply physically exists at any price.
Cuba Special case: distorted substrate medium nonenonenone82%
Not a Hormuz case. Already inside the wreckage from US sanctions cutting Venezuelan oil. Hormuz amplifies via global prices. Fossil fuels 82%, renewables 18% If it persists: Venezuelan supply cut by US policy. Russian humanitarian shipments. No market alternatives at scale. Decrepit domestic generation infrastructure. Capacity to ration: Already beyond rationing into collapse. Not a Hormuz case — this is blockade + infrastructure decay + loss of benefactor supply. Hormuz is an amplifier through global price pressure.Grid collapse well-documented; not primarily Hormuz-driven; Venezuelan/US sanctions are the primary mechanism
AuditEverything below is the page checking itself: limits, method, and the source record.
Audit · what the marks mean
● Observed
Read off a cited source. Has a unit and a receipt.
◌ Derived
Computed here. Shows its rule; carries no receipt.
▌Interpreted
Authored judgment. Publishable — and not evidence.
⊘ Refused
Declined or unavailable. Shown, not omitted.
What this page does not claim. That the disruption will persist. That any country will ration, or when. That the assessments here are forecasts, rankings of suffering, or probabilities.
Audit · boundary conditions

What this page does not show

The five countries with no reserve figure are not the least exposed. They are the least measured.
Audit · method

How the bands were assigned

They were assigned by judgment. No derivation rule was ever specified for stress_band, and none can honestly be reconstructed.

The field was introduced in commit 1ecc1de (2026-03-29) and never edited. That commit names the inputs — “12 countries scored on 7 dimensions” — then enumerates which country landed in which band. It states no threshold, weight, or ordering rule; its own framing is “stress bands, not countdown clocks.” The repository's schema.yaml governs the U.S. case corpus only. No threshold table exists anywhere in the repository or its siblings.

Nor are the classes a scale. “Already stressed” is a present state, “acute if disruption persists” a conditional, “price shock first, physical shortage later” a sequencing claim, “resilient but expensive” a cost judgment. They describe different kinds of thing, and nothing in the corpus ranks them — so nothing here does. The map's colours are categorical for that reason, and an earlier draft of this page that ordered the classes and called two of them “severest” was wrong: that ranking was invented here, not found in the corpus.

A formula could be back-fitted to the twelve current values. That would be an invention presented as a method, so it has not been done. stress_band therefore renders as interpreted everywhere it appears — including on the map above, which is drawn in neutral ink for that reason.

The named inputs, so the judgment is at least inspectable: chokepoint oil share, chokepoint LNG share, reserve days, generation mix, current status, alternative supply, rationing capacity. Four of the twelve are missing at least one; Cuba is missing three.

The map's geography is not the corpus's. countries.geojson carries a hand-drawn outline per country — 9 to 31 vertices, roughly the right place, not the right shape. Rendering those implied a cartographic fidelity the corpus does not have, so they are no longer drawn. National boundaries here are vendored substrate — Natural Earth 1:50m Admin 0, v5.1.2, public domain (vendor/boundaries/PROVENANCE.md) — joined to the corpus by ISO country code at build time. They are reference geography: they carry no claim, and the build fails if a corpus country has no boundary rather than dropping it. The corpus polygons are left untouched in the data; they are simply not the render path.

Everything derived on this page. Two computations, both stating their rule where they appear: the evidence window (2026-01-01 → 2026-08-12) is min/max of all receipt dates, and the qualifier flag is the rule given under the reserve table. Everything else is presentational and encodes nothing: sort order, the 60-day line, the shaded quadrant, label placement on both figures, and the minimum mark size that keeps Singapore visible on the map — which exaggerates its area, one reason area is not an encoding there. Confidence is the corpus's own field, reproduced unchanged; it rates a country's inputs, never the band assignment.

Audit · receipts

Source material

All 31 citations, and what each supports.
No citation in this corpus has a retrieval date or an archive snapshot. Those fields do not exist in the schema yet, so nothing here records when a source was last seen or preserves it against removal. Stated once rather than repeated as an empty column.
SupportsPublisherPublished
Bangladesh The Conversation 2026-07-01
Bangladesh Reuters 2026-03-06
Cuba Reuters 2026-03-05
Cuba NPR 2026-03-26
India India Briefing 2026-03-15
India IEA 2026-01-01
India Lloyd's List Intelligence 2026-08-12
Japan Energy Tracker Asia 2026-03-08
Japan Al Jazeera 2026-03-23
Japan Lloyd's List Intelligence 2026-08-12
Pakistan The Conversation 2026-07-01
Pakistan Reuters 2026-03-13
Philippines The Conversation 2026-07-01
Philippines Reuters 2026-03-24
Singapore EMA Singapore 2026-01-01
Singapore CNBC 2026-03-03
Singapore Lloyd's List Intelligence 2026-08-12
South Korea Reuters 2026-03-24
South Korea Zero Carbon Analytics 2026-03-10
South Korea Lloyd's List Intelligence 2026-08-12
Sri Lanka The Conversation 2026-07-01
Sri Lanka Atlantic Council 2026-03-20
Taiwan Low-Carbon Power 2026-01-01
Taiwan Zero Carbon Analytics 2026-03-10
Taiwan Lloyd's List Intelligence 2026-08-12
Thailand Reuters 2026-03-10
Thailand CNBC 2026-03-03
Thailand Lloyd's List Intelligence 2026-08-12
Vietnam Reuters 2026-03-23
Vietnam Reuters 2026-03-26
Vietnam Lloyd's List Intelligence 2026-08-12
31 citations · 20 unique URLs