Counting the months in 2026 where the ERCOT monthly energy price moved far enough from its baseline to swing a buyer's bill by more than $500,000 — and showing why the answer is a table, not a number.
$500,000 is not a price move. It becomes one only when you divide it by a volume. For a book averaging 100 MW (~74,000 MWh a month), a $500K swing needs a move of just $6.72/MWh — and five of the seven settled months in 2026 cleared that on day-ahead prices. For a 10 MW customer the same $500K needs $67.20/MWh, and no month in 2026 came close.
Across the plausible mid-market range — 50 to 250 MW average load — the count lands at 3 to 6 material months out of 7, i.e. roughly 0.4 to 0.9 per month. But that average is doing dishonest work: strip January out and the typical monthly deviation collapses from $16.85/MWh to $4.57/MWh.
And $500,000 is not an industry number — nobody publishes one. Section 05 tests the same seven months against thresholds that are citable, including the SEC's codified 10% rule, which needs no volume assumption at all and therefore gives every customer the same answer: 6 of 7 months on day-ahead, 5 of 7 on real-time — but only one of them an overrun.
These are ERCOT-wide average energy prices — the load-weighted average settlement point price for the month — published by Potomac Economics in its role as ERCOT's Independent Market Monitor. Day-ahead and real-time are shown separately because in 2026 they told very different stories.
| Month | Hours | Gas $/MMBtu | DA $/MWh | Δ vs 6-mo avg | Δ vs Jan | RT $/MWh | Δ vs 6-mo avg | Δ vs Jan |
|---|
January 2026 is the whole story. Winter Storm Fern (23–26 January) drove Texas spot gas to a record $30.57/MMBtu on 26 January. Day-ahead cleared at $95.47/MWh for the month — 2.4× the six-month average — while real-time settled at $57.45/MWh, because actual demand came in below forecast and the grid held. That $38.02/MWh gap between what day-ahead buyers paid and what the market actually cleared is the single largest number in the dataset, and it is a basis event, not a level event.
Deviation of each month's price from the January–June 2026 six-month average. The vertical scale is not kind to the other six months — that squashing is the finding.
The whole question reduces to one line of arithmetic. A move of Δ dollars per MWh, applied to V MWh of delivered energy, moves the bill by Δ × V. So the move that costs exactly $500,000 is Δ = 500,000 ÷ V. Put your own load in and the page recounts.
In plain terms: the price sat far enough from the baseline that, once multiplied by the energy actually taken, the bill landed more than half a million dollars away from what the baseline implied. It is a bill test, not a price test — materiality belongs to the customer, not to the market. "Material month" is a term coined for this study; it is not industry vocabulary, and the $500,000 threshold is not an industry number. Section 05 tests the same data against thresholds that are.
Each row is a book size. Each strip shows the seven settled months of 2026 in order — a filled square means that month's price move swung the bill by more than $500,000 against that baseline.
| Avg load | MWh / month | Break-even move | Day-ahead | n | Real-time | n |
|---|
| Avg load | MWh / month | Break-even move | Day-ahead | n | Real-time | n |
|---|
Baseline B is a trap, and worth knowing why. Anchoring on January means anchoring on a storm month. Every subsequent month then looks like a $60–75/MWh miss, and almost any book above 25 MW registers a "material move" every single month. That is arithmetically true and commercially meaningless. If you use a start-of-period reference in a customer-facing model, expect it to produce numbers that flatter the product and won't survive a sophisticated buyer's scrutiny.
The test above uses absolute value, so a month that came in $500K cheaper than baseline counts exactly the same as one that came in $500K more expensive. Split them out and the picture changes a great deal — 2026 has been a cheap year in ERCOT, and every month except January sat below the six-month average.
| Avg load | DA either way | DA overrun | DA saving | RT either way | RT overrun | RT saving |
|---|
This decides how the number can be used. If the proposition is "we protect you from cost shocks", the honest 2026 count at 100 MW is 1 in 7, not 5 in 7 — and against the January baseline it is zero, because no month was more expensive than January. If the proposition is "we prove what you actually paid versus what you should have", two-sided is correct: a buyer who locked at $39.50 and then watched the market trade at $24 was demonstrably worse off, and that is a real claim.
There is no published industry standard that says "a move of $X/MWh is significant." No regulator, rating agency or major consultancy defines one, and the numeric triggers in real C&I supply contracts are bilateral and confidential. But four anchors do exist, and they are citable.
| Anchor | The number | Status | Applied to ERCOT 2026 |
|---|---|---|---|
| SEC Reg S-K Item 305 Instruction 3.A |
≥ 10% of end-of-period price | Codified rule text. The only hard bright line found. | $3.95/MWh day-ahead $3.40/MWh real-time |
| REP unit margin index products |
$3–8/MWh adder | Broker-quoted market practice. Directional. | A move past this erases the whole margin on an index book |
| REP unit margin fixed-price retail |
$26–50/MWh | SEC-filed. Via Renewables, blended ~20 states. | Only January cleared it — and it cleared by $56 |
| SEC SAB 99 SAB Topic 1.M |
5% of pre-tax income | A screen the SEC explicitly refuses to treat as dispositive. | Company-specific; needs the customer's P&L |
The SEC's sensitivity-analysis floor is a percentage of price, not a dollar total, so it needs no volume assumption at all. That removes the flaw that made the $500K question unanswerable as a single number: every customer gets the same count. It is also the number a CFO's auditors already recognise.
| Month | DA move | % of baseline | ≥10%? | RT move | % of baseline | ≥10%? |
|---|
Under the SEC test, 6 of 7 months qualify as significant on day-ahead and 5 of 7 on real-time — but only one (January) is an overrun. April is the only other month that finished above baseline, and only on real-time.
ERCOT and the PUCT do define hard price numbers — they just define extreme, not significant. Worth knowing because they bound the distribution:
| Mechanism | Value | Since | Meaning |
|---|---|---|---|
| High System-Wide Offer Cap | $5,000/MWh | 1 Jan 2022 | Cut from $9,000 after Winter Storm Uri |
| Low System-Wide Offer Cap | $2,000/MWh | 2021 | Flat figure; the old "50× gas index" formula was deleted |
| Peaker Net Margin trigger | $315,000/MW-yr | 2012 basis | 3 × CONE of $105,000. Crossing it switches the cap to LCAP for the rest of the year |
| Unsecured credit | $0 | 1 Oct 2023 | NPRR1112 eliminated it entirely — every dollar of exposure is now collateralised |
| Collateral call trigger | 100% of TPE (advisory at 90%) | current | Post by close of the 2nd business day |
One market-structure change worth flagging. ERCOT's real-time market was redesigned on 5 December 2025 — RTC+B, co-optimising energy and ancillary services, the largest change since 2010. The ORDC price adder is being replaced by per-product Ancillary Service Demand Curves, and the single system-wide offer cap has split into separate day-ahead and real-time caps. Any historical ERCOT price analysis that spans that date is comparing two different markets. This study sits entirely after it, which is fortunate — but the 2025 comparison in section 06 does not.
Retire the $500,000 figure. Use two thresholds instead, and say which one you are using every time you quote a count:
For the defensible, audit-facing claim — the SEC 10% test. It is codified, it needs no volume assumption, and a CFO's finance team already reports market risk this way.
For the commercial claim — a multiple of the customer's own unit margin. This is what a retailer actually means by "significant": a move that eats the margin on the book. For an index product that is $3–8/MWh; for a fixed-price retail book, $26–50/MWh. It reframes the pitch from "prices are volatile" to "this move was N times your margin on that volume", which is a sentence a counterparty can act on.
The evidence that the second framing bites: every ERCOT retailer that failed in 2021 had a unit margin in the tens of dollars per MWh and met a move of $8,000–9,000/MWh sustained for 77 consecutive hours — hundreds of times the entire margin on the book, realised in days. Griddy went to Chapter 11; Just Energy filed a $418.4M one-time hit; Entrust received a single ERCOT invoice near $300M and was rendered insolvent; Brazos was billed $2.1bn. Even NRG, hedged and vertically integrated, took a $967M first-quarter loss.
Seven monthly observations, one of which sits roughly eleven standard deviations from the mean of the other six. You can report this as an illustration of magnitude. You cannot report it as a frequency, a rate, or an expected value, and any investor with a trading background will say so within thirty seconds.
Mean absolute deviation from the six-month average is $16.85/MWh on day-ahead. Remove January and it is $4.57/MWh — a 73% collapse. The break-even load moves from ~68 MW to ~148 MW. Fat-tailed is the honest description; "on average X moves per month" is not.
Running the identical test on 2025's twelve months: at 100 MW, day-ahead crossed $500K in 1 month out of 12; real-time in zero. Only above ~250 MW did 2025 produce regular crossings. So 2026 is not a typical year — but "the year with a storm in it" is exactly the year a risk product has to justify itself on. Caveat: ERCOT's real-time market was redesigned on 5 December 2025, so the 2025 series and the 2026 series are not strictly the same market.
January's $38.02/MWh DA–RT gap is bigger than any month's deviation from the six-month average except January's own. A 100 MW book that scheduled day-ahead paid roughly $2.8M more that month than one settling in real-time, for the same energy. If FortyEight's proposition is evidencing what a buyer actually paid versus what they should have, that spread is a more defensible entry point than monthly level dispersion.
These are monthly averages. Inside January, ERCOT had hours at the $5,000/MWh cap and hours near zero. A buyer exposed hour-by-hour saw far more than seven decision points — settlement risk is intraday, and a monthly count is a floor on the number of material moves, never a ceiling.
ERCOT-wide average energy price for each month of 2026, day-ahead and real-time, read from the Independent Market Monitor's monthly reports. Seven months are settled and published (January–July); August is not yet available.
Baseline A is the arithmetic mean of January–June 2026 — one fixed strike for the whole period. Baseline B is the January 2026 price, held flat. Both were requested; both are reported.
Δm = Pm − baseline, in $/MWh, for each month m.
Vm = average load (MW) × actual hours in month m × load factor. Hours used: Jan 744, Feb 672, Mar 743 (DST), Apr 720, May 744, Jun 720, Jul 744.
A month counts when |Δm × Vm| > $500,000. The count is reported as hits out of comparable months, and as a per-month rate.
Solid: the source is ERCOT's own Independent Market Monitor, the series definition is unambiguous, and the same figures were read from three separate report vintages (the January, June and July 2026 monthly reports) with agreement to the cent on every month except January real-time, which rendered as $57.43 / $57.45 / $57.48 across vintages. $57.45 is used here; the ±$0.05 spread changes no conclusion.
To harden before external use: these values were read off charts in published PDFs rather than pulled from raw settlement data. Before any of this goes to an investor or a design partner, re-derive the series from ERCOT's own Historical DAM and RTM Settlement Point Price files, which give hourly prices by hub and load zone. That also unlocks the two things this analysis cannot do at monthly ERCOT-wide granularity: zone-specific prices (Houston, North, West and South diverge materially) and hourly exposure, which is where the real distribution lives.
Two things this section could not find, stated plainly. First, no regulator, rating agency or major consultancy publishes a named benchmark for what wholesale price move is "significant" to a retailer or a C&I buyer — the anchors above are borrowed from adjacent disciplines, not lifted from an energy standard. Second, no public C&I supply agreement discloses a numeric material-adverse-change or repricing trigger; those are bilateral and confidential. Industry blogs quote swing/bandwidth tolerances of 10%, 20% and 25% with equal confidence and no primary source, so "±10% is standard" should not be repeated as fact. Neither gap is a research failure — both are genuine absences, and both are arguably an opportunity.