The Strategic Petroleum Reserve Covers 65 Days or 106, Depending on the Rate

The U.S. Strategic Petroleum Reserve held 286.6 million barrels of crude oil in the week ending August 28, 2026, on the Energy Information Administration's weekly series WCSSTUS1. It is the figure that travels, and the least informative of the three numbers that decide what the reserve can do.

An emergency stockpile is governed by three constraints that are not interchangeable: how much oil sits in the caverns, how fast that oil can physically leave them, and how much of it has already been promised to something that is not an emergency. The first is published every Wednesday. The second is a property of caverns, pipelines and terminals, and the Government Accountability Office publishes two different values for it in the same table. The third is a sales schedule Congress has mandated. Choosing between the two published rates moves the answer to "how many days of cover" by 41.0 days before a single barrel moves.

What the inventory series actually recorded

The monthly series is the cleaner place to start. EIA's MCSSTUS1, released 8/31/2026, shows the reserve at 358.0 million barrels at the end of January 2024, rising in twenty-four of the next twenty-five months - the exception is a 27 thousand barrel dip from June to July 2025 - to a peak of 415.4 million at the end of February 2026. By the end of June 2026 it stood at 322.2 million. Twenty-five months of net accumulation were undone in four.

GAO put a cause on the record. In GAO-26-106918, published May 29, 2026, the agency wrote that "In March 2026, DOE began a 172-million-barrel emergency release planned in response to the war in Iran." That release was the U.S. share of an IEA collective action. GAO adds that "The U.S. percentage share of an IEA collective action release is currently set at just over 40 percent" - and 172 million of the 400 million barrels the IEA members agreed to release is 43.0 percent.

The announced program is a gross plan; the EIA series is a net stock measurement. Between the February peak and the week ending August 28 the reserve fell 128.8 million barrels, which is 74.9 percent of the announced 172 million. It does not follow that 74.9 percent of the program has been delivered. The series nets emergency deliveries against injections, exchange returns and other movements, and does not label them.

Across the five weeks from July 24 to August 28, 2026 the reserve fell 21.0 million barrels - an average of 4.2 million a week, or 601.3 thousand barrels a day - with weekly changes ranging from 2.8 million to 6.1 million and no settled direction.

Two release rates, four answers

GAO states the design requirement plainly: "Consequently, the SPR is designed and required to be able to sustain a drawdown of 4.4 million barrels per day for up to 90 days, and to begin releasing crude oil within 13 days of being directed to do so, though actual market impact may depend on available commercial pipeline, storage, and refinery capacity."

One point follows immediately. The 13 days are an upper bound on the delay, not a fixed one: the requirement says how late the first barrel can be, not how early. A release can begin sooner.

But 4.4 is the design figure, and the same report publishes a second one. Table 2 of GAO-26-106918 gives the reserve's overall drawdown capability as a current effective rate against a design rate: 2.700 / 4.415, or 61 percent. The gap is documented site by site. Big Hill, which carries 1.100 of the 4.415 design rate - 24.9 percent of it - is recorded at zero, annotated "Currently no drawdown capability because of construction outage." Bayou Choctaw and West Hackberry are both marked "Some limitations because of low cavern inventory."

That last annotation matters: part of the shortfall is a function of how full the caverns are, so the release rate is not independent of the level it divides into.

GAO also records that DOE has not revisited its own drawdown and fill rate requirements: "Relatedly, DOE has not reassessed the technical and performance criteria that establish the SPR's operational requirements for drawdown and fill rates to reflect significant changes in crude oil markets over the last three decades."

Days of cover therefore has four answers, and the third constraint supplies the second axis. GAO reports that "After these cancellations, an additional 92.6 million barrels of mandated sales to raise federal revenue remained planned through 2031, as of May 2026." Those barrels are scheduled to be sold to raise revenue, independent of any emergency. Subtract them from the August 28 stock and 194.0 million barrels remain uncommitted, 67.7 percent of the headline.

Against the 90-day endurance the design requirement names, sustained release needs 396.0 million barrels at 4.4 a day and 243.0 million at 2.700. Three of the four cells fall short of 90 days. The one that clears it uses the degraded rate and ignores the sales schedule.

The same report carries a projection the 2026 data have already overtaken: "The reserve was then expected to decrease to a historical minimum of between 330 and 340 million barrels-half of its capacity-after the remaining 92.6 million barrels of congressionally mandated sales are completed at the end of fiscal year 2031." The monthly series crossed below 330 million barrels between the end of May 2026 (355.4 million) and the end of June (322.2 million). A floor projected for 2031 was reached in mid-2026, because emergency releases and mandated sales draw on the same caverns.

The refill side runs on its own rate

Table 2 carries the injection side as well: fill capability of 0.440 against a design rate of 0.785, or 56 percent. The same Big Hill outage removes 0.225 of the 0.785 design fill rate, 28.7 percent of it, and West Hackberry is annotated "Significant limitations because of brine disposal rate."

The observed rates sit below both. From the end of January 2024 to the end of February 2026 the reserve gained 57.4 million barrels across 759 days - 75.7 thousand barrels a day, which is 17.2 percent of Table 2's effective fill capability and 9.6 percent of its 0.785 design rate. On the other side, the 128.8 million barrels that left between February 28 and August 28, 2026 spread over 181 days is 711.8 thousand barrels a day: 26.4 percent of Table 2's effective drawdown capability and 16.1 percent of its 4.415 design rate.

Setting the two observed rates against each other gives 9.4 days of refill for every day of drawdown. Both of Table 2's own ratios are smaller: 4.415 out against 0.785 in is 5.6, and 2.700 against 0.440 is 6.1. The observed asymmetry is wider than either. That gap is not accounted for by the rate table, and GAO does not attribute it. Replacing 128.8 million barrels at 75.7 thousand barrels a day takes 1,703 days, about 4.7 years - into 2031, the fiscal year the remaining mandated sales are due.

What the 90-day benchmark stopped measuring

The familiar yardstick is the IEA obligation: members hold stocks equal to 90 days of the previous year's net imports. That test no longer binds the United States. GAO writes that "As a net exporter, the United States is no longer subject to the 90-day import-based reserve requirement for the foreseeable future but would still be asked to contribute to any IEA collective releases."

The metric that made reserve adequacy legible for decades is undefined for the country holding the reserve, while the obligation to contribute to collective releases survives. Cover then has to be stated as days at a named rate - which is why the pairs 65.1 and 106.1, or 44.1 and 71.9, say more than the barrel count they come from.

History supplies a base rate. GAO records that "The 2022 drawdown required the SPR to achieve just more than 20 percent of its designed maximum drawdown rate (about 1 million barrels per day) and to release about 31 percent (about 180 million barrels) of the SPR's inventory (about 580 million barrels at the time)." The 2026 decline of 128.8 million barrels from a February base of 415.4 million is 31.0 percent - the same share of the reserve, taken from a base about 165 million barrels smaller.

Where This Doesn't Apply

The day counts above are quotients. A lower achievable release rate produces more days on paper and fewer barrels a day into a disrupted market, which is why 106.1 is not better news than 65.1. GAO records the concern as DOE's, not its own: "DOE has identified concerns that significant remaining issues with aging SPR infrastructure will increasingly limit the SPR's capability to meet fill or drawdown directives."

• The 92.6 million is a plan, not a settled fact. GAO's wording refers to cancellations that already reduced the schedule, and Congress can cancel more - which would raise the uncommitted 194.0 million without a barrel moving.
• Exchanges are loans. Where a release is an exchange rather than a sale, barrels are contracted to return with a premium: inventory falls now and rises later, and the series shows only the net.
• Commercial stocks are not zero. Commercial crude excluding the SPR stood at 424.5 million barrels on August 28, 2026, above the reserve itself. Total crude including the SPR was 711.1 million, of which the reserve was 40.3 percent, down from 43.2 percent five weeks earlier.
• The observed rates are records, not forecasts. The 75.7 thousand barrels a day through February 2026 describes what happened, not what is possible. Table 2 records limitations on the fill and drawdown rows at three of the four sites, Bryan Mound excepted, and does not decompose the remaining shortfall.

What to Watch Next Week

• The Wednesday weekly petroleum status report. The question is whether the weekly change on WCSSTUS1 stays inside the 2.8 million to 6.1 million barrel range of the last five weeks, or turns positive.
• Whether commercial crude offsets the reserve. Commercial stocks rose while the SPR fell through August. If both fall together, the 40.3 percent share stops being a composition story.
• DOE notices of sale, exchange or purchase. A purchase solicitation is what would move the observed refill rate off 75.7 thousand barrels a day; its absence is itself information.
• The end of the federal fiscal year. Mandated sales are scheduled by fiscal year, so that boundary is where the 92.6 million barrel schedule can visibly change.

Concrete Framework

A repeatable way to read a reserve number:

1. Name the release rate before saying anything about adequacy. GAO publishes two: the 4.4 million barrels a day design requirement and the 2.700 current effective rate in Table 2.
2. Give both answers, not the convenient one. At 286.6 million barrels the two rates give 65.1 and 106.1 days. The 41.0-day spread is the size of the modelling choice.
3. Subtract committed volumes on a second axis. The 92.6 million barrels of mandated sales cut the uncommitted stock to 194.0 million, which gives 44.1 and 71.9 days on the same two rates.
4. Check whether the announced program and the published series are the same object. A 172 million barrel gross plan and a 128.8 million barrel net decline are not.
5. Read the 13 days as an upper bound. The requirement says how late the first barrel can be, not how early.
6. Re-derive rather than reuse projections. A floor projected for the end of fiscal year 2031 was passed in mid-2026. Projections built on a pre-drawdown baseline expire when the baseline moves.

The reserve is not close to empty, and the data do not support that framing. What they support is narrower: the number that gets quoted moves fastest, while the two that decide outcomes move slowly, are published in more than one version, and are not independent of each other.