Natural Gas +44 Bcf vs +49 Est — 29 Bcf Below the Five-Year Normal, and South Central Drew
Fundamentals · 2026-09-17
Working gas in storage rose 44 Bcf in the week to 11 September against a 49 Bcf consensus, after +40 the week before — a 5 Bcf bullish miss, and the second consecutive week the build has undershot. The seasonal comparison is the story: the five-year average build for this week is +73 Bcf, so this print came in 29 Bcf below normal, and a year ago the same week added 90 Bcf. Total stocks are 3,298 Bcf — 122 Bcf (-3.6%) below last year but 118 Bcf (+3.7%) above the five-year average. South Central drew 5 Bcf, a withdrawal in the middle of injection season, with salt caverns down 24.0% year over year. East (+22) and Midwest (+26) did all the building. At this pace the end-October peak lands near 3,600 Bcf against roughly 3,800 on a normal build. Next release 24 September.
What It Changes
- The build has now undershot the five-year normal by a wide margin two weeks running, and the gap is widening: 12 Bcf below normal last week, 29 Bcf below this week. That is the injection season failing to refill at the usual rate with seven weeks left.
- South Central drew gas in September, which is not what injection season looks like. The region fell 5 Bcf, its salt caverns fell 5 Bcf, and salt storage is 24.0% below where it was a year ago. That is the region that supplies the Gulf Coast LNG export terminals.
- The stock level is genuinely ambiguous and both halves get quoted selectively. 3,298 Bcf is comfortably above the five-year average and clearly below last year. Anyone citing only one of those is making an argument rather than describing the data.
- The trajectory matters more than the level now. Seven weeks of injection season remain. At the five-year average pace the peak lands near 3,800 Bcf; at this week's pace it lands near 3,600. That 200 Bcf difference is what winter gets priced against.
- It is a bullish print into an already tight energy complex — import natural gas prices are up 102.6% on the year, crude settled above $100 this week, and diesel set a record on Monday.
Impact
- Natural Gas (NG) — Bullish — a smaller build than expected, 29 Bcf below the seasonal norm, with the export-adjacent region drawing down. Every component of this release points the same way.
- The salt-cavern number is the one traders should hold. South Central salt storage at 222 Bcf is 24.0% below a year ago and 8.3% below its own five-year average. Salt caverns cycle fast and serve peak demand, so a thin salt position is what turns a cold snap into a price spike rather than a drawdown.
- The counterweight is the level, and it is real. Total stocks remain 118 Bcf above the five-year average and within the historical range. This is a tightening trajectory from a comfortable starting point, not a shortage.
- USD — Neutral — gas storage has no dollar transmission.
- US Indices (ES / NQ / YM / RTY) — Neutral to slight bearish — only through the input-cost channel, and slowly. Utilities and gas-intensive manufacturers are the exposure, not the index.
- Gold (GC) — Neutral — no mechanism worth writing down.
- Crude (CL) — No direct read. Gas and oil storage are separate markets on separate fundamentals; the connection this week runs through the shared supply-disruption backdrop rather than through the barrel.
Inside The Number
Working gas in underground storage in the Lower 48 rose 44 Bcf in the week ending 11 September, taking total stocks to 3,298 Bcf. Consensus was +49 Bcf; the prior week was +40 Bcf. So the build missed by 5 Bcf, and it is the second consecutive week the injection has come in below the survey.
The measurement is precise enough that a 5 Bcf miss means something. EIA publishes sampling variability with every report: the standard error for the total net change this week is 0.9 Bcf, with a coefficient of variation of 0.4% on the stock level. A 5 Bcf gap against consensus is more than five standard errors of measurement — the miss is real, not noise in the survey.
But the consensus comparison is the least interesting one available. The number that matters is the seasonal benchmark. The five-year average build for this week of the year is +73 Bcf. This week delivered 44. That is 29 Bcf below normal, and it follows a week that came in 12 Bcf below normal. The shortfall is not just persistent, it has more than doubled.
Set it against the individual years and the gap is consistent rather than driven by one outlier: 2021 added 83 Bcf in the comparable week, 2022 added 77, 2023 added 57, 2024 added 58, and 2025 added 90. Every one of the last five years injected more gas in mid-September than this year did. Against last year specifically, this week is 46 Bcf smaller.
Now the level, which points the other way and is the reason this release is not a panic. At 3,298 Bcf, stocks are:
122 Bcf below the 3,420 Bcf held a year ago, a 3.6% deficit. And 118 Bcf above the five-year average of 3,180 Bcf, a 3.7% surplus. EIA notes that total working gas "is within the five-year historical range."
Both of those are true simultaneously, and which one gets quoted usually reveals the position of whoever is quoting it. The honest reading is that the US entered this injection season in good shape, has been refilling more slowly than normal for at least a fortnight, and still sits above its five-year norm — for now.
The arithmetic on what remains is the useful part. Injection season conventionally runs to the end of October, leaving roughly seven weeks. At the five-year average build of 73 Bcf a week, storage would peak near 3,809 Bcf. At this week's 44 Bcf, it peaks near 3,606 Bcf. At last week's 40, near 3,578. That spread of roughly 200 Bcf between a normal finish and the current pace is the quantity winter contracts are being priced against, and each week of sub-normal injection narrows the options.
The regional detail is where the release stops being about weather. Two regions did all the building: East +22 Bcf and Midwest +26 Bcf, 48 Bcf between them. Everything else was flat or negative: Mountain +2, Pacific -1, and South Central -5.
A South Central withdrawal in mid-September is the anomaly worth explaining. That region holds 1,039 Bcf, nearly a third of national storage, and it is the region that feeds the Gulf Coast LNG export terminals. It is 11.2% below last year. Inside it, salt caverns drew 5 Bcf to 222 Bcf, and are 24.0% below a year ago and 8.3% below their own five-year average, while nonsalt was exactly flat at 817 Bcf.
That split matters more than the headline. Salt caverns can be filled and emptied many times a season — they exist to meet sudden demand — while nonsalt (depleted reservoirs) cycles slowly and provides baseload. A salt position a quarter below last year means the system's fast-response capacity going into winter is materially thinner than the national number suggests.
Pacific is the mirror image. It drew 1 Bcf on the week and sits 2.4% below last year, but it is 9.9% above its five-year average — the most comfortable region in the country relative to its own norm. Mountain is similar: down 7.3% on the year, up 7.6% against the five-year.
So the regional picture is a country with plenty of gas in the wrong places: surplus in the West against its own history, deficit in the South Central where the export demand is.
Against the week's energy backdrop, the direction is consistent. Import natural gas prices are up 102.6% year over year — more than doubled — per Tuesday's import price data. Crude settled above $100 this week before Aramco began restoring roughly half the damaged East-West pipeline capacity. Diesel set a record on Monday. And the Philadelphia Fed survey this morning had 36% of manufacturers expecting energy markets to worsen as a constraint over the next three months, the most negative forward answer in its entire battery.
This release does not cause any of that. It does mean the one part of the energy complex with a visible, weekly, physical inventory number is refilling more slowly than it should be, with seven weeks to go.
The Internals
The headline, against every benchmark that matters:
Comparison · Value · Gap · Read
This week's build · +44 Bcf · n/a · The print
Consensus · +49 Bcf · -5 Bcf · Second consecutive undershoot
Prior week · +40 Bcf · +4 Bcf · Slightly larger than last week
Five-year average build, this week · +73 Bcf · -29 Bcf · The comparison that matters
Same week a year ago · +90 Bcf · -46 Bcf · Half of last year's injection
Standard error on the net change · 0.9 Bcf · n/a · The 5 Bcf miss is well outside measurement error
The comparable week in each of the five benchmark years:
Year · Week ending · Build · Stocks that week
2021 · 10 September · +83 Bcf · 3,006 Bcf
2022 · 9 September · +77 Bcf · 2,771 Bcf
2023 · 8 September · +57 Bcf · 3,205 Bcf
2024 · 13 September · +58 Bcf · 3,445 Bcf
2025 · 12 September · +90 Bcf · 3,423 Bcf
Average · n/a · +73 Bcf · 3,180 Bcf
2026 · 11 September · +44 Bcf · 3,298 Bcf
Stock levels and the two comparisons that disagree:
Measure · Bcf · vs current · Read
Total working gas, 11 September · 3,298 · n/a · Within the five-year historical range
Total, prior week · 3,254 · +44 · The net build
Year-ago stocks · 3,420 · -122, -3.6% · Below last year
Five-year average stocks · 3,180 · +118, +3.7% · Above the five-year norm
Regional detail, Bcf:
Region · 11 Sep · 4 Sep · Net · vs year ago · vs 5-year average
Midwest · 934 · 908 · +26 · +2.0% · +3.9%
East · 795 · 773 · +22 · +2.2% · +5.7%
Mountain · 241 · 239 · +2 · -7.3% · +7.6%
Pacific · 289 · 290 · -1 · -2.4% · +9.9%
South Central · 1,039 · 1,044 · -5 · -11.2% · -0.2%
South Central salt · 222 · 227 · -5 · -24.0% · -8.3%
South Central nonsalt · 817 · 817 · 0 · -6.8% · +2.4%
Total Lower 48 · 3,298 · 3,254 · +44 · -3.6% · +3.7%
Where the injection season ends, on three different paces:
Pace · Weekly build · Implied end-October stocks · Against a normal finish
Five-year average · +73 Bcf · About 3,809 Bcf · The benchmark
This week's build · +44 Bcf · About 3,606 Bcf · Roughly 200 Bcf short
Last week's build · +40 Bcf · About 3,578 Bcf · Roughly 230 Bcf short
Sampling variability, which EIA publishes and almost nobody reads:
Region · Coefficient of variation on stocks · Standard error on the net change
Total Lower 48 · 0.4% · 0.9 Bcf
South Central · 0.8% · 0.8 Bcf
South Central salt · 1.2% · 0.7 Bcf
Midwest · 0.8% · 0.4 Bcf
South Central nonsalt · 1.0% · 0.4 Bcf
East · 0.6% · 0.3 Bcf
Mountain · 2.9% · 0.3 Bcf
Pacific · 0.0% · 0.0 Bcf
Regional And Facility Detail
Two regions built and three did not. East added 22 Bcf and Midwest added 26 Bcf — 48 Bcf between them, against a national net of 44. Mountain added 2, Pacific drew 1, and South Central drew 5. The regional changes reconcile exactly to the published national figure.
South Central is the region to watch and the reason is structural. It holds 1,039 Bcf, close to a third of all US working gas, and it sits alongside the Gulf Coast LNG export terminals. It is the only major region below its five-year average, at -0.2%, and it is 11.2% below last year. Drawing gas in the middle of September means demand in that region — power burn, industrial load, or pipeline deliveries to liquefaction — is outrunning the supply available to inject.
The salt-versus-nonsalt split inside it is the most actionable detail in the report. Salt caverns fell 5 Bcf to 222 Bcf. They are 24.0% below a year ago and 8.3% below their own five-year average. Nonsalt was flat at 817 Bcf, 6.8% below last year but 2.4% above the five-year average.
Those two facility types do different jobs. Salt caverns can be injected and withdrawn rapidly and multiple times in a season, which makes them the system's shock absorber for cold snaps and demand spikes. Depleted reservoirs — nonsalt — hold far more gas but cycle slowly, and serve as baseload winter supply. A nonsalt position roughly in line with normal and a salt position a quarter below last year means the US has adequate winter volume but noticeably less ability to meet a sudden call on it.
The West is the surplus, and it is a surplus against its own history rather than an absolute one. Pacific holds 289 Bcf, down 2.4% on the year but up 9.9% against its five-year average — the largest relative surplus of any region. Mountain holds 241 Bcf, down 7.3% on the year but up 7.6% against its five-year norm. Both regions are in better shape than their own histories suggest, and neither can send much gas to the Gulf Coast.
East and Midwest are doing the season's work and are in reasonable shape. East at 795 Bcf is 2.2% above last year and 5.7% above the five-year average; Midwest at 934 Bcf is 2.0% above last year and 3.9% above the average. These are the two regions that serve the largest winter heating demand, so a comfortable position there is genuinely reassuring for the season ahead.
On the survey's precision, EIA reports a coefficient of variation of 0.4% on the national stock level and a standard error of 0.9 Bcf on the weekly net change. Pacific is measured essentially exactly, at 0.0% variation. Mountain carries the loosest measurement, at 2.9% — so its +2 Bcf build is the one regional figure this week that could plausibly be a different number. South Central's net change carries a 0.8 Bcf standard error, which means its 5 Bcf withdrawal is a genuine draw rather than a rounding artifact.
Against This Morning's Open
- The Open tabled this release, and the row carried two errors — The Open listed it as "EIA natural gas storage (w/e Sep 12)" with a prior of +49 Bcf.
- The week ended 11 September, not the 12th. EIA's gas storage weeks end on Friday; the report states stocks "as of Friday, September 11, 2026." The Sep 12 date belongs to jobless claims, whose weeks end on Saturday, and which sat two rows above it in the same table.
- The prior was +40 Bcf, not +49. The +49 was the *consensus* for today, placed in the prior column. Our own 10 September brief reported the previous week at +40 Bcf taking stocks to 3,254, and today's EIA release confirms that 3,254 figure exactly.
- This is the right kind of error to catch, because the fix exists. The calendar has been complete on both days since the enumeration step went into the prompt — nine rows today, against three and five omissions on Wednesday. Completeness is solved; the remaining gap is accuracy inside the rows, which is what the second-source pass is supposed to cover.
What This Sets Up
- Next release — Thursday 24 September at 10:30 ET, covering the week to 18 September.
- Whether the sub-normal build becomes a trend or a fortnight. Two weeks at 12 and then 29 Bcf below the five-year norm is suggestive; a third would make the end-October peak arithmetic hard to ignore.
- Whether South Central keeps drawing. A withdrawal in September is unusual. A second consecutive one, with salt already 24% below last year, would be the clearest signal yet that Gulf Coast export demand is outrunning regional supply.
- The 3,600 versus 3,800 question is the whole season. Seven weeks remain. Each week below 73 Bcf moves the likely peak further from a normal finish, and the peak is what the winter strip prices against.
- Watch the salt position specifically into the first cold snap. Salt caverns are the system's fast-response capacity, and at 222 Bcf they are a quarter thinner than a year ago.
What Is This?
- What it is: The Weekly Natural Gas Storage Report, published by the Energy Information Administration every Thursday at 10:30 ET for the week ending the previous Friday. It reports working gas — the volume that can actually be withdrawn and sold — held in underground storage across the Lower 48, broken into five regions, with South Central further split into salt caverns and nonsalt depleted reservoirs. Each figure comes with comparisons against the same week a year earlier and against the five-year average, and EIA publishes sampling variability alongside. The data come from Form EIA-912.
- Why it matters: Natural gas has no meaningful alternative storage and demand is intensely seasonal, so the inventory number is the market's single most important physical signal. The year divides into an injection season, roughly April to October, when storage refills, and a withdrawal season through winter. How much gas is in the ground when injection ends determines how much cushion exists for winter heating and power demand — which is why the trajectory in September matters more than the level.
- How to read it: Four things. Compare the build against the five-year average for that specific week, not against consensus — the survey tells you what traders guessed, the seasonal average tells you whether the system is actually keeping pace. The two level comparisons often disagree, as they do this week, with stocks below last year and above the five-year average; quote both or you are arguing rather than reporting. Salt and nonsalt are different products — salt cycles fast and meets spikes, nonsalt is slow baseload — so a national total can look adequate while the responsive half is thin. And the weekly figure is a survey with published error bars, currently 0.9 Bcf on the national net change, which is small enough that misses of a few Bcf against consensus are real but large enough that single-Bcf regional moves are not worth interpreting.
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._