U.S. Crude Oil Inventories Since 1982: The Complete Weekly Record

Every Wednesday at 10:30 a.m. Eastern, the U.S. Energy Information Administration publishes the most argued-over number in oil markets: how much crude oil is sitting in American storage tanks. That weekly release has been running since August 1982, and it adds up to 2,297 observations of U.S. commercial crude oil stocks. This page is the whole record, computed end to end from the EIA's official weekly series (WCESTUS1, the same feed this site refreshes for its markets page). The headline results: stocks peaked at 540.7 million barrels in June 2020 and bottomed at 247.3 million in January 2004; the current reading of 424.1 million barrels (week ending October 2, 2026) is higher than roughly four out of every five weeks in the 44-year record; and the Strategic Petroleum Reserve, at 283.0 million barrels, has been lower in only six weeks since 1982, all of them from the year the series begins. Every table below can be cited as-is. The year-end table in section 11 is the complete dataset in print form.

1. The record at a glance

Start with the six numbers everything else hangs off. Commercial crude means crude oil held by refiners, tank farms, and pipelines, excluding anything the government holds in the Strategic Petroleum Reserve (SPR). That is the series traders mean when they say "inventories," and it is the series the EIA's Weekly Petroleum Status Report leads with.

Source: EIA weekly petroleum stocks series WCESTUS1 (commercial crude oil, excluding SPR), 2,297 weekly observations, August 20, 1982 to October 2, 2026, computed from the official EIA API on October 8, 2026. Five-year range from the site's weekly inventory feed. A step-by-step guide to reading the Wednesday release itself is our how-to-read-the-weekly-petroleum-report explainer.

2. How this dataset is built, and where its edges are

The figures on this page come from three EIA weekly series. WCESTUS1 is U.S. ending stocks of crude oil excluding the SPR, in thousand barrels. WCSSTUS1 is the crude oil held in the SPR itself. W_EPC0_SAX_YCUOK_MBBL is stocks at Cushing, Oklahoma, the delivery point for the WTI futures contract. The first two series run the full span used here: 2,297 consecutive weekly observations from Friday, August 20, 1982 through Friday, October 2, 2026. The Cushing series starts April 9, 2004 and contributes 1,174 observations. Averages, percentiles, decade means, records, and weekly changes on this page are arithmetic on those series, run on October 8, 2026. Nothing is interpolated and nothing is seasonally adjusted; the EIA publishes levels, and levels are what is tabulated here.

Three edges deserve stating rather than footnoting. First, the EIA revises weekly figures as late reports arrive, so a level pulled today can differ by a few hundred thousand barrels from the same week's first print; the records below are computed on the current vintage of the series, which is the right basis for comparing across decades. Second, "commercial" is a moving fence: the series counts crude at refineries, in tank farms, and in pipelines, and excludes the SPR, so total U.S.-held crude is the commercial figure plus the SPR figure (707.1 million barrels on October 2, 2026, if you add the two). Third, percentile ranks treat every week since 1982 as an equal draw, which flatters recent levels because the market now moves far more oil per day than it did in 1985. Days-of-supply framing corrects for that, and it is covered in the weekly report explainer rather than repeated here.

3. Five decades, five different oil markets

Average inventory by decade is the fastest way to see that "normal" has moved. The 1980s and 1990s look alike: stocks in the low 300s, drifting with the size of a market that imported most of its crude. The 2000s are the tight decade, with the lowest average in the record at 293.7 million barrels and a year-end print of 254.5 million in 2003. Then the shale era reprices what storage means. The 2010s average 396.6 million, about 103 million above the 2000s, and the 2020s average 445.5 million, the highest of any decade in the series. Half of that 2020s lift is the 2020 spike working through the mean; the other half is a market that produces roughly 13 to 14 million barrels a day and holds proportionally more oil in the system on any given Friday.

The threshold counts make the regime change sharper than the averages do. Stocks have closed at or above 450 million barrels in 251 weeks. Every one of those weeks falls in the 2010s (135) or the 2020s (116). At or above 500 million, the count drops to 60 weeks, clustered in the 2016-2017 shale glut (34 weeks) and the 2020-2021 pandemic build and drawdown (26 weeks). Before 2015, a 400-million-barrel print was itself rare enough to make news; the year-end table in section 11 shows the level crossing 400 for the first time at year-end in 2015 and never closing a year below it since. On the low side, 464 weeks (about one week in five) sit below 300 million, and they are almost all pre-2010 weeks. The practical read: anyone quoting a "normal" inventory level needs to say which decade's normal they mean, because the series carries at least three.

4. The low-water mark: January 2004 and the tight decade

The cheapest oil of the modern era, in storage terms, was the oil that was not there. Commercial stocks fell through the early 2000s as demand grew and U.S. production kept sliding, and the series touched its all-time floor of 247.3 million barrels in the week ending January 23, 2004. The neighborhood of that print is as striking as the print itself: 254.2 million in October 2002, 254.5 million in the last week of 2003, 255.8 million in February 2003, 257.1 million in February 2004. Four of the five lowest readings in 44 years landed inside seventeen months. Year-end levels tell the same story at annual grain: 262.1 million to close 2002, 254.5 million to close 2003, and no year-end print above 310 million anywhere in the decade until 2009 closed at 308.0. A market running on 250 to 300 million barrels of commercial cover had very little slack for a refinery outage or a hurricane, which is part of why price spikes in that era could be violent: the buffer was thin before the shock arrived, not after.

For calibration against the present: at 424.1 million barrels, current stocks are 71.5 percent above the 2004 floor. Put differently, the entire drawdown-era cushion problem has inverted. The question the Wednesday print answers now is rarely whether there is enough crude in the system; it is whether the stock level is rising or falling fast enough, in the right region, to say something about refinery demand and export pull. The level records in this section are the background that makes the weekly change legible.

5. The shale glut: 2015 to 2017

The largest sustained build in the record arrived with shale. U.S. output grew faster than pipelines, export rules, and refinery configurations could absorb, and storage absorbed it instead. Year-end stocks jumped from 353.0 million barrels at the close of 2014 to 455.1 million at the close of 2015, a 102.1 million barrel rise in twelve months and the biggest year-end jump in the series (the next largest is not close: 2020 added 63.6 million year over year into the pandemic peak). 2016 closed at 479.0 million, still the highest year-end print on record. The weekly peak of that era, 535.5 million in the week ending March 31, 2017, stood as the all-time record until 2020 passed it. Of the 60 weeks in history at or above 500 million barrels, 34 belong to this window: 12 in 2016 and 22 in 2017.

The glut also reset the stock geography. Cushing, Oklahoma filled toward tank tops and set its own record of 69.4 million barrels in the week ending April 7, 2017, within days of the national peak. The Cushing delivery-point explainer covers why that one tank farm carries outsized weight in the WTI contract. The drawdown that followed was slow: stocks ended 2017 at 424.5 million and 2018 at 441.4 million, and the level did not revisit the low 400s on a sustained basis until the 2020s. Anyone who traded the era remembers the phrase "storage is the shock absorber." The data version is simpler: between the last week of 2014 and the spring 2017 peak, the United States added roughly 182 million barrels of commercial crude to storage, an increase of more than half the 2000s decade average, in twenty-seven months.

6. Spring 2020: the fastest fill on record

The all-time peak was not a slow glut. It was an emergency. Commercial stocks stood at 455.4 million barrels in the week ending March 20, 2020. Thirteen weeks later, in the week ending June 19, 2020, they printed 540.7 million, the highest reading in the series, 85.3 million barrels added in a single quarter while demand had collapsed and supply was still catching down. The weekly steps were violent by the standards of this series: a 19.3 million barrel build in the week ending April 10, 2020 (the second largest single-week build ever), and 518.6 million by April 17 on the way up. The same episode drove WTI futures negative in April 2020, a price event inseparable from the storage arithmetic: when tanks at the delivery point approach full, the contract has to pay someone to take the barrels. The WTI-Brent spread explainer traces how that stress showed up in pricing.

The exit took longer than the entry. Stocks ended 2020 at 493.5 million and drew through 2021 to a 417.9 million close, a 75.6 million barrel year-over-year fall that is the largest year-end drop in the record. Since then the series has lived in a narrower band: every year-end print from 2021 through 2025 sits between 415.6 and 431.1 million barrels. The current 424.1 million is inside that band and just below the five-year average of 433.0 million. Measured against the two most extreme episodes in the record, today's market is holding a mid-range tank: full by the standards of 2004, unremarkable by the standards of 2020.

7. The biggest weekly moves in 2,297 weeks

Weekly changes are where the series turns from history into instrument. The average absolute move over 44 years is a few million barrels; the tails are driven by weather, refinery outages, and holidays. The five largest builds and five largest draws in the full record are tabulated below. Two entries deserve their context in the text. The record build, 21.6 million barrels in the week ending February 26, 2021, landed immediately after the Texas freeze knocked out refinery capacity across the Gulf Coast: crude kept arriving at tank farms whose downstream customers were offline, and in the same week gasoline stocks posted their largest draw on record, 13.6 million barrels, because the gasoline could not be replaced. One weather event holds both ends of two tables. The record draw, 17.1 million barrels in the week ending July 28, 2023, came at the height of summer driving season with exports running strong; draws of that size are demand events, not accidents.

Read the dates as a group and a pattern shows: three of the five largest builds are January or February prints, when refinery maintenance season and winter demand lulls let crude accumulate, and both 2023 entries on the build side are rebound weeks after holiday-shortened reporting. The draw side leans the other way, toward late summer and year-end, when driving demand and export loadings run hardest and when a single hurricane week in the Gulf can pull double digits out of storage. None of that makes a given Wednesday predictable. It does mean a 10-million-barrel surprise in February and the same surprise in August carry different information, and the record tables are the reference for deciding which one you just saw. The site's weekly report explainer works through that reading order step by step.

8. The SPR runs a different cycle, and it just set a record of its own

The Strategic Petroleum Reserve is not part of the commercial series, and its history is a policy story rather than a market story. The EIA record begins in August 1982 with the reserve at 270.5 million barrels, still filling toward a design capacity built for a different era. It peaked at 726.6 million barrels in January 2010 and then sat in a narrow 690 to 727 million band for most of a decade: the 2010s average is 690.1 million, nearly the full cavern system held as insurance. The 2020s average, 467.9 million and falling, records the unwind. In calendar 2022 alone the SPR fell from 593.4 million barrels (January 7) to 372.4 million (December 30), a 221.0 million barrel drawdown in fifty-one weeks. Our SPR depletion mathematics deep dive works through the release-rate arithmetic behind moves like that.

Which brings up the freshest finding in this dataset, and it belongs to 2026, not 2022. The five largest single-week SPR drawdowns in the entire 44-year record all landed in May and June 2026: 9.9 million barrels in the week ending May 15, 9.1 million in the weeks ending May 22 and June 19, 8.9 million in the week ending June 12, and 8.6 million in the week ending May 8. The biggest 2022 week, 8.4 million barrels in September 2022, ranks sixth. By the week ending October 2, 2026 the reserve stood at 283.0 million barrels, 61.1 percent below the 2010 peak. In 2,297 weeks of record, only six weeks sit lower, and all six are from 1982, when the caverns were still being filled for the first time. Only 30 weeks in history, twenty-one of them in 1982-83 and nine in 2026, have closed below 300 million barrels. Whatever framework a reader applies to energy security, the input is simple: the emergency stockpile is at its smallest sustained level since its first year of operation, and the fastest weekly drawdowns ever recorded happened this past spring.

9. Cushing: five percent of the barrels, most of the argument

Cushing, Oklahoma held 24.7 million barrels in the week ending October 2, 2026, which is 5.8 percent of the commercial crude total. Its own record, from a series that starts in April 2004, peaked at 69.4 million barrels in the week ending April 7, 2017, days after the national stocks peak of the shale glut, and its decade averages have stepped down from 44.1 million in the 2010s to 32.6 million in the 2020s as pipelines and export routes around the hub multiplied. The current reading sits at roughly the 31st percentile of Cushing's own history: low-normal for the hub, unremarkable next to its 2020 spike, when it added 6.4 million barrels in a single week (April 3, 2020, its largest build) and then shed 5.6 million in a week that May (its largest draw). The reason a 25-million-barrel tank farm leads market coverage while 424 million barrels nationally get a table is the futures contract: WTI settles at Cushing, so Cushing's level is the one that can force a price event when it nears either tank tops or operational minimums. The Cushing explainer covers the delivery mechanics. The point for this dataset is proportionality: track Cushing for contract stress, track the national series for market balance, and do not let a 2-million-barrel move in 5.8 percent of stocks stand in for the other 94.2 percent.

10. The products side of the ledger, and how to use this record on a Wednesday

Crude is the input; gasoline and distillate are what drivers and truckers actually burn, and their stock histories rhyme with crude's without copying it. Gasoline stocks (EIA series WGTSTUS1, 1,915 weekly observations from January 1990) set their record at 263.2 million barrels in the week ending April 17, 2020, the same demand collapse that filled crude tanks, and their low at 178.7 million in September 2008, printed the week after Hurricane Ike shut Gulf Coast refining. Distillate, the diesel and heating oil barrel (WDISTUS1, full 1982 span), is the one series here whose record high is its oldest story: 186.0 million barrels in December 1982, a level set when heating oil dominated winter demand, against a low of 87.2 million in April 1996. Today's product readings sit low in their own histories even as crude sits high: gasoline at 204.7 million barrels (about the 21st percentile since 1990) and distillate at 105.1 million (about the 12th percentile since 1982). That split, comfortable crude cover against lean product cover, is the recurring signature of a system whose bottleneck sits at refineries rather than in the ground or in transit, and it is the first comparison worth making when a product price spikes on a Wednesday headline.

For readers who open the Wednesday release with this page in a second tab, the record reduces to a five-step check. One, place the level before the change: 424.1 million barrels at the 80th percentile means a 5-million draw tightens a comfortable system, while the same draw at the 20th percentile would bite. Two, compare the weekly move with the tails in section 7: anything inside plus or minus 10 million barrels is a large but precedented week; outside that band, something physical happened (a freeze, a hurricane, a holiday calendar quirk) and the report's refinery and trade blocks will say what. Three, check the region behind the headline, starting with Cushing's share: a national draw concentrated at the delivery point moves futures differently than the same draw spread across the Gulf Coast. Four, run the four-week average before reacting to one print; the EIA revises, holiday weeks distort collection, and the series' own history shows single-week records reversing within a month. Five, read the SPR line separately: a commercial draw paired with an SPR release week is a policy-supplied market, and 2026 has shown that the reserve can now post the largest weekly drawdowns in its history, which has not been true of any prior year in the record, including 2022.

Two limits on all of this bear repeating. Stocks are a level, not a flow: the table says nothing about how fast crude is arriving or leaving, which is why the same 424 million can accompany a rising or a falling price depending on the direction of the flows around it. And national totals hide the map: crude stranded in a region without pipeline outlet does not relieve a shortage two states away, the exact failure mode the Cushing record keeps demonstrating. Used inside those limits, the 44-year record does one job well: it converts each Wednesday's number from a headline into a measurement, with a percentile, a precedent list, and a full set of extremes to measure it against.

11. The complete year-end record, 1982 to today

The table below is the dataset the rest of this page summarizes: the last weekly reading of every year since 1982 for commercial crude stocks and the SPR, in million barrels, with the current week appended. Year-end levels are the grain at which annual reviews, textbooks, and market histories usually quote inventories, and they are reproduced here in full so the figures can be lifted directly. Suggested citation: U.S. Energy Information Administration weekly series WCESTUS1 and WCSSTUS1, tabulated by PetroEyes, October 8, 2026. The live series continues on the markets page, and the rig counts page carries the supply side of the same system. What the annual grain shows at a glance: a flat 1980s-1990s shelf in the 275 to 340 range, the tight 2002-2004 trough, a gradual 2005-2014 rebuild, the two-step shale jump of 2015-2016 (455.1, then 479.0, still the highest year-end close), the 2020 spike year at 493.5, the 2021 washout, and the narrow 415 to 431 band the market has closed in every year since. The SPR column reads as an arc: build for twenty-eight years, peak at the 2009-2010 closes (726.1 and 726.5), a decade near the brim, then the 2022 step down of 221.0 million barrels and the continued slide to 283.0, below every year-end close in the table including 1982.

Source: EIA weekly series WCESTUS1 and WCSSTUS1, last observation of each calendar year, computed October 8, 2026 from the official EIA API. Values in million barrels (thousand barrels divided by 1,000), rounded to one decimal. The 1982 row is the last reading of a series that begins August 20, 1982.

12. Frequently asked questions

13. Related reading

• How to read the EIA Weekly Petroleum Status Report: the Wednesday release, block by block, with a worked days-of-supply example.
• SPR depletion mathematics: release rates, refill mechanics, and how fast a drawdown of a given size actually moves the reserve.
• Cushing, Oklahoma: the WTI delivery point: why 5.8 percent of national stocks carries the futures contract.
• Drilled but uncompleted wells (DUCs), explained: the shale inventory that sits between the rig count and production, 2013 to 2024.
• Markets: the current week's inventory, price, and production figures, refreshed weekly from the same EIA feed.

Data: U.S. Energy Information Administration, Weekly Petroleum Status Report series (WCESTUS1, WCSSTUS1, W_EPC0_SAX_YCUOK_MBBL), retrieved via the official EIA API on October 8, 2026. This page is educational analysis of published government data. It is not financial advice and takes no position on any security or commodity contract.