The Energy Information Administration's August 2026 Short-Term Energy Outlook , released on August 11, puts one number at the centre of this winter's natural gas story. The outlook states that “we expect natural gas inventories to be a record 3,985 billion cubic feet (Bcf) at the end of October 2026,” a figure “which is an increase of 19 Bcf compared with the July STEO and 5% above the five-year average.” Ten years earlier the same commodity produced almost the same reading. On the EIA's weekly Lower 48 working gas series , the last October report of 2016, for the week ending October 28, recorded 3,963 Bcf. The two figures sit 22 Bcf apart, about half a percent. What changed is everything the stock has to be measured against. A stock figure is a numerator. On its own it is a volume, not a condition. This piece takes up a narrow question: when the numerator is flat across a decade and the denominator is not, which denominator should a reader use, and do the pl...
The Energy Information Administration's August 2026 Short-Term Energy Outlook, released on August 11, puts one number at the centre of this winter's natural gas story. The outlook states that “we expect natural gas inventories to be a record 3,985 billion cubic feet (Bcf) at the end of October 2026,” a figure “which is an increase of 19 Bcf compared with the July STEO and 5% above the five-year average.”
Ten years earlier the same commodity produced almost the same reading. On the EIA's weekly Lower 48 working gas series, the last October report of 2016, for the week ending October 28, recorded 3,963 Bcf. The two figures sit 22 Bcf apart, about half a percent. What changed is everything the stock has to be measured against.
A stock figure is a numerator. On its own it is a volume, not a condition. This piece takes up a narrow question: when the numerator is flat across a decade and the denominator is not, which denominator should a reader use, and do the plausible candidates agree? They do not, and the disagreement is where the information sits.
Two Series, Not One
Before any arithmetic, the two figures in the 3,900s need separating. The weekly number belongs to the Lower 48 working gas series, dated to a Friday and published each Thursday; as of the August 27, 2026 release, its most recent entry was 3,184 Bcf for the week ending August 21. The outlook's number is a month-end inventory that the STEO describes as U.S. working natural gas in storage, without specifying Lower 48. They are close relatives rather than the same measurement, so nothing below subtracts one from the other.
That distinction also caps how precisely the remaining injection season can be described. From 3,184 Bcf on August 21 there are ten more weekly reports before the end of October, running from August 28 to October 30. Closing a gap of roughly 800 Bcf across ten reports implies an average of about 80 Bcf a week. Treat that as an order of magnitude, not a target: the endpoint comes from a different series than the starting point.
What the Outlook Says Is Driving the Build
The outlook is explicit about causation, and the sentence matters because it runs against the intuition the rest of this piece might otherwise invite. It reads: “Our lower price forecast reflects reduced LNG feedgas demand and record natural gas production, which we expect will leave natural gas inventories at their highest level heading into winter since 2016.” The sentence immediately after it supplies the price consequence: “We expect Henry Hub prices to rise gradually in the coming months but remain relatively low because inventories are well above the five-year average.”
Two qualifications travel with that clause. First, the reduction in feedgas demand has a dated physical cause, which the outlook names: “Maintenance at Freeport LNG began on July 10 and is expected to be completed in late August, affecting 2.0 Bcf/d of nominal export capacity in the short-term.” It adds that “Maintenance at Freeport and other LNG export terminals have reduced feedgas demand on the Gulf Coast in June and July.” The same section puts third-quarter 2026 LNG exports at 16.5 Bcf/d, “down 0.2 Bcf/d compared with last month's forecast,” while the outlook's annual table still carries LNG exports rising from 17.4 Bcf/d in 2026 to 18.6 Bcf/d in 2027. Second, the accompanying press release of August 11 frames the build in the language of comfort, quoting the Administrator: “More natural gas in inventories in the fall season provides a cushion for increased heating-related consumption during the winter.” The cushion is real. The thing being cushioned is also larger, and the release does not scale one against the other.
The Denominator Rose 45 Percent
The natural denominator for a pre-winter stock is the offtake it will face. EIA publishes both components monthly. Total U.S. natural gas consumption covers everything burned domestically; total exports covers everything that leaves by pipe or by ship. Their sum is what left the domestic system in a month; dividing by the days in the month gives a rate.
January is the stress month, so each October stock should be paired with the January that followed it. January 2017 is the January that the 3,963 Bcf of October 2016 actually had to face. Domestic consumption that month was 2,913,823 million cubic feet, or 2,913.8 Bcf, and total exports 269,960 million cubic feet: together 3,183.8 Bcf across 31 days, a rate of 102.7 Bcf/d. In January 2026, the most recent January on the books, consumption was 3,789,224 million cubic feet, or 3,789.2 Bcf, and exports 833,493 million cubic feet, giving 4,622.7 Bcf and 149.1 Bcf/d. The rate rose 45.2 percent while the pre-winter stock rose 0.6 percent.
Most of the export half of that increase is one thing. In January 2017, liquefied natural gas exports were 51,342 million cubic feet, or 51.3 Bcf, a rate of 1.7 Bcf/d; the same series had read 26 million cubic feet in January 2016. In January 2026 the same series read 539,203 million cubic feet, or 539.2 Bcf, a rate of 17.4 Bcf/d that sits alongside the outlook's 17.4 Bcf/d annual average for 2026. Non-LNG exports, mostly pipeline flows, went from 218.6 Bcf in the month to 294.3 Bcf. Of the 18.2 Bcf/d increase in the total January export rate, LNG supplied 15.7, roughly six-sevenths. Of the 46.4 Bcf/d increase in the total January offtake rate, LNG supplied about a third and domestic consumption about three-fifths.
The Balance Identity Closes
That offtake arithmetic can be checked against a figure EIA publishes independently. In any month, offtake minus the supply that production and imports delivered should approximate the net withdrawal from storage, which the Lower 48 net withdrawal series reports.
In January 2017, dry natural gas production was 2,187.4 Bcf and imports 291.7 Bcf, giving 2,479.1 Bcf of supply against 3,183.8 Bcf of offtake. The gap is 704.7 Bcf; the reported net withdrawal was 672.7 Bcf, a residual of 4.8 percent. In January 2026, production of 3,371.5 Bcf and imports of 339.8 Bcf gave 3,711.3 Bcf against 4,622.7 Bcf of offtake, a gap of 911.4 Bcf against a reported net withdrawal of 903.5 Bcf, a residual of 0.9 percent. The residual is EIA's balancing item and nothing here attempts to decompose it. The check matters only because it confirms both sides describe the same month.
Days of January Offtake, Winter by Winter
Define the ratio plainly. Take the last October reading of the weekly working gas series in year Y, divide by the January offtake rate of year Y+1, and the result is the number of days of that January's offtake the pre-winter stock was equal to. EIA does not publish this ratio. Eleven completed winters are available on the current vintage of both series.
| Winter | Oct stock, Bcf | Jan offtake rate, Bcf/d | Days of offtake | Jan net withdrawal, Bcf | January draws |
|---|---|---|---|---|---|
| 2015-16 | 3,931 | 105.2 | 37.4 | 728.7 | 5.39 |
| 2016-17 | 3,963 | 102.7 | 38.6 | 672.7 | 5.89 |
| 2017-18 | 3,775 | 117.5 | 32.1 | 893.5 | 4.22 |
| 2018-19 | 3,143 | 122.2 | 25.7 | 705.5 | 4.46 |
| 2019-20 | 3,695 | 123.8 | 29.9 | 566.1 | 6.53 |
| 2020-21 | 3,919 | 125.8 | 31.2 | 705.5 | 5.56 |
| 2021-22 | 3,611 | 135.3 | 26.7 | 991.1 | 3.64 |
| 2022-23 | 3,501 | 126.6 | 27.7 | 454.8 | 7.70 |
| 2023-24 | 3,779 | 142.1 | 26.6 | 841.5 | 4.49 |
| 2024-25 | 3,864 | 148.9 | 26.0 | 1,008.3 | 3.83 |
| 2025-26 | 3,905 | 149.1 | 26.2 | 903.5 | 4.32 |
The series falls, but not smoothly. The two highest readings are the first two winters, 37.4 and 38.6 days. The lowest is 25.7 in the winter of 2018-19, when the stock itself was the smallest in the table at 3,143 Bcf. The compression at the end is more telling. Since the winter of 2021-22 the ratio has sat in a narrow band, 26.7, 27.7, 26.6, 26.0 and 26.2, while the stock behind those readings ranged from 3,501 to 3,905 Bcf. Once the January offtake rate passed roughly 135 Bcf/d, ordinary variation in the stock stopped moving the ratio much. The denominator had taken over.
Applying the outlook's 3,985 Bcf to the January 2026 offtake rate gives 26.7 days. January 2027 has not happened, so this holds the rate at its most recent observed value; a colder January or higher feedgas would push the figure lower, a milder one higher. Against the 38.6 days of the 2016-17 winter, that is a decline of 11.9 days, or 30.7 percent, for a stock 0.6 percent larger.
The Other Scaling Says Something Different
Days of offtake is not the only defensible denominator, and the alternative is less flattering to the argument. Storage does not have to cover all offtake. It covers only the part that production and imports do not, which is exactly the net withdrawal used in the balance check above. Dividing the same October stock by the following January's net withdrawal gives a second ratio: how many Januaries of actual storage draw the pre-winter stock was worth.
That ratio has no trend worth the name. It was 5.89 for the winter of 2016-17 and 4.32 for 2025-26, but in between it fell to 3.64 in 2021-22 and then rose to 7.70 the very next winter. Two consecutive winters differing by more than a factor of two is not a signal about the size of the demand base; it is a measure of how hard each individual winter drew on storage. On the offtake scaling those same two winters read 26.7 and 27.7, barely apart at all.
The reason the two scalings diverge is supply. Dry production in January 2017 ran at 70.6 Bcf/d; in January 2026 it ran at 108.8 Bcf/d, an increase of 54.1 percent against the 45.2 percent rise in the offtake rate. The demand base grew, and the supply that meets it directly, without ever passing through a cavern, grew slightly faster. Reported January net withdrawal rose only 34.3 percent between those two Januaries, from 672.7 Bcf to 903.5 Bcf. Both statements are true at once: the stock covers far fewer days of total offtake than it used to, and it still covers a comparable number of actual January draws.
Where This Doesn't Apply
The offtake ratio is a scaling device, not a scarcity measure. Nothing in it says the United States is short of gas this winter, and the outlook's own price path points the other way. It answers one question only: whether a pre-winter stock that reads the same as a decade ago sits in front of the same job. It does not.
Three limits deserve stating. First, the base year is EIA's choice, not a neutral one. January 2017 carried the lowest domestic consumption of the eleven Januaries in the table, 2,913,823 million cubic feet. A low denominator raises the 2016-17 ratio and so flatters the decline that follows it. Measured instead from January 2016, the offtake rate rose 41.8 percent rather than 45.2 percent, and the ratio fell from 37.4 days to 26.7, a decline of 28.5 percent rather than 30.7.
Second, the ratio says nothing about deliverability. A cavern holding a given volume and a cavern able to release gas at a given daily rate are different constraints, and a cold snap tests the second one. No figure in this piece measures it.
Third, the 2026-27 entry is a forecast on one side and a held constant on the other. The 3,985 Bcf is a projection that already moved 19 Bcf between the July and August editions, and the January 2027 offtake rate is unknown and simply frozen at the 2026 value. Treat that bar as a construction, not an observation.
What to Watch Next Week
- The weekly storage reports. The working gas table names September 3, 2026 as its next release date, with another the following Thursday. What matters is not the headline injection but whether the pace is near the roughly 80 Bcf a week implied by the gap between 3,184 Bcf on August 21 and an end-October figure in the high 3,900s.
- The September Short-Term Energy Outlook, scheduled for September 9. The 3,985 Bcf figure gained 19 Bcf between the July and August editions. A further increase widens the numerator; a cut narrows it. Either way the offtake denominator will not move in the same release.
- The Natural Gas Monthly on September 30. July 2026 data lands then. LNG exports are the component of offtake with the least weather sensitivity, so a July reading against the 3Q26 forecast of 16.5 Bcf/d is the cleanest early check on whether the outlook's feedgas assumption is holding.
- The gap between the two export narratives. The August outlook trimmed 3Q26 LNG exports while keeping annual figures of 17.4 Bcf/d for 2026 and 18.6 Bcf/d for 2027, and the April export note projected that “In 2027, net exports increase another 10% to 20.5 Bcf/d.” Watch whether the near-term trim survives into the annual numbers.
Concrete Framework
A monitoring checklist for keeping this ratio current rather than taking it on trust.
- Pull the numerator from one series and stay in it. The weekly Lower 48 working gas table and the STEO month-end inventory are different measurements. Mixing them inside a single subtraction produces a number that belongs to neither.
- Rebuild the denominator every January. Add that month's total consumption and total exports, divide by 31. Both publish on the same monthly schedule, next dated September 30, 2026.
- Compute both ratios, not one. Days of offtake tracks the structural demand base; Januaries of draw tracks what storage was actually asked to do. When they point the same way the reading is strong. When they diverge, as they do now, the divergence is the finding.
- Check the balance identity before trusting either. Offtake minus production minus imports should land within a few percent of the reported net withdrawal. In January 2017 the residual was 4.8 percent and in January 2026 it was 0.9 percent. A residual far outside that range means a series revision, not a discovery.
- Separate the base year from the finding. Any decade comparison inherits the weather of both endpoints. Run the arithmetic from at least two starting winters and report the range, as the 41.8 and 45.2 percent figures above do.
- Keep deliverability out of the conclusion. Volume ratios cannot answer questions about peak-day flow. If the question is what happens during a two-week cold snap, this framework is the wrong tool.
Comments
Post a Comment