Crude Draws 640K vs 1.6M Est — API Said +7.1M and EIA's Balancing Item Swung 10M Barrels

Fundamentals · 2026-09-16

Commercial crude stocks fell 640,000 barrels to 423.4 million in the week to 11 September, against a 1.6 million draw expected — a smaller draw than forecast, and the second consecutive week the draw has undershot. The API had reported a 7.14 million barrel BUILD for the same week, a roughly 7.8 million barrel divergence. EIA's own balancing item swung 1,465 kb/d week on week — 10.3 million barrels, sixteen times the size of the headline. Underneath: crude exports 4,831 kb/d, up 1,414; refinery runs 17,330 kb/d, down 256, with utilization 96.8%; Cushing 21.5 million, down 342,000 and -8.8% y/y; SPR 285.0 million, -29.8% y/y. Products: gasoline +794,000 (-4.6% y/y), distillate +1.585 million but -13.5% y/y. Production 13,944 kb/d. Next release 23 September.

What It Changes

• The headline is bearish and close to meaningless this week. A 640,000 barrel draw against a 1.6 million forecast is a bearish miss, but EIA's unaccounted-for adjustment moved by 10.3 million barrels over the same week. The balancing item is sixteen times the headline. Read the flows, not the number.
• The flows are the opposite of the headline, and they are bullish. Crude exports jumped 1,414 kb/d — nearly 10 million barrels over the week — as the Saudi outage pulled US barrels abroad. Refineries ran at 96.8% utilization. Cushing fell again, to 21.5 million, and is down 8.8% on the year.
• API and EIA disagree by 7.8 million barrels, and that has to resolve. The API reported a 7.14 million build; EIA reported a 640,000 draw. The two were within 90,000 barrels of each other the week before, so this is not a standing methodology gap — it is a genuine one-week conflict.
• The strategic buffer keeps shrinking into a supply crisis. The SPR released another 403,000 barrels and now sits 29.8% below where it was a year ago, 120.8 million barrels lighter, with the East-West pipeline outage unresolved.
• It lands three and a half hours before the FOMC, and it will not survive the 14:00 decision as a market driver. The information value is the export and Cushing picture, not the print.

Impact

• Crude (CL) — Mixed, lean bearish on the print and bullish underneath — a smaller draw than expected is a bearish headline, but exports surging and Cushing draining are not bearish facts.
• The Open had WTI settling $105.83 on Tuesday, up 4.38% on the Yanbu loading suspension, and trading around $103.41 overnight as part of that spike unwound. A 960,000 barrel miss against consensus is a rounding error next to an unresolved 4-to-5 million barrel-a-day supply disruption.
• Cushing at 21.5 million is the number to hold onto. It is the NYMEX delivery point, tank bottoms sit near 20 million, and it has fallen for a second week to 8.8% below last year. Thin delivery-point inventory is what turns a supply scare into a squeeze at expiry.
• The conflicting reports argue for treating this week's inventory data as low-confidence in both directions. When API says build 7.14 and EIA says draw 0.64 in the same week, the honest position is that neither number should carry a position on its own.
• USD — Neutral — inventories do not price the dollar, and today belongs to 14:00.
• US Indices (ES / NQ / YM / RTY) — Neutral to slight bearish — the transmission is through fuel costs and the inflation path rather than through the print. Futures came into the session +0.2% to +0.4%.
• Gold (GC) — Neutral — no mechanism worth writing down from a weekly inventory report.

Inside The Number

Commercial crude oil inventories excluding the SPR fell 640,000 barrels to 423.4 million in the week ended 11 September, against a consensus of a 1.6 million barrel draw. Stocks are 1.9% above the same week last year. Including the SPR, total crude fell 1.043 million barrels to 708.4 million, which is 13.7% below a year ago.

Before anything else, the caveat that should govern how you read this week. EIA's crude balance contains a line called the adjustment — unaccounted-for crude, the residual that makes production plus net imports plus transfers minus refinery runs equal the observed stock change. This week it printed +577 kb/d. Last week it printed -888 kb/d. That is a 1,465 kb/d swing, or 10.26 million barrels over a single week, against a headline commercial draw of 640,000 barrels.

Run the balance without it and the week looks entirely different. Production 13,944, net imports 2,227, transfers 433, refinery input 17,330: that arithmetic gives a stock change of -726 kb/d, or a 5.08 million barrel draw. The adjustment is what turned a five-million-barrel draw into a one-million-barrel one. The balancing item is not a footnote this week; it is the number.

The API disagreed violently, and that is unusual. The American Petroleum Institute, whose survey is voluntary and published the evening before, reported a 7.14 million barrel build for the same week. EIA reported a 640,000 barrel draw. That is a 7.8 million barrel divergence. Note what makes it stand out: the week before, API reported a 300,000 barrel draw and EIA reported 391,000 — within 91,000 barrels. So this is not a persistent methodological offset that can be mentally adjusted away. One of the two is wrong about this week by a very large margin, and next Wednesday's revision to the prior week is where it starts to resolve.

Now the flows, which tell a coherent story the headline does not. Crude exports rose 1,414 kb/d to 4,831 kb/d — nearly 10 million barrels over the week — the clearest single move in the report. Imports rose 234 kb/d to 7,058, so net imports collapsed by 1,180 kb/d to 2,227. With Saudi Arabia's East-West pipeline down and loadings suspended at Yanbu, the marginal barrel is being pulled out of the US Gulf toward buyers who have lost their usual supply. That is the mechanism connecting this week's geopolitics to this week's tank levels, and it is why stocks fell at all in a week when refineries ran *less*.

Refineries eased off a very high base. Crude input fell 256 kb/d to 17,330 kb/d and utilization slipped a point to 96.8%. Put that in context: utilization was 93.3% this week last year and 92.1% the year before. At 96.8% the system is running about as hard as it physically can, which is what you would expect with crack spreads where they are and diesel at a record. Gross inputs fell 180 kb/d to 17,449. The Gulf Coast and Midwest each cut 107 kb/d.

Cushing is the number with the most leverage. Stocks at the Oklahoma hub fell 342,000 barrels to 21.482 million, now 8.8% below a year ago and 5.4% below two years ago. Cushing is the delivery point for the NYMEX WTI contract, and operational tank bottoms are generally reckoned at around 20 million barrels. At 21.5 million there is not much cushion between the current level and the point where physical delivery gets awkward. This is the second consecutive weekly decline.

Regionally the draw was a Gulf Coast story, which fits the export read: PADD 3 fell 1.315 million barrels, the largest regional move, while the West Coast built 1.010 million and Alaska in-transit rose 1.308 million. The Midwest fell 441,000 and the East Coast 193,000.

The SPR continues to drain, and the annual figure is the one that matters. Another 403,000 barrels left the reserve, taking it to 284.957 million — down 120.8 million barrels, or 29.8%, from a year ago. The strategic buffer has shed close to a third of its volume in twelve months, and it is being drawn down during an active Middle East supply disruption rather than being rebuilt ahead of one.

Products built, but the annual comparisons are what carry the information. Gasoline rose 794,000 barrels to 207.7 million yet remains 4.6% below last year. Distillate rose 1.585 million to 107.9 million and is still 13.5% below last year — a build from a badly depleted base, in the same week diesel set a record price. Jet fuel fell 695,000, propane fell 1.378 million, residual fuel rose 968,000.

One number to distrust. Total products supplied — EIA's implied demand proxy — jumped 1,942 kb/d to 21,255 kb/d, which reads as a demand surge. But 1,588 kb/d of that came from "other oils", the residual catch-all and by far the noisiest line in the report. Finished motor gasoline supplied rose a modest 247 kb/d to 8,798, and distillate supplied actually fell 177 kb/d. Products supplied is a calculated residual, not a measurement, and a week where the catch-all provides 82% of the move is a week to wait for the four-week average.

The Internals

Stocks, in millions of barrels, week ended 11 September:

Series · This week · Last week · Change · Year ago · vs year ago

Commercial crude excluding SPR · 423.429 · 424.069 · -0.640 · 415.361 · +1.9%

Strategic Petroleum Reserve · 284.957 · 285.360 · -0.403 · 405.728 · -29.8%

Total crude including SPR · 708.386 · 709.429 · -1.043 · 821.089 · -13.7%

Cushing, Oklahoma · 21.482 · 21.824 · -0.342 · 23.561 · -8.8%

Total motor gasoline · 207.732 · 206.938 · +0.794 · 217.650 · -4.6%

Distillate fuel oil · 107.859 · 106.274 · +1.585 · 124.684 · -13.5%

Kerosene-type jet fuel · 45.338 · 46.033 · -0.695 · 43.899 · +3.3%

Propane and propylene · 109.092 · 110.470 · -1.378 · 98.931 · +10.3%

Residual fuel oil · 23.018 · 22.050 · +0.968 · 20.798 · +10.7%

Total stocks excluding SPR · 1,251.248 · 1,248.641 · +2.607 · 1,282.421 · -2.4%

Commercial crude by region, in millions of barrels:

Region · This week · Change · vs year ago

Gulf Coast, PADD 3 · 246.172 · -1.315 · +3.7%

Midwest, PADD 2 · 99.663 · -0.441 · -3.2%

Cushing, within PADD 2 · 21.482 · -0.342 · -8.8%

West Coast, PADD 5 · 45.989 · +1.010 · +2.2%

Rocky Mountain, PADD 4 · 23.409 · +0.298 · +3.6%

East Coast, PADD 1 · 8.195 · -0.193 · +11.2%

Alaska in transit · 3.737 · +1.308 · -19.4%

Supply and demand flows, thousands of barrels per day:

Measure · This week · Last week · Change · Read

Crude exports · 4,831 · 3,417 · +1,414 · The dominant move in the report

Crude imports · 7,058 · 6,824 · +234 · Up, but swamped by exports

Net crude imports · 2,227 · 3,407 · -1,180 · Collapsed on the export surge

Crude input to refineries · 17,330 · 17,586 · -256 · Easing from a very high base

Refinery utilization · 96.8% · 97.8% · -1.0pp · Was 93.3% a year ago

Domestic crude production · 13,944 · 13,947 · -3 · Flat, and +462 on the year

EIA adjustment · +577 · -888 · +1,465 · Unaccounted-for crude

Total products supplied · 21,255 · 19,313 · +1,942 · 82% of it from "other oils"

The balance, and why the adjustment dominates it:

Component · kb/d · Over the week, million barrels

Domestic production · +13,944 · +97.6

Net imports · +2,227 · +15.6

Transfers to crude supply · +433 · +3.0

EIA adjustment · +577 · +4.0

Crude input to refineries · -17,330 · -121.3

Published stock change, all crude · -149 · -1.04

Same balance with adjustment set to zero · -726 · -5.08

Week-on-week swing in the adjustment alone · +1,465 · +10.26

Products supplied, the implied-demand detail:

Product · This week · Change · Read

Total products supplied · 21,255 · +1,942 · Looks like a demand surge

Other oils · 5,877 · +1,588 · The residual catch-all; 82% of the total move

Finished motor gasoline · 8,798 · +247 · Modest, and flat against last year

Propane and propylene · 1,038 · +381 · Seasonal and volatile

Kerosene-type jet fuel · 1,800 · +15 · Steady, +4.4% on the four-week average

Distillate fuel oil · 3,501 · -177 · Fell, despite record diesel prices

Residual fuel oil · 241 · -112 · Small and noisy

Where the two surveys disagree:

Source · Week to 11 September · Week to 4 September · Read

EIA commercial crude · -0.64 million · -0.391 million · The official series

API commercial crude · +7.14 million · -0.30 million · Voluntary survey, published a day earlier

Gap between them · 7.78 million · 0.09 million · Last week they agreed; this week they do not

Where The Barrels Actually Went

Exports did the work, and nothing else comes close. Crude exports rose 1,414 kb/d to 4,831 kb/d — an extra 9.9 million barrels shipped out over the week. Set that against the entire commercial draw of 640,000 barrels and the scale is obvious: the export surge is fifteen times the size of the headline. With Saudi Arabia's 4-to-5 million barrel-a-day East-West pipeline down and Red Sea loadings suspended at Yanbu, buyers who normally take Saudi crude are bidding for Atlantic Basin barrels, and the US Gulf is the marginal seller.

The Gulf Coast confirms it. PADD 3 holds 246.2 million barrels, and it fell 1.315 million — the largest regional decline and more than the national commercial draw. Nearly all US crude export capacity sits on that coast. Meanwhile the West Coast, which exports almost nothing, built 1.010 million barrels.

Refineries were not the reason stocks fell. Crude input dropped 256 kb/d. All else equal, refiners taking less crude *builds* inventory. Stocks fell anyway, which is only possible because the barrels left the country instead. That inversion — draws on falling runs — is the signature of an export-led week rather than a demand-led one.

Cushing is where this gets structurally interesting. Down another 342,000 barrels to 21.482 million, a second consecutive weekly fall, and 8.8% below last year. The hub is the physical settlement point for WTI futures, and the working range between operational minimum and full is much narrower than the headline national number suggests. A hub this thin during a live supply disruption is the setup where the front of the curve can move independently of the national inventory picture.

The SPR is going the wrong way for a supply shock. Another 403,000 barrels out, leaving 284.957 million — 120.8 million barrels, or 29.8%, lower than a year ago. Whatever the policy logic, the practical consequence is that the buffer available to respond to a Middle East disruption is nearly a third smaller than it was the last time this question came up.

On the product side, read the annual column and ignore the weekly. Gasoline built 794,000 barrels and distillate built 1.585 million, which reads comfortable. But gasoline is 4.6% below last year and distillate 13.5% below, and the distillate build arrived in the same week diesel set a record price. Building stock off a base that low, at prices that high, is a thin kind of comfort: it says refiners are prioritizing distillate, not that the shortage is over.

Production is flat and that matters more than it looks. Domestic output was 13,944 kb/d, down 3 kb/d on the week and up 462 kb/d on the year. The US is producing near record volumes and still drawing crude stocks — because exports are taking the incremental barrel. There is no production response available here that would ease a global supply problem; the barrels are already being pumped.

Against This Morning's Open

• This release was not on The Open's calendar, which makes five missing from today's table — The Open carried three rows: NAHB at 10:00, the FOMC at 14:00, the press conference at 14:30. EIA crude publishes at 10:30 every Wednesday.
• On the substance, though, The Open led with exactly the right story. It opened on the 10-year at 2007 highs "after oil surged 4.4% Tuesday," put WTI's settle at $105.83 and the overnight at $103.41, and framed the Saudi disruption as the live catalyst.
• It also flagged the contradiction that this report indirectly speaks to. The Open noted Energy Secretary Chris Wright saying the pipeline could resume "within days," against Reuters sources citing five to six weeks, and said plainly that neither claim is confirmed. A 1,414 kb/d jump in US crude exports is what the market looks like when it is pricing the longer estimate.
• Where the read needs refining: The Open treated the oil move as a price story. This report says it is now also a *flows* story — barrels physically leaving the US, and the WTI delivery hub thinning out.

What This Sets Up

• Next EIA report — Wednesday 23 September at 10:30 ET, covering the week to 18 September, which is the first full week of the Yanbu loading suspension.
• The API-EIA gap has to close, and the revision is where to watch. API said +7.14 million, EIA said -0.64 million. Next week's report revises this week's figures; if the prior week gets marked substantially higher, API was closer and this week's draw was an artifact.
• Whether the adjustment normalizes. A +577 kb/d adjustment after a -888 kb/d one is not a stable series. If it stays large in either direction, treat the weekly headline as unusable and read the four-week averages instead.
• Whether Cushing breaks 21 million. Two consecutive draws have taken it to 21.482 million against operational minimums near 20 million. That is the level where the front of the WTI curve stops following national inventories.
• Whether exports hold above 4.8 million barrels a day. This week's surge is the transmission channel from the Saudi outage into US tank levels. If it persists, draws continue regardless of what domestic demand does.

What Is This?

• What it is: The Weekly Petroleum Status Report, published by the Energy Information Administration every Wednesday at 10:30 ET for the week ending the previous Friday. It reports US stocks of crude oil and refined products, refinery activity, production, imports and exports, and an implied-demand measure called products supplied. The headline traders watch is commercial crude excluding the Strategic Petroleum Reserve. The report also breaks stocks down by the five PADD refining districts and reports Cushing, Oklahoma separately, because Cushing is the physical delivery point for the NYMEX WTI futures contract.
• Why it matters: It is the most detailed and most timely inventory picture for the world's largest oil consumer, and it is the only weekly read on whether a supply or demand shock is actually reaching tank levels. With Saudi Arabia's export infrastructure disrupted and WTI above $100, it is the series that shows whether the disruption is moving physical barrels or only moving price. It is also, through gasoline and diesel, a direct input into the inflation data the Fed is currently focused on.
• How to read it: Four cautions, and the first two decide most weeks. The adjustment can dominate the print — it is a balancing residual for unaccounted-for crude, it routinely runs to hundreds of thousands of barrels a day, and when it swings it can be larger than the headline change itself, as it was this week. Products supplied is calculated, not measured, so treat a weekly jump as noise until the four-week average moves. The API's survey, released the evening before, is voluntary and frequently differs, sometimes by millions of barrels, and neither series should be treated as ground truth when they conflict. And the weekly series is revised, so the cleanest read on direction is the four-week average and the year-on-year column, not the headline change.

Want the full explainer? Economic News Events, Explained breaks down this release and every other one we cover, in plain English.

_For informational purposes only. Not investment advice._