On August 12, 2026, Brent crude traded at $89.53 a barrel. That week, between seven and nine million barrels of oil a day were moving through the Strait of Hormuz, against roughly twenty million before the closure. Ten vessels had crossed the strait the previous day, against a pre-war norm near 130. Fresh attacks had just dented what remained of the hopes for a reopening. And the price stood about twenty-four percent above where it had been in late February, before any of this began. Half the oil was gone, and the market was charging a quarter more for it.

In late April the same barrel had touched $126, its highest level in four years. Nothing between April and August had been resolved. The strait was still contested, insurers were still refusing routine cover, and every attempt at a settlement had failed. The price came down anyway, and by August analysts were describing $85 to $90 as the level the market would support.

The explanation offered at the time was adaptation. Modern economies use far less oil per unit of output than they did in 1973. Efficiency had improved, substitutes existed, and the International Monetary Fund still projected three percent global growth for the year. Analysts wrote that the world had absorbed the largest disruption to energy supply since the 1970s and had barely broken stride. It was a reassuring account, and it had the advantage of being partly true.

It also had a problem. Through that summer an American motorist was paying $4.15 a gallon for gasoline, thirty-nine percent above the level of late February, and $5.90 for diesel, a record. A European motorist was experiencing something closer to relief. The two were buying refined products made from the same barrel, in the same week, and reading opposite signals from it. A price that resolves so differently in two markets is not reporting scarcity. It is reporting the position of whoever is not in the room.

Between April and August, the change the market narrative largely missed was the withdrawal of one buyer. Not efficiency, not substitution, and not the restoration of supply. China temporarily removed the world’s largest bid from a broken oil market and paid for months of absence out of its own storage. By late summer the buffer had reached the point where it had to come back. The world mistook the resulting discount for recovery. The market did not price the shortage. It priced how long the largest buyer could postpone confronting it.

The Buyer That Stopped Buying

In the second quarter of 2026, China imported 8.1 million barrels of crude a day, thirty-two percent below the first quarter. In May and June the figure fell below 8.0 million barrels a day for the first time since 2016. The American Petroleum Institute measured the fall from February to May at forty percent. Bloomberg called it an eight-year low. Whatever the baseline, the direction was not ambiguous and the magnitude was not marginal. The world’s largest importer removed something in the order of three and a half million barrels a day of buying from a market that was already short.

The first explanation to reach for is substitution. China has spent years building overland supply from Russia and Kazakhstan, and in the opening weeks of 2026 it had been buying Russian crude at pace. Imports from Russia reached 21.8 million tonnes in January and February, up 40.9 percent year on year, more than a fifth of total crude by volume, and at a discount steep enough that the dollar value rose only 5.8 percent. If the Gulf closed, the pipelines and the discounted seaborne barrels would carry the load. That was the design.

The design did not hold, because the substitution did not happen. The Energy Information Administration’s breakdown of the waterborne decline between the first and second quarters shows reductions distributed across suppliers rather than concentrated in the Gulf. Iraq fell by 910,000 barrels a day, Russia by 640,000, the United Arab Emirates by 600,000. Pipeline volumes from Russia held roughly flat. Russian seaborne crude, the supposed alternative, fell alongside the Gulf barrels it was meant to replace.

That pattern rules out the comfortable reading, and the overland route could not have carried the load in any case. The Oxford Institute for Energy Studies puts pipeline deliveries at roughly 1.1 million barrels a day under optimistic assumptions, against a five-year average of 10.9 million barrels a day arriving by sea. The pipelines cover about a tenth of the requirement. They were always a partial hedge against the strait rather than a replacement for it, which leaves storage as the only instrument large enough to matter. A country switching suppliers shows one source falling while another rises. China’s sources fell together. The constraint was not access to any particular barrel. It was the price of all of them, and the decision not to pay it.

China did not change suppliers. It stopped taking delivery.

The Buyer With a Pass

There is a second explanation to dispose of, and disposing of it is what makes the rest of this argument possible. The obvious reason a country stops importing oil during a naval crisis is that it cannot get the oil. In the spring of 2026 that was true of a great many buyers. Daily transits through the strait collapsed from more than a hundred ships to single digits. Around two thousand vessels were left waiting for clearance. Protection and indemnity cover was withdrawn, which closed the waterway commercially while leaving it technically open, and the United States imposed a naval blockade on Iran. For most of the world’s shipowners the question of whether to buy Gulf crude did not arise, because no underwriter would cover the voyage.

China was not in that position. Iran ran the strait as a vetting system rather than a closure, and it published the guest list in everything but name. Vessels linked to China, Russia, India, Iraq, Pakistan, Malaysia, Thailand and later the Philippines were cleared to pass. Vessels linked to the United States and Israel were not. Transits ran through what Gulf News described as a pre-approved toll booth mechanism, and Iran’s parliament moved to put formal transit fees into law. Bloomberg reported that charges of up to two million dollars a voyage were being sought on an ad hoc basis, payable in Chinese yuan, which Iran denied. On July 5, Iran’s ambassador to China, Abdolreza Rahmani Fazli, said publicly that China and other friendly nations would receive special considerations on those fees.

So the country with the best access to Gulf crude of any large importer, the one whose ambassador was being offered a discount on a toll nobody else could avoid, cut its imports by roughly a third. Access alone was not the binding constraint. Neither was insurance alone. Neither could explain why Gulf and non-Gulf suppliers fell together, since the discounted Russian barrels were falling at the same time as the Iraqi and Emirati ones. What remains, once those are removed, is a decision about price, taken by a buyer who had somewhere else to get its barrels from. The somewhere else was not another country.

The one buyer who could still pass through the strait was the one who chose not to.

Fifty-Four Percent Came Out of a Tank

Where does three and a half million barrels a day of unpurchased crude go? It does not simply vanish from the domestic economy, because Chinese factories, trucks and aircraft were still running through the spring. Erica Downs and Michal Meidan at Columbia University’s Center on Global Energy Policy took the question apart, and their answer splits cleanly in two. Slightly more than half of the import reduction was covered by drawing down stored crude. The rest came from refining less of it.

The inventory half is the more striking. In the second quarter of 2025, China had been adding 1.57 million barrels a day to storage. In the second quarter of 2026 it was withdrawing 363,000 barrels a day. The swing between those two positions is 1.93 million barrels a day, and it accounts for fifty-four percent of the fall in imports. Through the second quarter, more than half of China’s apparent immunity to the Hormuz crisis was a tank being drawn down.

There is a distinction inside that figure which changes what the return to the market means. The Oxford Institute for Energy Studies reports that Beijing resisted strategic reserve releases in April and initially constrained stock draws more broadly, preferring refiners to cut runs. As the crisis continued, managed commercial destocking became one of the few remaining levers. The Columbia account corroborates the first half of that from the other side, describing Beijing as reluctant to let refiners tap commercial reserve sources, which is why they cut operating rates instead. The sequence is also visible in the arithmetic. Observable draws through April were minimal, yet the quarter as a whole averaged 363,000 barrels a day of withdrawal, which puts most of the drawdown in the weeks after Beijing ran out of cheaper instruments. Neither institution is working from figures Beijing publishes, because Beijing publishes none. If the strategic layer remained largely untouched, the threshold reached in the summer was not the bottom of China’s tanks. It was the point at which the commercial buffer and the policy protecting the state reserve began to collide. A barrel in storage is only supply if the state is willing to release it.

The tanks were unusually full when the war started. Entering 2026, Chinese crude inventories stood at roughly 1.4 billion barrels, of which around 360 million were government reserve and the remainder commercial stock held by refiners and state companies. That is approximately 121 days of import cover, comfortably above the ninety days the International Energy Agency treats as a security benchmark. It had been built deliberately. In the early weeks of 2026, before the strikes, China was still accumulating at around 1.24 million barrels a day. Beijing had spent years buying cheap barrels against exactly this contingency, and when the contingency arrived the buffer performed as designed.

Performing as designed is not the same as solving the problem. A reserve is measured in barrels, not in barrels per day. It answers the question of how long, and it answers no other question. Efficiency reduces the requirement permanently. Substitution moves the requirement elsewhere permanently. Storage does neither. It sets a date.

A strategic reserve is not a source of supply. It is a source of time.

The Margin Where the Shock Was Parked

The other forty-six percent of the import decline came from refineries processing less crude. Chinese throughput fell 1.6 million barrels a day year on year, to 12.8 million in the second quarter, and in June reached 12.5 million, the lowest level since the first months of 2020. The independent refiners in Shandong, known in the trade as teapots, cut operating rates to their lowest since 2017.

The reason was not, at least initially, a collapse in Chinese demand for fuel. It was arithmetic on a refining margin. Wholesale crude costs rose by roughly forty dollars a barrel by early April. Beijing holds administrative caps on retail fuel prices, and those caps did not move far enough or fast enough to follow. A refiner buying at world prices and selling at capped prices loses money on every barrel it processes, and the rational response is to process fewer.

That is a specific and locatable transfer. The cap protected Chinese households and Chinese industry from the full force of the crude price. Somebody still paid the difference, and the somebody was the refining sector, disproportionately the independents in Shandong who have no upstream production to offset a negative downstream margin and no state balance sheet standing behind them. Households and hauliers absorbed part of it too. Chinese gasoline demand fell five percent year on year in the second quarter and diesel demand fell thirteen percent, in the form of trips not taken and freight not moved. But the first and largest absorber was a margin.

Described as national resilience, this looks like strength. Described mechanically, it is a subsidy paid by one industry to postpone a price signal reaching everyone else. It worked, in the narrow sense that Chinese consumers experienced a milder shock than American consumers did. It could not compound, because a negative margin is not a renewable resource either.

The price cap did not absorb the shock. It chose who would carry it.

The Fleet That Was Given the Credit

By the summer the story had acquired a hero. More than half of new vehicles sold in China are now electric, and that statistic travelled widely, attached to the conclusion that China had escaped the oil shock because it had already stopped needing oil. The statistic is accurate, and electrification is genuinely bending China’s long-run oil trajectory. The conclusion drawn from it is still wrong, and the reason is timescale.

When Downs and Meidan looked for the electric contribution in the second quarter data, what they found was thin. Anecdotal reports of increased taxi use and more electric vehicle rentals. That is a description of consumers economising during a price spike, not of a fleet that has structurally displaced gasoline. The Columbia analysis attributes the demand side of the adjustment to price and the supply side of it to storage and to run cuts. The electric fleet does not appear in the arithmetic that closes the gap.

There is a structural reason for that, and it is not a criticism of electrification. A vehicle fleet turns over on a timescale of a decade or more. Even at a sales share above fifty percent, the share of kilometres driven on electricity rises by a few points a year, because most of the cars on the road in 2026 were bought before 2026. J.P. Morgan puts the permanent element of the Chinese gasoline decline at around 126,000 barrels a day, roughly seventy percent of a 180,000 barrel loss. That is real, and it is structural, and it is less than four percent of the import reduction this article is trying to explain.

The confusion matters because of what is being inferred from it. If China rode out the largest oil disruption since the 1970s because it had electrified, that is an argument for electrification as energy security, and Europe is currently making precisely that argument to itself. If China rode it out by drawing down a reserve and squeezing its refiners, the security lesson is different and considerably less comfortable, because neither of those things can be done a second time.

EVs changed China’s oil trajectory. They did not make three and a half million barrels a day disappear in one quarter. Tanks did that. Refinery cuts did that. Price did that.

The Threshold in Shandong

Storage is measurable, if you are willing to count roofs. By late July, crude stocks in Shandong had fallen to an eight-month low, with around 35 million barrels drawn in that month alone, the largest monthly decline in the series Energy Aspects has compiled since 2016. Shandong is not the whole country, and the national drawdown was broader than one province. But Shandong is where the effect surfaces first, because the independent refiners there hold the least cushion and have the least room to wait.

The reversal arrived in the same month. Chinese crude imports rebounded twenty-two percent in July, to an average of 8.45 million barrels a day. Domestic stocks of finished gasoline and diesel had fallen back to December 2025 levels, which meant refiners had to run in order to rebuild them, which meant they had to buy crude in order to run. Kpler’s forecast in mid-August had crude intake climbing from 12.64 million barrels a day in July to 12.92 in August, 13.29 in September and 13.54 by October.

Sourcing the crude to match that recovery is the difficulty. Persian Gulf cargoes were arriving two to three weeks behind schedule. Chinese refiners committed to 23 million barrels of August-loading Saudi crude but kept September nominations low, because Houthi attacks were forcing cargoes around the Cape of Good Hope and adding roughly twenty days to the voyage. Competition for the alternatives tightened at once. Russian ESPO cargoes for September were bought before formal trading opened, and the grade moved from four dollars below Brent in mid-July to a dollar above it. The Shandong teapots began buying Iranian crude again in August, for the first time in months.

The detail worth holding onto is where the returning barrels came from. Millions of barrels of Iranian crude had been positioned for Asian delivery during a window between mid-June and early July, when the American blockade restrictions briefly lifted, and the teapots moved on them once Shandong stocks hit their floor. The marginal supply for the world’s largest importer, at the moment its buffer hit the threshold, was sanctioned crude loaded inside the Gulf. A restocking cycle that is supposed to demonstrate Chinese independence from the strait is being fed through the strait, by the country the blockade is aimed at, into refineries that had run at their lowest rates since 2017. The dependency did not move. It only became harder to see on a customs form.

Kpler’s own conclusion, published on August 13, was that unless the recovery in refinery runs stalled again, China would need to return to the international crude market more meaningfully, and sooner, than the market expected. That sentence is the thesis of this article, issued weeks in advance as a routine note to shipping clients.

The buffer did not fail. It reached the level where drawing on it stopped being cheaper than buying.

Storage is a countdown, not a policy.

The Price Moved Without the Ships

On September 7, Brent traded around $97 a barrel, up nine percent in five days and nineteen percent over the month. West Texas Intermediate reached $92.27. Both were at six-week highs, and Brent stood roughly eight dollars above where it had traded in the middle of August.

The ships had not come back. Ten vessels crossed the strait on August 11. In the ten days to September 7, an average of ten commodity ships a day crossed it. That is the same waterway, at the same rate, a month apart, running at a fraction of a pre-war norm near 130 transits a day. There were incidents in the interval. American forces struck three Iranian tankers, Saudi Aramco’s Jizan refinery was hit for the second time in a month, and the Islamic Revolutionary Guard Corps claimed attacks on three vessels with American links. There had been comparable weeks through the summer that moved the price by a dollar or two. The physical throughput that supposedly sets this price did not change between August and September. The price moved by eight dollars.

One clarification is owed, because it is the obvious objection. Gulf exports and Hormuz transits are not the same measurement. Saudi Arabia can load at Yanbu on the Red Sea and the Emirates at Fujairah, both outside the strait, so regional export volumes can recover while transits stay flat. It does not change the point, because the transit figure is stable across the two dates and the price is not.

The ships did not return. The buyer did. A market that had spent the summer clearing without its largest participant had to clear with it again, at a moment when the alternative grades were already bid up and Gulf cargoes were three weeks late.

China was no longer buying for consumption alone. It was buying consumption and inventory at the same time, and that is the part the market has not finished pricing. Kpler’s crude intake path implies close to nine hundred thousand barrels a day of additional refinery demand between July and October, and the crude to feed it has to travel further than it did in February, because the Cape of Good Hope routing that Houthi activity has forced on Saudi cargoes adds roughly twenty days to each voyage and therefore ties up tonnage that would otherwise be making another trip. Longer voyages consume the same fleet twice. The effective supply of transport falls even when the supply of oil does not.

Iran contests the strait. Insurers withdraw cover and freight reprices. Physical flows fall and crude rises. China declines to pay the new price and draws on stored barrels instead. The largest buyer leaves the market and the price falls back. The buffer reaches its threshold and the buyer returns at whatever the price has become. That is not a market finding its level. That is a machine with a delay built into it.

Hormuz was still broken. What changed was that China could no longer afford to stay out of the market.

The price was measuring absence.

The Listing That Did Not Move

The buffer in Shandong has a counterpart in London, and unlike most of this story it leaves a paper trail with reference numbers on it. The Lloyd’s Joint War Committee, a body of roughly fifteen private underwriters with no public mandate, publishes the listed areas that decide whether a voyage is commercially insurable at all. On March 3 it issued circular JWLA-033, adding Bahrain, Djibouti, Kuwait, Oman and Qatar, and defining the listed water as the Persian and Arabian Gulf, the Gulf of Oman, the Indian Ocean, the Gulf of Aden and the southern Red Sea. By March 18 the listing covered the entire Gulf, the strait, the Gulf of Oman and the northern Arabian Sea, the widest Gulf listing in the committee’s history.

Then it published nothing for almost five months. The next revision, JWLA-034, is dated July 29 and took effect on August 8, four days before the price this article opens on. It amended Saudi Arabia and Eritrea, adjusted the boundaries of the Gulf composite area, and deleted Pakistan. It did not remove the Persian Gulf, the Strait of Hormuz or the Gulf of Oman. In the only revision the committee published between the first week of the war and the second week of August, covering the waterway the war was being fought over, the listing came off a country and stayed on the strait. No rationale was given, because none is ever given.

The premiums behaved differently, and the difference is the point. War risk cover for a Gulf transit ran at two to five hundredths of a percent of hull value in 2024 and five to ten hundredths in January 2026. In the first week after the strikes it reached one and a half to two and a half percent, and Reuters reported a three percent rate quoted on a tanker worth $250 million, about $7.5 million for a single passage against roughly $625,000 before the war. By March 18 it had already fallen back to between half a percent and one and a half, and on April 13 Argus put the additional premium near one percent, with a no claim bonus of thirty-five to fifty percent for vessels that stayed inside the Gulf. The rate came down while the listing was still expanding. The listing then did not move for five months.

That divergence is not a detail of underwriting practice. A rate is a price a shipowner can decide to pay. A listing is a condition on whether the voyage is written at all, and it carries its own machinery: applications forty-eight hours before entry, quotes valid for forty-eight hours and sometimes twenty-four, seven days of cancellation notice and seventy-two hours where one of the five powers is involved. A market that reprices in a week and reclassifies twice in half a year is not slow because it is careless. It is slow because classification is the part that cannot be undone cheaply.

A rate tells a shipowner what the voyage costs. A listing tells him whether there is a voyage.

Read the two together and the pattern in this article repeats one level up. Beijing decided which barrels counted as available. Fifteen underwriters in London decided which water counted as passable. Neither decision was published with a reason, both were taken by people with no obligation to explain them, and between them they governed what a barrel cost in a market that believed it was watching a war.

The Comparison That Flattered Everyone

The reason a drawdown could be mistaken for adaptation for months is that a ready-made comparison was available and everyone reached for it. The disruption of 2026 was, in absolute terms, the largest interruption to world energy supply since the 1970s. Roughly twenty million barrels of oil and refined products a day moved through the strait before the closure. By August it was seven to nine million. Daily transits fell from more than a hundred ships to single digits and stayed there. In March the International Energy Agency approved the largest emergency release in its history, four hundred million barrels, and Brent was trading near $91 when it did.

And then the world did not collapse. Brent came off its April peak. The International Monetary Fund kept its three percent growth projection. The Organisation for Economic Co-operation and Development raised its United States inflation forecast to 4.2 percent, 1.2 points higher than before, which is a real cost but not a 1970s cost. The natural conclusion, drawn almost universally by the summer, was that the comparison to 1973 had been overdrawn, that four decades of efficiency gains and fuel switching had made the modern economy structurally harder to hurt through a chokepoint. Every part of that conclusion is defensible on its own. Assembled into an explanation of the August price, it is wrong, because it explains a resilience that had not yet been tested.

The tell was sitting in the same datasets. Global inventory draws reached an estimated 5.1 million barrels a day in the second quarter of 2026. A world that is drawing down stocks at that rate is a world consuming more than it produces, which is the classic setup for prices to rise, not fall. Prices fell anyway. When the physical balance and the price move in opposite directions for a quarter, the price is being set by something other than the physical balance, and the candidate explanations are few. One of them was visible in the customs data of a single country.

The world was not absorbing the shock. It was watching the largest buyer decline to bid on it.

What Europe Bought With the Discount

European buyers experienced the summer as relief, and this is where precision is needed, because the relief was smaller than the headline. Brent fell by roughly thirty percent from its April peak to its August range, but European pump prices are dominated by excise duty and retail margin, so only a fraction of that decline reached a forecourt in Rotterdam or Lyon. Gas told a cleaner story. Qatari interruptions to liquefied natural gas output had driven European gas up around fifty percent in March, with Goldman Sachs estimating that a full Hormuz disruption could carry it 130 percent higher, and the summer easing in that market was material.

Whatever did reach European consumers was borrowed. It was financed by barrels sitting in Chinese tanks that are no longer there, and by Chinese refining margins that cannot be compressed a second time. Europe did not benefit from a market that had healed. It benefited from a competitor that had temporarily stepped back, for reasons of its own, using a resource that has since been consumed.

This bears directly on an argument Europe is currently having with itself. The strategic case for accelerating the energy transition is partly a security case. Electrify, and the Gulf loses its hold. China is the exhibit most often cited. But the exhibit does not show what it is taken to show. In 2026 China’s insulation came from stored crude and administered prices, not from its electric fleet. And the substitution Europe is being offered runs through supply chains that China controls more completely than anyone has ever controlled oil, from the manufacture of the turbines themselves to the separation of the rare earth elements they depend on.

The asymmetry is worth stating precisely, because it is the part that survives any disagreement about the numbers. Oil can be stored. Europe holds ninety days of it by treaty obligation, the United States holds a strategic reserve, China holds around a hundred and twenty days, and that storage is what bought everyone time this year. There is no equivalent for a rare earth supply chain. A country cannot stockpile the industrial capacity to separate dysprosium, and the lead time to build it is measured in years rather than in tanker voyages. The dependency Europe is escaping came with a shock absorber. The dependency it is acquiring does not have one.

None of that is an argument against electrification, which rests on reasons that have nothing to do with this war. It is an argument against citing the summer of 2026 as evidence for it. The insulation Europe admired was a drawdown, and the dependency it is being invited to adopt in exchange has no strategic reserve behind it at all.

Europe read a Chinese inventory statement as a Chinese energy policy.

The Strongest Counterargument

The strongest counterargument does not dispute a single figure above. It accepts the import collapse, the inventory swing, the run cuts and the July reversal, and reads them as competence rather than constraint. On this account China behaved as a sophisticated buyer always behaves. It accumulated cheaply before the war, including 21.8 million tonnes of discounted Russian crude in January and February alone, declined to chase a spiking market, ran down stock while prices were high, and resumed buying once the peak had passed. That is arbitrage executed well, not a buffer exhausted. The run cuts fit the same reading, because the price caps that made refining unprofitable are Beijing’s own instrument and can be lifted whenever Beijing decides the trade-off has changed. And the September price rise has serviceable supply-side explanations that require no reference to Chinese demand at all: strikes on tankers, a second hit on Jizan, Houthi attacks pushing cargoes around the Cape.

A second version of the objection is sharper still, and it comes from the forecasters. J.P. Morgan reads the whole episode as larger-than-expected demand destruction rather than deferred demand, and puts China’s structural crude requirement about a million barrels a day below prior expectations. On that reading the buyer who returned in the autumn is permanently smaller than the buyer who left in the spring, and the word doing the work in this article, temporary, is the wrong word. Both objections are structurally serious, and on one point they are stronger than the reading offered here. China publishes nothing about its reserve position. Every inventory figure in this article is a reconstruction by Kpler, Energy Aspects and Vortexa from tanker movements and tank-farm imagery, and reconstructions carry error bars that official statistics do not. The reading offered here does not claim that China was driven to the edge of its capacity, nor that Chinese buying alone determined the summer price. It claims something narrower. A substantial part of the mid-2026 price relief was produced by the temporary absence of the largest buyer, that absence was financed from a stock rather than from any structural change in consumption, and the size of the stock therefore set the length of the relief. Strategy and constraint are not alternatives in this account. A buyer can time a drawdown perfectly and still be governed by how much there is to draw.

Return to August 12 and the $89.53. Read as a market signal, it said that the world had metabolised the largest interruption to oil supply since the 1970s in under six months. Read accurately, it said that half the oil was still missing, that the largest importer was not bidding that week, and that the tanks in Shandong still had something in them.

The headlines said the world had adapted. The tank data said China was drawing down. The price said the crisis was easing. The inventories said the crisis was only being settled later. Nobody was responsible for the difference between those two readings. There is no institution whose function it is to publish the second one. China does not disclose its reserve position. The Joint War Committee does not explain its listed areas. The refiners running at a loss in Shandong file no statement that a European motorist will ever see. The most consequential number in the oil market for months was assembled by private analysts photographing the tops of storage tanks, sold to the clients who could afford it, and reported to everyone else as weather.

Every participant in this market held a position and knew what it was. The refiner knew its margin. The underwriter knew its classification. The state knew its inventory to the barrel. The trader knew which of them to watch. The motorist was handed a number, told it meant the crisis was easing, and given no way to see that it meant a tank was being drawn down eight thousand kilometres away.

Hormuz was visible and could not explain the price. Shandong was barely visible and could.

Every participant in this market had a position. The one who paid the price had only the price.

Evidence Map

Core claim: A substantial share of the mid-2026 fall in crude prices was produced by the temporary withdrawal of China from the import market, financed from stored crude and compressed refining margins rather than from any structural reduction in demand, and it ended when the drawdown reached the level at which China had to return to the market. The market did not price the shortage. It priced how long the largest buyer could postpone confronting it.

Observed conditions (high confidence, directly documented): Chinese crude imports of 8.1 million barrels a day in the second quarter of 2026, thirty-two percent below the first quarter, and below 8.0 million in May and June for the first time since 2016 (Energy Information Administration). Waterborne declines between the first and second quarters: Iraq 910,000 barrels a day, Russia 640,000, United Arab Emirates 600,000. Refinery throughput of 12.8 million barrels a day in the second quarter, 12.5 million in June. Chinese gasoline demand down five percent and diesel down thirteen percent year on year. July imports of 8.45 million barrels a day, up twenty-two percent. Brent at $126 in late April, $89.53 on August 12 and approximately $97 on September 7; West Texas Intermediate at $92.27. Ten vessels transiting the strait on August 11 and an average of ten commodity ships a day in the ten days to September 7, against a pre-war norm near 130. Oil flow through the strait of seven to nine million barrels a day in August against roughly twenty million before the closure. Joint War Committee circular JWLA-033 of March 3, 2026 adding Bahrain, Djibouti, Kuwait, Oman and Qatar, and circular JWLA-034 of July 29, 2026, effective August 8, amending Saudi Arabia and Eritrea and deleting Pakistan while leaving the Persian Gulf, the Strait of Hormuz and the Gulf of Oman listed.

Documented structural dependencies (medium-high confidence): An inventory swing of 1.93 million barrels a day between accumulation in the second quarter of 2025 and withdrawal in the second quarter of 2026, accounting for fifty-four percent of the import decline (Downs and Meidan, Center on Global Energy Policy). Shandong crude stocks at an eight-month low in late July, down roughly 35 million barrels within that month, the largest monthly fall in the Energy Aspects series since 2016. Retail fuel price caps compressing refining margins against a crude cost increase of about forty dollars a barrel by early April. Iranian selective passage granting transit to Chinese, Russian, Indian, Iraqi, Pakistani, Malaysian, Thai and Philippine-linked vessels while denying it to American and Israeli-linked ones.

Analytical inferences (medium confidence): That Chinese absence, rather than efficiency or substitution, was the largest single removable factor behind the August price level. That the September recovery in price without any recovery in transits reflects the return of that buyer. That the electric vehicle fleet cannot account for a quarter-scale import reduction given the arithmetic of fleet turnover and J.P. Morgan’s estimate of roughly 126,000 barrels a day of permanent gasoline loss. That the drawdown fell on commercial rather than strategic stocks and was rationed rather than released freely, making the summer threshold a policy boundary rather than a physical floor. That war risk rates and listed area classifications are governed by different institutional clocks, the first repricing within a week and the second revised twice in five months, which is documented in the circulars but the causal reading of it is inference. OIES and Columbia describe that restraint from different sides, the first as resistance to releases and the second as reluctance to let refiners tap commercial sources; both are hedged and neither works from published Chinese figures. The phasing within the quarter is inferred from minimal observable draws through April against a quarterly average of 363,000 barrels a day.

What would confirm this: Chinese import volumes sustaining above nine million barrels a day through the fourth quarter while transits remain suppressed. Continued Shandong inventory decline into the fourth quarter. Kpler and Vortexa crude intake series tracking the forecast path toward 13.5 million barrels a day. Any official Chinese disclosure of reserve drawdown for the period.

What would disprove this: Chinese imports falling back below eight million barrels a day in the fourth quarter without a corresponding fall in price. Evidence that the August price level coincided with a supply increase sufficient to explain it independently. Material divergence between the Kpler, Energy Aspects and Vortexa inventory reconstructions for July. Evidence that the decline in Chinese gasoline demand was driven by fleet electrification rather than by price. Brent settling toward J.P. Morgan’s $80 fourth-quarter average while Chinese imports stay high, which would indicate that supply recovery rather than Chinese absence was the governing variable all along.

Watchlist: Monthly Chinese customs crude import data. Shandong tank-farm inventory series. Any signal that Beijing has authorised strategic reserve draws or invoked the October 2021 supply directive. Lloyd’s Joint War Committee listed area revisions for the Gulf, by circular number and date, and specifically the first circular that removes Gulf water rather than adding it. Saudi Aramco loading nominations to Chinese buyers. ESPO differentials to Brent. Teapot operating rates and Iranian cargo arrivals. Daily transit counts against the seven to nine million barrel flow figure, since the two can move independently.

Confidence assessment: Claim B | Transmission B | Causality C | Synthesis B. The observed volumes, transits and prices are documented. The inventory reconstruction is estimated rather than disclosed. The causal weight assigned to Chinese absence in the summer price is the weakest link and is stated as an inference.

Sources: Import volumes and the supplier breakdown are from the U.S. Energy Information Administration. The decomposition of the import decline into inventory draws and refinery run cuts is from Erica Downs and Michal Meidan at the Center on Global Energy Policy, Columbia University. Crude intake forecasts, cargo delays and grade differentials are from Kpler, August 13, 2026. Shandong inventory movements are from Energy Aspects as reported by Bloomberg. Reserve policy is from the Oxford Institute for Energy Studies, China’s crude levers, May 2026. Prices, transit counts and flow volumes are from Al Jazeera of August 12 and September 7, 2026, and from Trading Economics. The selective passage arrangements are from Gulf News and from Al Jazeera’s report of July 5, 2026 on the remarks of Iran’s ambassador to China. European gas estimates are from Goldman Sachs. Forward price expectations are from J.P. Morgan Global Research. The Joint War Committee circulars JWLA-033 and JWLA-034 are published by the Lloyd’s Market Association; war risk premium levels are from Reuters for March and Argus for April 13, 2026.

This mechanism has two documented layers in the archive. Oil Falls While Hormuz Stays Closed: The Arithmetic established the supply side of the same puzzle, tracing the suppressed price to the largest emergency reserve release in history and an American reserve at its lowest level since 1983. Everyone Is Watching Hormuz. Nobody Is Watching the Other Five Passages. set out why a chokepoint is a political instrument before it is a geographic one.