What it costs to run, and which half of that we can prove
Two numbers come out of this page. One is arithmetic on an efficiency threshold and a state’s average electricity price, with no run-hour assumption. The other needs to know how many hours a year your unit runs, which nobody publishes for your house — so it arrives with a band and a label rather than false precision.
Heating in New York, per 1,000 hours of run time
$460
Cooling costs $236 for the same 1,000 hours. Both figures are arithmetic on the efficiency threshold you picked and New York’s average residential price of 29.93 cents per kilowatt-hour (EIA, May 2026, preliminary; historical figure). They assume the unit performs exactly at that seasonal threshold and that its heating capacity equals the head size; no run-hour assumption is involved.
A full year lands near $954, and honestly it could be anywhere from $620 to $1,288. That band exists because the two run-hour coefficients above are ours, not anybody’s published figure. Trust the per-hour numbers more; treat the annual one as a sketch.
Our own assumption converting a state’s cooling degree days into equivalent full-load run hours; we could not obtain a citable hours table at a stable public address.
What moves this most: the heating season; on our assumed run hours this state needs more heating hours than cooling hours, so the HSPF2 figure matters more than the SEER2 figure.
Your state, a standard head size and an efficiency tier
Sets both the electricity price and the weather.
Sizes from our own standard head-size list; use the size the sizing tool gave you.
The same head, five climates
Every row below is a 12,000 BTU/h head at the ENERGY STAR thresholds, run through the same calculation with only the state changed (prices: EIA, May 2026, preliminary, historical; degree days: NOAA, 2025; hours: our assumption). The spread is not about the machine. It is about the price of electricity where you live and how many hours the weather asks the machine to work.
| State | Cents per kWh | Heating, per 1,000 h | Cooling, per 1,000 h | Assumed hours, heat / cool | Annual band |
|---|---|---|---|---|---|
| Maine | 28.63 | $440 | $226 | 2,296 / 230 | $691–$1,435 |
| New York | 29.93 | $460 | $236 | 1,775 / 577 | $620–$1,288 |
| California | 33.25 | $512 | $263 | 758 / 761 | $382–$794 |
| Texas | 16.44 | $253 | $130 | 456 / 2,571 | $292–$606 |
| Florida | 15.17 | $233 | $120 | 185 / 3,114 | $270–$562 |
Maine and Florida are the instructive pair. Maine has more than 12 times Florida’s heating degree days (7,652 against 617 in NOAA’s 2025 statewide series) and a fraction of its cooling; Florida is the mirror image. California’s electricity costs 2.2 times what Florida’s does, so California’s cost per run hour is higher than either. Any national average annual figure is therefore averaging across a spread this wide, which is why this site does not publish one.
The arithmetic, written out
The arithmetic divides the head size by the efficiency threshold and by a thousand to give an average draw in kilowatts, assuming the unit performs exactly at that threshold:
draw (kW) = head size in BTU/h / (efficiency threshold x 1000)cost per 1,000 hours = draw x 1,000 x state price per kWh
A 12,000 BTU/h head at the federal minimum 14.3 SEER2[2] draws 0.84 kW on average while cooling, on that arithmetic. The same head at 15.2 SEER2 [3] draws 0.79 kW. At the national average residential price of 18.44 cents per kilowatt-hour[1] (EIA, May 2026, preliminary), that difference is worth about $9 per thousand cooling hours. Whether that repays the price step depends on your hours, which brings us to the part we cannot prove.
To get from an hourly cost to a yearly one you need equivalent full-load hours. Some utility Technical Reference Manuals publish equivalent full-load hours; we could not obtain one at a stable public address. Rather than quietly picking a number and printing a confident annual bill, this site multiplies your state’s degree days by two coefficients of our own — 0.8 [5] for cooling and 0.3[6] for heating — and puts a 35 percent band[7] around the result. If your utility publishes a load-hours figure for your city, it is a better guide than ours. Our own assumption converting a state’s cooling degree days into equivalent full-load run hours; we could not obtain a citable hours table at a stable public address. The same conversion on the heating side, with the same missing source. The band we place around any annual figure, chosen to reflect how little confidence the run-hour assumption deserves.
One more thing the arithmetic does not answer: HSPF2 is a single seasonal figure; it does not tell you the capacity the unit delivers at 5 F, which is why the cold-climate page exists as a separate check rather than a footnote here.
- Every state, one tablePrice, heating and cooling degree days, and running cost per 1,000 hours for all 51 rows.
- The cold-morning numberWhy a seasonal efficiency figure does not answer the five-degree question.
- Sources and limitsWhere each figure came from, when it was retrieved, and which ones are ours.
Questions people actually ask about this
Are mini-splits cheap to run?
Per hour, the SEER2 threshold implies an average draw of about 0.79 kW for a 12,000 BTU/h head at the ENERGY STAR threshold while cooling. Per year is a different question, and it depends heavily on how many hours the unit runs.
Why do you give a cost per 1,000 hours instead of a monthly bill?
Because the per-hour figure rests on fewer assumptions than a monthly bill. Head size divided by the efficiency threshold gives an average draw in kilowatts; multiply by your state's average price and you have a number with no run-hour assumption in it (it does assume the unit performs exactly at the threshold). Run hours are the guess, so we keep the guess visibly separate.
Where do the annual run hours come from?
From your state's NOAA degree days multiplied by a coefficient we chose ourselves: 0.8 for cooling and 0.3 for heating. We could not obtain a citable equivalent-full-load-hours table at a stable public address, so those two numbers are ours and are labelled assumed everywhere they appear.
Why is there no estimate for Alaska, Hawaii or DC?
NOAA's statewide degree-day series does not cover them. Their electricity prices are published, so the per-hour figures still work; the annual estimate does not, and borrowing a neighbouring state's weather to fill the gap would be inventing data.
Where each figure on this page came from
- U.S. Energy Information Administration, Electric Power Monthly, July 2026 issue, Table 5.6.A (May 2026 data, preliminary) — Historical figures — EIA Electric Power Monthly, July 2026 issue (archived), Table 5.6.A, May 2026 data, marked preliminary by EIA (read 2026-10-05 and saved that day). These May 2026 values were first read on 6 August 2026 from EIA’s rolling table page, which has since moved on to later months and no longer shows them. EIA marks this value preliminary.
- 10 CFR 430.32(c)(5), Energy and water conservation standards (first read 2026-08-06; re-checked 2026-10-05 against a saved copy)
- ENERGY STAR (U.S. EPA), Air-Source Heat Pumps key product criteria (first read 2026-08-06; re-checked 2026-10-05 against a saved copy)
- ENERGY STAR (U.S. EPA), Air-Source Heat Pumps key product criteria (first read 2026-08-06; re-checked 2026-10-05 against a saved copy)
- Assumed: our own labelled figure, no direct source; disclosed on the methodology page.
- Assumed: our own labelled figure, no direct source; disclosed on the methodology page.
- Assumed: our own labelled figure, no direct source; disclosed on the methodology page.
This worksheet’s state list covers 51 jurisdictions; 48 have published degree days for the annual estimate, and the rest carry a price but no degree days.