Why Every Estimate of China's Oil Stockpile Disagrees, and Which Numbers to Trust
China publishes almost nothing about its oil reserves, so every figure you have ever read is a reconstruction. This dossier explains the three methods analysts use, shows the real spread between them, and gives you a defensible way to cite the number without pretending it is a fact.
- No source has the real number. China treats stock levels as strategically sensitive, and the last meaningful official disclosure covered mid 2017: 37.73 million tonnes, roughly 280.7 million barrels, released by the National Bureau of Statistics.
- There are two independent methods, and they measure different things. Satellite gauging reads visible above ground tanks. The supply balance method infers a residual from customs and refining data. They are not interchangeable.
- Satellites are structurally blind to part of the system. Underground rock caverns cost far less per barrel than steel tanks, which gives China an incentive to keep building storage that cannot be photographed.
- The denominator decides the headline. Net imports, gross imports and total consumption give three different day counts for one identical stockpile.
- Cite a range with attribution, never a point estimate. As of mid 2026 the defensible position is roughly 1.2 to 1.4 billion barrels, converging on the lower end, at about 110 to 130 days of net import cover.
Beijing does not publish it, and that is the whole problem
China does not release routine crude inventory data. There is no weekly stocks report, no equivalent of the EIA's Wednesday release, no audited annual filing. Inventory levels are treated as strategically sensitive, which is a rational position for a country whose principal energy vulnerability is a maritime chokepoint.
The exception proves the rule. In 2018 the National Bureau of Statistics disclosed a national reserve total of 37.73 million tonnes, about 280.7 million barrels, as of mid 2017 (South China Morning Post). It was the first time an official total had ever been published, it arrived months after the period it described, and nothing on that scale has been repeated on a regular schedule since.
So every figure in circulation, including ours, is a reconstruction. That is not a criticism of the analysts doing the reconstructing. It is a fact about the data, and it should change how you cite the number.
Project 54Measured from orbit, inferred from customs data, confirmed by almost nobody: how China's reserve numbers are actually built.Three methods, measuring three different things
Understanding why the numbers disagree requires understanding that they are not all trying to measure the same object.
Satellite gauging
Providers such as Ursa Space Systems, Kpler and Vortexa photograph tank farms and read the shadow inside a floating roof tank. The roof floats on the liquid and falls as oil is drawn down, so the shadow is a gauge. Ursa tracks more than 4,000 floating roof tanks across roughly 130 Chinese locations. It measures visible above ground tanks only.
Supply balance
The dominant method for Reuters, the EIA and OIES. Domestic production plus imports plus pipeline receipts, minus refinery throughput and exports. Whatever is unaccounted for is assumed to have gone into storage. It measures a residual, not a tank.
Official disclosure
Rare, partial and late. The 2018 release covering mid 2017 remains the clearest data point. Xinhua publishes occasional commercial tank figures that cover only part of the system and exclude the strategic layer.
The spread, in real numbers
The divergence is not academic. The EIA put China's strategic inventories at nearly 1.4 billion barrels as of December 2025, after average additions of 1.1 million barrels a day through 2025 (EIA). Kpler's satellite and flow based tracking had onshore inventories at about 1,232 million barrels in late May 2026, down from a peak of 1,251 million in early May (Kpler, 25 May 2026). Reuters columnist Clyde Russell, working from the supply balance, put combined commercial and strategic stocks at at least 1.2 billion barrels, with Vortexa data cited alongside at a record 1.24 billion in April 2026.
The rate of accumulation splits even wider than the level. The Oxford Institute for Energy Studies found published implied stock build estimates for 2025 ranging from 0.43 to 0.9 million barrels a day, against its own central case of 0.75, and noted the spread between analysts averages about 0.5 million barrels a day and can reach 1.1 million (OIES, May 2026). A disagreement of 1.1 million barrels a day is larger than the entire daily output of several OPEC members.
The most instructive case is historical. Analysts at Rice University's Baker Institute, working with Orbital Insight satellite data, wrote that "at times in mid to late 2017, Orbital Insight's data suggested total crude oil stockpiles in China were more than three times as large as the figures reported by Xinhua, a potential discrepancy of more than 500 million barrels" (Collins and Hung, Baker Institute, 2018). Half a billion barrels is the gap between a partial official figure and an independent measurement of the same country.
| Source | Estimate | What it measures | Date | Why it differs |
|---|---|---|---|---|
| China NBS (official) | 37.73 million tonnes, approx. 280.7 million barrels | Disclosed national reserve total | Released 2018, covering mid 2017 | Only official figure; excludes most commercial stock; not repeated |
| EIA | Nearly 1.4 billion barrels | Combined strategic and commercial inventories | December 2025 | Broadest definition, includes commercial NOC stocks |
| Kpler | Approx. 1,232 million barrels | Onshore visible crude inventories | 25 May 2026 | Satellite and flow based; misses underground storage |
| Reuters / Clyde Russell | At least 1.2 billion barrels | Implied total from customs and refining data | March to April 2026 | Residual method; sensitive to teapot run rates |
| Vortexa (cited by Reuters) | Approx. 1.24 billion barrels, a record | Tracked commercial plus strategic stocks | April 2026 | Different tracking universe and definitions |
| OIES | Stock build 0.43 to 0.9 million b/d (central 0.75) | Rate of accumulation, not level | 2025 data, published 2026 | Different refinery run and output assumptions |
Four technical reasons the gap does not close
Invisible storage. A growing share of China's strategic capacity is built in mined underground rock caverns rather than above ground steel. Shadow based satellite analysis cannot read it at all. The Baker Institute notes that underground caverns can cost more than 60 percent less per barrel of capacity than above ground tanks, so China has both a security incentive and a cost incentive to keep building storage that satellites cannot see.
Commercial and strategic barrels are not separable from outside. The buffer is two blended layers: a state reserve and a much larger pool of commercial stock held by CNPC, Sinopec and CNOOC that Beijing can lean on in a crisis. An outside analyst cannot attribute a given barrel to one layer, so estimates diverge simply on where the author draws the line.
Refinery runs are the weak input. Independent teapot refiners in Shandong are roughly a quarter of national capacity and move run rates fast: the national average fell to 66.3 percent in May 2026, with Shandong plants at 50.5 percent in one week. Because throughput is one of two big inputs to the supply balance, an error there lands directly in the implied stock build.
Apparent demand is a proxy, not a measurement. China's apparent demand is refinery processing plus net product imports. There is a persistent gap of roughly 1.1 to 1.4 million barrels a day between crude supply and what refineries actually process. Some of that gap is statistical noise across customs, NBS and shipping data. The rest becomes the stock build. Small measurement errors therefore become large disagreements about storage.
The denominator is doing more work than the barrels
This is the part most readers miss. Days of supply is a fraction, and the numerator is only half the argument.
The IEA measures its members' 90 day obligation against net imports of the prior calendar year, a stable and audited figure. China's headline day counts are frequently quoted instead against gross imports, or against total consumption, which runs well above net imports. Both are larger denominators, so both mechanically produce a smaller day count for exactly the same barrels.
That is how one credible source says 90 days and another says 130 while describing an identical stockpile. Neither is lying. They are dividing by different things, and almost nobody says which. We set out the benchmark mechanics in the IEA 90 day stockholding dossier, and the working day count in the days of supply dossier.
A defensible position, as of mid 2026
Treat every China stockpile figure as a range and always name the method and the denominator. As of mid 2026, total Chinese crude inventories, state reserve plus commercial, most likely sit between about 1.2 and 1.4 billion barrels, converging toward the lower end after the drawdowns through the Iran conflict period. On a net import basis that is roughly 110 to 130 days of cover. On a gross import or total consumption basis the same barrels look like fewer days.
The practical rule for anyone writing, briefing or selling on this: cite the range with attribution, say plainly that Beijing publishes no confirming figure, and note that the number moves in near real time with China's buying and refining behaviour, unlike the IEA's fixed annual benchmark. A single number with no method attached is not more precise. It is just less honest.
This is also why the opacity itself is the strategy, and why the estimates matter commercially: the same ambiguity that frustrates analysts is what gives Beijing optionality in a crisis.
Listen & take it with you
Prefer audio, or need the deck for an internal review? The full briefing is available as a podcast episode and a downloadable slide presentation.
When you see a figure for China's oil reserves, what do you check first?
Frequently asked
Because Beijing does not publish routine oil inventory data. Analysts reconstruct the total from satellite readings of visible tanks and from the gap between customs reported imports, domestic output and refinery throughput. The method is credible but indirect, so every published figure carries a margin of error and is an estimate, not an official statistic.
Because sources use different methods and different denominators for the same stockpile. A total of roughly 1.2 to 1.4 billion barrels can be quoted as anywhere from about 110 to 180 days of cover depending on whether it is divided by net imports, gross imports or total consumption, and estimates of the rate of stock build differ by as much as 1.1 million barrels a day between analysts, according to OIES.
No. Satellites read the shadow cast by a floating roof on an above ground tank, which captures a large share of visible tank farms, more than 4,000 tanks in Ursa Space Systems' coverage alone. They cannot see oil in mined underground rock caverns or fixed roof tanks, and China has both a cost incentive and a security incentive to keep building exactly that kind of storage.
It is a residual calculation: domestic production plus imports plus pipeline receipts, minus refinery throughput and exports, using China customs and National Bureau of Statistics data. Whatever crude is not refined or exported is assumed to have gone into storage. Because it depends on estimating refinery runs, especially at volatile independent teapot refiners, small errors there produce large swings in the implied build.
Rarely and partially. The clearest case is the National Bureau of Statistics figure of 37.73 million tonnes, about 280.7 million barrels, disclosed in 2018 for mid 2017. Xinhua publishes occasional commercial tank data that covers only part of the system, which leaves satellite and supply balance estimates as the primary tools. For why the silence is deliberate, see our dossier on why China does not publish its oil reserves.
Get the next intelligence drop
Join energy and industrial leaders getting our marketing, AI-growth and revenue-architecture intelligence, direct, no filler.
You're on the list
Welcome to The Energy Growth Brief, watch your inbox for the next dispatch.