THE PHANTOM BARRELS
China crude balance 2022–2026
Dossier · China Crude Balance · 2022 – July 2026

THE PHANTOM BARRELSdemand that never was

China's crude imports since 2023 systematically overstated its real economy. The gap was a strategic war reserve, accumulated as Project 2025 telegraphed an energy offensive against its three discounted suppliers: Iran, Russia, Venezuela. The 2026 import collapse is not demand destruction. It is the reserve doing its job.

0.5→1.7B
bbl total crude reserves, start-2022 to start-war (EIA / Kpler / CFR)
+1.1
mb/d avg. added to inventories in 2025 (EIA)
−5.3
mb/d import cut Feb→Jun '26; June seaborne lowest since 2016
1/5×
domestic price volatility vs Brent through the March spike
$74
Brent, Jul 7: +3% (+5.6% AH) as Iran struck 3 tankers; GL X revoked
01

The wedge

Apparent demand (imports + output)
Est. true consumption
Wedge = stockpile builds
Inversion = drawdown
Drag the timeline · tap the pins
171615 14131211 MB/D CRUDE 202220232024 20252026 JUL '26 STOCKPILE BUILDS · ≈0.6–1.9 MB/D ≈1.2B BBL BUILT · 0.5B → 1.7B, 2022–FEB 2026 DRAWDOWN −1.5–1.7 mb/d APR 2023 · PROJECT 2025 PUBLISHED NOV 2024 · ELECTION + ENERGY LAW WAR · HORMUZ SHUT FEB 28 '26 NOV '25: imports 12.43 mb/d 1.88 mb/d to tanks MAY '26: 7.8 mb/d · decade low JUN '26: seaborne 6.4 · lowest since 2016 APR '23: diesel peaks 4.7 mb/d, then structural decline
passage without harm
02

The event ledger

Each entry maps to a pin on the timeline above. Tap a pin to highlight its entry; tap an entry to locate its pin. Black badges are Chinese moves, red badges are US and war events.

2022–23 · China
Discount absorption begins

China absorbs 1+ mb/d of surplus, mostly Russian, discounted crude into tanks even as demand growth cools. (Stillwater)

Apr 2023 · US
Mandate for Leadership

Heritage publishes Project 2025. It names China, Iran, Russia, Venezuela and North Korea as the five priority countries, and frames re-hemisphering energy away from distant suppliers. Beijing can read.

2024 · China
NOC stockpile directive

Beijing directs national oil companies to add emergency oil to commercial tanks: a shadow second SPR. (EIA, reported 2024)

Aug 2024 · China
Real demand rolls over

Gasoline −14% y/y; NEVs near half of car sales; LNG trucks hit 42% of heavy-duty sales; refinery runs fall 3.3% in 2024. (EIA / CGEP / NBS)

Nov 2024 · both
Election + Energy Law

Trump wins. The same month, Beijing passes the Energy Law enshrining corporate stockpiling in statute, effective Jan 1 2025. (OIES)

Jan 2025 · US
Energy dominance orders

Inauguration; Unleashing American Energy; National Energy Dominance Council. The Project 2025 energy chapter becomes policy nearly verbatim.

Jun 2025 · war
12-day air war

Israel and Iran exchange strikes; Operation Midnight Hammer degrades Iran's nuclear program. Chinese refiners, already stocked, cut buying into the price spike.

Jan 3 2026 · US
Venezuela operation

US captures Maduro; Washington states all Venezuelan oil now flows through authorized channels, cutting the discounted flow to China. (CGEP)

Feb 15–20 2026 · China
Frontloading spree

Iran triples its export rate days before the strikes; Jan–Feb imports run 11.7+ mb/d. The last fill-up. (CRS; GAC)

Feb 28 2026 · war
Epic Fury; Hormuz shut

US and Israeli strikes begin. Mar 4: Iran declares the strait closed; ~10 mb/d of Gulf exports lost, the largest supply disruption in oil-market history. (IEA/CRS)

May 14–15 2026 · both
Xi–Trump summit

Beijing summit; both agree the strait must open. Brent holds near $100. China's cut, not diplomacy, is the price cap. (CNBC / Morgan Stanley)

Jun 27–Jul 2026 · both
MoU + reopening

US–Iran memorandum, 60-day framework, temporary sanctions waiver; JMIC widens the corridor; Hormuz flows top 10 mb/d. Brent returns to pre-war $72, the biggest monthly drop since 2020. (Al Jazeera / TradingEconomics)

Jul 7 2026 · war
Re-escalation; lanes contested

Iran strikes 3 tankers on the US-Navy southern lane (Qatari LNG Al-Rekayyat, Saudi VLCC Wedyan); US retaliates on air-defense and anti-ship sites; Treasury revokes GL X. JMIC threat "severe." Brent +3% to $74, +5.6% AH. (CNBC / CBS)

03

Five legs of the thesis

Real demand peaked first

Transport fuels (gasoline + diesel + jet) fell in 2024 to ≈8.1 mb/d, 2.5% below 2021. Diesel: 4.7 → 4.0 mb/d (Apr '23 → Apr '25). Gasoline −14% y/y in Aug '24. Sinopec and CNPC's own institutes date the transport-fuel peak to 2023. Only petrochemicals grew (+5%).

IEA · EIA · Rhodium · Energy Intelligence
The wedge, quantified

Implied stock builds ran ≈0.75–1.1 mb/d in 2025 (OIES vs EIA), effectively all marginal import growth, ≈7–10% of headline "demand." Reserves went from ~0.5B bbl at the start of 2022 to ~1.7B at the start of the war — a ~1.2B build. EIA's conservative onshore-only tally reached ~1.4B by Dec '25 (~360M government + ~1B commercial); including floating and bonded crude the full number runs to ~1.7B.

EIA Apr '26 · OIES Feb '26 · Kpler
The timing telegraphs

Builds accelerate in sequence: P2025 published (Apr '23) → NOC directive ('24) → Energy Law passed the month Trump won → record 12.43 mb/d imports Nov '25 → frontloading as Iran triples exports days before the war. A Jun '24 Baker Institute testimony had already flagged energy stockpiling as the classic pre-war indicator.

Baker Institute · OIES · GAC customs
The proof by inversion

Imports cut 11.7 → 7.8 mb/d with no fuel crisis: domestic prices moved at one-fifth of Brent's volatility. A −3.9 mb/d cut absorbed calmly is only possible if the marginal barrels were never consumption. China alone = 74% of the global crude-trade decline, capping Brent near $100 amid a 10 mb/d supply loss.

JPMorgan · SocGen · Morgan Stanley
The reserve trajectory
From ~0.5B to ~1.7B barrels, 2022 to the war

China entered 2022 with roughly 500M barrels of crude reserves — an SPR of ~400M against ~500M of government-system capacity, plus a thin commercial base. By the start of the war it held roughly 1.7B barrels. EIA's deliberately conservative onshore-only tally reached ~1.4B by December 2025 (~360M government + ~1B commercial) and was still building at ~1.24 mb/d into February; adding floating and Iranian bonded crude that EIA excludes lifts the fuller figure toward 1.7B. Either way the build is ~1.2B barrels, which ties directly to the wedge.

1.81.41.00.6 2022202320242025war EIA onshore ~1.4B ~0.5B ~1.7B CHINA TOTAL CRUDE RESERVES · BILLION BBL
Leg 5 · the domestic-production offset
China ramped its own barrels to fill the gap

The wedge is not the whole story. Chinese crude production sits on a decade-long linear uptrend and hit an all-time high near 4.5 mb/d in March 2026 (EIA), up from ~3.7 mb/d in 2018. When Beijing decided to cut imports, it could raise domestic output immediately, offsetting the barrels it declined to buy. Every incremental home-produced barrel is a barrel of import demand removed from the seaborne market — and unlike commercial buying it is price-insensitive and state-directed, so it subtracts from global demand precisely when the state wants prices contained.

Stacked against the import cut, the offset compounds: a ~0.3–0.5 mb/d production ramp on top of a ~1 mb/d stockpile drawdown means the seaborne market lost well over a full mb/d of Chinese demand that had nothing to do with consumption. The same lever runs in reverse later: when domestic fields plateau, marginal demand snaps back to imports.

4.54.23.93.6 201720202023'26 ATH 4.49 mb/d CHINA CRUDE OUTPUT · MB/D · EIA (monthly) decade uptrend
The paper trail · what Beijing read, April 2023
“…provide tools for U.S. policymakers to assist our allies and deter our adversaries.”
— Department of Energy chapter, Mandate for Leadership (2023), on energy dominance as foreign policy
“…embraced and enriched Communist China and its genocidal Communist Party…”
— C. Miller, Department of Defense chapter, on 30 years of engagement policy
“…restore the free flow of energy among the hemisphere's largest producers…”
— Department of State chapter, on a hemisphere-centered energy policy
The document's through-line was legible from Beijing: energy as leverage, re-hemisphering supply away from producers it considers open to manipulation, a nuclear buildup the Bulletin of the Atomic Scientists called the largest since Reagan, and a State chapter naming the five countries the next administration should prioritize: China, Iran, Russia, Venezuela, North Korea. Three of the five were China's discounted crude suppliers. The fourth was China.
04

The case against, taken seriously

Objection 01
Price opportunism, not politics

The builds tracked price, not geopolitics. Fills clustered whenever Brent traded below $70; the Q2 '25 binge followed a four-year price low. A rational importer buys dips. No foresight required.

Response

Partly true, and insufficient. The Energy Law codified stockpiling in the month Trump won, and OIES documents tolerance for building even into the $70s once the mandate landed. Price set the pace; policy set the floor. The Feb '26 frontloading happened into rising prices, the opposite of dip-buying.

Objection 02
The cut was forced, not chosen

Imports fell because the blockade and counter-blockade physically removed Iranian and Gulf barrels. Recasting a forced outage as strategic discipline overstates Beijing's agency.

Response

Conceded in part, and the thesis stands anyway. China declined to bid for replacement barrels at $100+ while domestic prices moved at a fifth of Brent's volatility. A forced outage without a deep reserve looks like 1973: a 7% supply loss then drove prices +134%. A 14% loss now drove +30%. The difference is the tanks.

Objection 03
Taiwan, not Project 2025

The reserve is a Taiwan-2027 contingency asset and an all-purpose sanctions hedge. Stockpiling predates Project 2025 by two decades; the correlation with one document is cherry-picked.

Response

Not mutually exclusive. Both scenarios price the same adversary and the same chokepoints, and a hedge that pays in either world is still a hedge. What P2025 changed is the probability mass: it put the supplier-decapitation scenario in writing, and the observable build rate roughly tripled after its playbook started becoming policy.

Objection 04
Data mirage

Apparent-demand arithmetic on opaque data can manufacture a wedge. Build estimates span 0.4 to 1.1 mb/d; teapot quota cycles and petchem feedstock shifts distort the residual. The phantom may be measurement error.

Response

The spread is real and disclosed on every figure here. But independent methods converge: EIA balances, OIES flow tracking, and satellite tank gauging (Kayrros, Kpler, Vortexa) all show multi-hundred-million-barrel accumulation. Error bars change the wedge's size, not its existence. May 2026 was the out-of-sample test, and a 3.9 mb/d cut cleared it.

05

The route forward

Regime update · Jul 8 2026 · escalation

The June reopening has fractured. On Jul 7 Iran struck three tankers on the US-Navy-protected southern (Omani) lane — the Qatari LNG carrier Al-Rekayyat and the Saudi VLCC Wedyan among them — and the US retaliated against air-defense and anti-ship-missile sites. Treasury revoked GL X, the waiver that had let Iran sell oil. The Joint Maritime Information Center raised the Hormuz threat to "severe." The strait is now two contested corridors: a southern lane under attack and an Iran-approved northern lane. June throughput ran ~4.3 mb/d against 15–20 pre-war. Brent jumped +3% to $74, +5.6% to $76 after-hours, off four-month lows near $70.

Two structural confirmations for the thesis landed at once. First, China has confirmed it returns as an aggressive buyer in the $65–75 spot zone — the price-sensitive, state-directed demand the wedge predicted, now observed. Second, China drew inventories before it resumed buying, which reveals the drawdown floor it will defend: a realistic comfort minimum, almost certainly still far above true operational minimums. JPMorgan pegs ~3 mb/d of the import cut as temporary, reversing from August.

The naive read says less Chinese buying means lower Brent. That holds in one of three configurations, and we now have the missing piece on tape: China is a confirmed price-sensitive buyer in the $65–75 band and a domestic-production offsetter above it. Supply state sets the level, Chinese elasticity sets the cap while stocks last, Chinese restock policy sets the floor, and a political suppression bid caps rallies into November. The base case is a managed 75–100 range; the tails are fat.

Config 1 · Mar–Jun 2026, now exited
Buying less because barrels don't exist

Blockade removes the supply; low imports and high prices coexist because the cut is the shock absorber, not the shock. China's elasticity was the price cap: why Brent peaked at $118 rather than $200+ and held near $95 into June. This configuration governed Mar–Jun; the tape exited it in late June as the strait reopened.

Config 2 · The trapdoor, now live
Buying less by choice

Post-normalization, if Beijing skips the restock: demand-led market, and the intuition is right. Consensus models still carry ~1 mb/d of "demand" that was never consumption. Low imports, low price. June ran this play: record-low buying into a reopening market, depressing the price Beijing will restock at.

Config 3 · The floor
The restock bid

Now confirmed, not inferred: China buys hand over fist in the $65–75 zone and stands aside when price is elevated. A floor and a volatility compressor, not a driver. It drew stocks before buying, exposing the inventory floor it defends — a comfort minimum well above true operational minimums, so the latent bid is deeper than the observed one. The 2023–25 phantom era was this configuration running continuously; the difference now is we have the trigger band on tape.

Base: imports + output
Est. true consumption
Bear: no restock
Bull: restock race
Restock wedge
16151413 H2 '2620272028 RESTOCK WEDGE (BASE) bull base bear MB/D CRUDE · PROJECTION · ALL PATHS ARE SCENARIO, NOT DATA
The import-threshold ladder

Post-normalization run-rates, crude imports. The first two clean months of customs prints identify the regime within ~60 days; with the August restock pulse JPMorgan expects, that means the Aug–Sep prints, released Sep–Oct.

War floor May '26 print
7.8
No restock maintenance only; consensus demand overstated 1.0–1.3
10.2
24-mo rebuild ~250M bbl over two years
10.6
12-mo rebuild ~250M bbl in one year
11.0
Buffer doctrine reserve deemed too small; 2–3 yrs; recreates the phantom optics
11.0–11.6
Reference: phantom-era average 11.1–11.5 · storage headroom exists: above-ground tanks were only 56% filled (Kayrros, Mar '25)
The demand arithmetic, crude basis
True crude consumption, pre-war (refinery runs + direct burn)14.5–14.7
of which crude processed for product exports0.9–1.1
domestic-use crude consumption13.5–13.7
Total liquids demand, CNPC peak (2025)15.4
Maintenance imports (consumption − 4.3 domestic output, zero stock change)10.2–10.4
Actual imports, 2023–2511.1–11.5
Standing overshoot = the phantom0.8–1.2

Wartime check: run cuts (−0.6), product export bans (−0.3) and substitution (−0.2) put consumption at 13.5–13.8, import requirement 9.2–9.5, actual buying 7.8–8.1, implied draw 1.1–1.7 mb/d. Consistent with the drawdown wedge. Cumulative draw through July ≈170–190M bbl; a grind into Q1 '27 reaches 400–500M, where the 60–90 day working-stock floor binds.

Bear · ~25–30% · the phantom unwind, capped
Brent 60–72 · contango · China defends the floor

Clean reopening; Beijing skips the restock; de-oiling becomes doctrine. OPEC+ spare (5–6 mb/d pre-war) returns, Venezuela ramps through authorized channels, Americas supply grows. The market then prices what this page argues: true consumption ≈14.5 and falling, and the ~1 mb/d stockpile bid that absorbed every surplus since 2022 is gone. Balance flips toward surplus. But the floor is now known, not hypothetical: China buys hand over fist under $75 and Trump wants no crash that dents US shale before November. So the unwind stalls in the low-60s rather than collapsing; the $52–60 tail needs the MoU to fully hold and China to stay out, which the confirmed bid argues against.

Cracks: the product-export ban is the trigger. Fully lifted, Chinese refined exports jump 88–160% from H1 levels (JPM) and Singapore gasoil compresses to $5–10. Atlantic diesel holds $15–20 on European closures. Jet the relative winner, the only Chinese product still growing. Expression: long complex Atlantic refiners against short flat price.

Base · ~45–50% · the managed range
Brent 75–100 · China out high, in low

The dominant path: bounce around 75–100 with China absent when price is elevated and buying hand over fist on every dip toward $75, while Hormuz stays a quagmire of unpredictable headlines. The August restock pulse lands (~3 mb/d, part state-directed; imports recovering toward 10.5–11.2); the suppression bid caps rallies into November; the US, Japan, Korea and India rebuild against thin cover (US stocks lowest since 2004, Cushing near its ~20M floor). Hard floor at the China bid, soft ceiling at the political line: fade rips, buy dips.

Cracks outlast crude: product stocks drew faster than crude, Chinese run cuts removed product supply, and the export ban stays partially in place. Diesel $14–22, jet $16–24, gasoline $10–15, normalizing through mid-2027. Complex margins stay good, off the wartime peak.

Bull · ~25–30% · break + squeeze
Brent 100–140 on a sustained break

The re-escalation is live: Jul 7 tanker strikes on the US-Navy lane, GL X revoked, JMIC at "severe." A sustained Hormuz closure or failed MoU removes 4–5 mb/d of Gulf-origin crude into working-floor-risk inventories. Then the accelerant: a record-adjacent short base violently covers, China's sub-$75 bid competes for scarce barrels, and the suppression ceiling breaks upward. Coordinated global restocking of 1.5–2.5 mb/d layers on top. This is the fat tail, and fatter than the tape given positioning.

Cracks blow out first: the CL2-CL3 spread already snapped back positive and the 3-2-1 (RBOB/HO) crack is near cycle highs. Diesel $35–50, jet spiking on military demand plus lost Gulf kerosene. Then demand destruction rolls cracks over before flat price; cracks weakening while Brent holds is the exit signal.

The sign-flip rule, applied. The supply hole sets the price level, Chinese elasticity sets the cap while stocks last, and Chinese restock policy sets the floor after normalization. True consumption trends down in every branch (CNPC dates the peak to 2025 at ~15.4 mb/d liquids; transport fuels decline 1–2% a year; petrochemicals grow ~5% on light feedstocks). The phantom supported the 2023–25 tape, the unwind is the bear case nobody is positioned for, and a buffer-expansion doctrine would recreate the illusion a second time. Watch the Aug–Sep customs prints against 10.2 / 10.6 / 11.0 / 11.5; the CFTC managed-money net for a covering trigger; the product-export ban (fully lifted, exports jump 88–160%); and the political calendar into Nov 3. The base case is a managed 75–100 range with China buying dips and standing aside on rips. The asymmetry is up: a supply break forces the short crowd and the suppression bid to unwind together.

06

The positioning powder keg

Two forces sit under the price that have nothing to do with barrels: a speculative short base built to an extreme during the slide to four-month lows, and a political imperative to keep gasoline cheap into the November midterms. Both are asymmetric. Both point the same direction on a shock.

01 · the crowded short
A violent unwind is the bull tail

The slide from $100+ to sub-$70 pulled managed money to a heavily net-short, bearishly-skewed book as the glut narrative took hold. Positioning that stretched is fuel, not a driver: it does nothing until a catalyst arrives, then it amplifies. The Jul 7 tanker strikes and the GL X revocation are exactly that kind of catalyst — Brent's +5.6% after-hours jump is a preview. A disorderly short-covering squeeze into thin summer liquidity, on top of a confirmed sub-$75 China bid and dangerously low inventories, is the mechanism by which the bull case turns violent rather than gradual.

CFTC COT (managed money) · CNBC
02 · the suppression bid
The state has a thumb on the scale until November

The political economy is explicit. Berenberg, Jul 3: ahead of the Nov 3 midterms, Trump wants low oil; losing the House and Senate raises impeachment risk. The toolkit is real — jawboning OPEC+ (which just added 188 kb/d), SPR mechanics, sanctions timing (the Iran waiver was a supply lever), pressure on allies, and, by your read, Treasury's capacity to lean on futures. Saudi cutting Arab Light to Asia by $11/bbl fits a coordinated soft-price posture. The effect is a managed ceiling that caps rallies but cannot manufacture barrels; when physical tightness overpowers it, the release is sharper for having been suppressed.

Berenberg · OPEC+ · Aramco OSP · author
The setup

Stack them: a record-adjacent short base, a state ceiling that holds price down into November, a confirmed Chinese bid that switches on under $75, inventories at working-floor risk, and a chokepoint that just re-escalated. That is a compressed spring. The suppression and the short crowd both borrow stability from the future; a genuine supply break — a sustained Hormuz closure, a failed MoU, a teapot-sanctions hit to Chinese refiners — forces both to unwind at once. The base case is quiet; the tails are fat and the upside tail is fatter than the tape implies.

07

Where the thesis lands

Position

The stockpiling-inflated-demand mechanism is real and quantifiable: ≈1 mb/d of builds, effectively 100% of marginal import growth in '24–'25, cumulating to ~1.2B barrels since 2022 (reserves ~0.5B → ~1.7B into the war). The honest attribution is general war-proofing against the world Project 2025 described: sanctions risk, supplier decapitation, Taiwan optionality. Beijing hedged the distribution of outcomes; the tail event happened. Two caveats bound the trade. The ~1.7B barrels cover ~150 nominal days but only 60–90 days of working stock before refiners cut runs. And the import cut was partly forced, not purely chosen. Then the picture sharpened on both sides. The confirmations: China buys hand over fist in the $65–75 band, offsets with record domestic output near 4.5 mb/d, and drew stocks before buying, exposing a comfort floor above true operational minimums. The escalation: Jul 7 tanker strikes on the US-Navy lane, GL X revoked, Hormuz threat "severe." Base case is a managed 75–100 range — China out on rips, in on dips, a political ceiling into November, Hormuz a permanent wildcard. But the asymmetry is up: a record-adjacent short base, thin inventories and a re-escalating chokepoint mean a supply break unwinds the suppression bid and the shorts together. Quiet base, fat tails, upside tail fattest.