Pick your state. A typical new heat pump is cheaper to run than a modern gas furnace in only 4 of the 43 states for which the government has published a 2025 gas price. In the other 39 it costs you more to run. Which side of the line your house falls on is decided by two numbers that are not on your quote: the price of electricity and the price of gas where you live.More
Then read the next sentence, because it decides whether any of that applies to you. The finding above is about GAS. Piped gas heats 59.4 million American homes, so it is a lot of people, but it is not most people. Another 44.4 million homes heat with electricity and have no heat pump, which means resistance heat, and on RUNNING COST a heat pump beats that in every state: it is the same electricity and two to three times the heat. If that is you, the uncomfortable half of this page is not about you. The one honest caveat is that about 2.3 million of those 44.4 million heat with portable electric heaters, which means no ducts and possibly no panel for one, so the running-cost win is real but the capital cost is a different question and this page does not price it.
A typical new heat pump is cheaper to run than a modern gas furnace in only 4 of the 43 states
for which the government has published a 2025 gas price. In the other 39 it costs you more to run. That is not an
opinion and it is not a model: it falls straight out of EIA's own electricity and gas prices and the physics of
the two machines.
More
Here is the arithmetic, and you can check it. A heat pump's running cost is electricity
divided by its seasonal COP. A furnace's is gas divided by its efficiency. Set them equal and you get the COP a
heat pump has to BEAT where you live: (electricity $/kWh / 3,412) x (0.95 x 100,000 / gas $/therm). In
Florida that comes to 1.74, so almost any heat pump wins. In Michigan it is 5.67 and in Alaska 6.09, and nothing on
the market gets close.
A stronger machine does not fix it. A top-tier cold-climate unit (COP 3.22) is
cheaper to run than gas in 16 of 43 states.
Now: does any of this apply to you? The finding is about GAS, and the Census housing survey says
exactly how many people that is. Piped gas heats 59.4 million American homes. It is a lot of houses and it is not
most of them. Another 44.4 million homes heat with electricity and have NO heat pump, which means resistance heat,
and for every one of those a heat pump is a straightforward win: it drinks the same electricity and delivers two to
three times the heat. Only 15.1 million homes, about one in nine, already run one. So the country splits nearly in
half and the honest version of this page has to tell you which half you are standing in, rather than shout the half
that makes the finding sound biggest.
And the rest of the case for a heat pump, which we are not going to bury. Everything above compares
running cost against a MODERN CONDENSING GAS FURNACE, which is the hardest thing a heat pump can be asked to beat,
and we picked it on purpose. Against oil and propane a heat pump wins comfortably. It is also your air
conditioner, so if you were buying AC anyway you are replacing two machines with one, which this page does
not price. And running cost is not emissions: on a clean grid a heat pump is far better for carbon even where it
costs more to run, and that is a real reason to buy one that we cannot put a price on.
What is not true is the sentence you will be sold: that a heat pump will cut your heating bill.
If you are on resistance heat, it will, and by a lot. Against gas, in most states, it will not.
- A typical new heat pump (HSPF2 8.1) is cheaper to RUN than a 95% gas furnace in only 4 of the 43 states for which EIA has published a 2025 gas price. In the other 39 it costs more. On the last complete gas year, where all 50 states are in, it is 3 of 50.
More
This is running cost, not purchase price, and it is against the hardest comparison a heat pump faces. The breakeven ranges from a COP of 1.74 in Florida, where almost any heat pump wins, to 6.09 in Alaska, where nothing on the market does.
- Even a top-tier cold-climate heat pump (HSPF2 11, COP 3.22) is only cheaper to run than gas in 16 of 43 states. Buying a stronger machine does not fix it.
More
The reason is that the breakeven is set by the ratio of two prices, not by the machine. Where electricity is dear and gas is cheap, no commercially available COP closes the gap.
- The arithmetic is physics, and you can check it. The COP a heat pump must beat is (electricity $/kWh divided by 3,412 Btu) times (furnace efficiency times 100,000 Btu divided by gas $/therm).
More
That is it. There is no model, no assumption and nothing of ours in that formula: it is two EIA prices and two efficiencies. Everything else on this page is commentary.
- If you have NO GAS LINE, throw all of the above away. Against electric resistance heating a heat pump wins in every state, usually by more than a factor of two, because it is drinking the same electricity and getting two to three times the heat out of it.
More
This is the case for a great many homes, particularly in the South, and it is where a heat pump is an unambiguously good buy. Change the fuel selector above and the whole page flips. We would rather show you that than let this read as an argument against heat pumps.
- A heat pump is also your air conditioner, and this page does not price that. If you were going to buy AC anyway, you are replacing two machines with one.
More
That can dominate the whole decision, and it is a capital argument rather than a running-cost one. We compare running costs because that is what can be sourced; the capital comparison depends on your ductwork and your panel, and it cannot be done from a web page.
- Running cost is not emissions. On a clean grid a heat pump is far better for carbon even in the states where it costs more to run.
More
That is a real reason to buy one and it is not a reason we can put a price on, so we are saying it plainly rather than leaving a hole where it should be.
2 more
- This page does not tell you about tax credits, and you should be suspicious of any page that does without showing you where it got the dates.
More
Credits change, they have end dates, and a salesperson quoting you a net price after a credit that has expired is quoting you a price you will not get. We had a confident paragraph about the federal credit here and we removed it, because when we went to check our own citation we found we had deleted it. Everything else on this page is traceable to a federal file we can name. That one was not, so it is not here. Check the credit with the IRS before you let anyone subtract it from a quote.
- The one thing on this page that could move the count in the heat pump's favour, and we are telling you rather than waiting to be caught: HSPF2 is rated in a single climate zone, and we apply it to Alaska and Florida alike.
More
A heat pump does better than its rating in a mild winter and worse in a hard one, because the colder it gets outside the less heat there is to move. Our four winning states (Arkansas, Florida, Georgia and Washington) all have mild winters, and in a mild winter the real seasonal COP runs a little ABOVE the label, so a more careful climate-by-climate model might find a fifth or a sixth. It would also push the cold states further out of reach, which is where 39 of the 43 already are. We could not source a state-by-state seasonal COP from a free federal file, so we use the label, name the limitation, and let you see that it points the one way that would soften our own headline.
Sourced: both prices, from EIA. Electricity is the Electric Power Monthly, Table
5.6.A, residential, by state, April 2026. Gas is EIA's 2025 annual residential average by state. The breakeven COP is
arithmetic on those two numbers and the two machines' efficiencies.More
Our one conversion, named: EIA publishes gas at dollars per thousand cubic feet and we need
dollars per therm, so we use EIA's own heat content for residential gas, about 1,037 Btu per cubic foot, giving
10.37 therms per Mcf. It is the only number here not read straight off a price table.
Our base case is the version that is KIND to heat pumps, and you should know that before you read the
headline again. We price with an April electricity snapshot and an annual gas average, and you heat your
house in winter, so neither is a heating-season price. We measured both against EIA's own monthly series. Gas is
CHEAPER in the heating season, not dearer: the annual average is inflated by the summer, where the standing charge
is spread over almost no gas, and December to March comes in at 89% of it. So we are pricing gas above what it
actually costs you to heat with, which flatters the heat pump. April electricity runs 6% above the winter it should
be compared with, which cuts the other way but by less. Correct both and a typical heat pump wins in
3 states rather than 4. The finding survives the correction that goes against us, which is the
only reason we are willing to print it. These are national seasonal factors applied to every state, not a
state-by-state measurement of your winter.
Ours: the annual heat demand, which is a box. It scales the saving up and down and it does not
change which way the comparison goes, because that is decided entirely by the two prices and the two efficiencies.
The machine costs about $4,450. The job does not have a price we are willing to print, and that is a
different sentence from "nobody publishes one". DOE and NREL keep a database of efficiency measures, and
in it are the listed prices of 85 air-source heat pumps at Home Depot, the AC Outlet, Lowe's and HVAC Direct. The
middle of that list is $4,450 for the machine, with the middle half running $2,553 to $5,729. That is what the box
costs, and it is a price you could go and pay today. It is not what the job costs.
And here is the part we got wrong, twice, and are not going to hide. This page originally said
that no free federal source publishes an installed price for a heat pump. That was false. The very same DOE file
has an "Installed Cost" column with 24 air-source models in it. We had already downloaded that workbook. We already
cite it by name on our kitchen page. We simply never opened the sheet, and then announced that the thing we had not
looked for did not exist.
Then the fix was wrong too. We said we would take the material price because it was "DOE's own
catalogue" and refuse the installed cost because it came from Homewyse and RSMeans. But nothing in that file is
DOE's own: DOE compiled other people's listings. And the 24 rows carrying an installed cost turned out to be
exactly the Homewyse and RSMeans rows, so we were taking the machine price off the very rows whose labour
we were refusing, and calling that a principle.
The honest rule is about what kind of number it is, not whose name is on it. A retailer's listed
price is an OBSERVED price: Home Depot has a heat pump and a price on it. Homewyse and RSMeans are cost-ESTIMATE
services, which do not observe your job, they model it. That is somebody's guess with a logo on it, and reprinting
it because it happens to sit inside a government workbook is exactly the laundering we exist to avoid. So we now
drop those 24 rows from BOTH columns, because a rule you apply to one and not the other is not a rule. The median
does not move either way. The point is that the criterion now separates what it claims to separate.
One caveat we will not bury: a listed price is not a transaction. Nobody in that file paid
anything. These are prices asked, not prices met, and they are HSPF 7.5 to 8.2, which is the old rating scale.
What the Census does and does not have, since we looked. The CONTRACTOR who fits it: 248 product
lines for the plumbing, heating and air-conditioning class, and not one of the 248 carries a quantity. Dollars, no
jobs, so no price per job. Worth knowing about that class while we are here, because it is the one your installer
is counted in: it is only 40.9% actual HVAC work, the rest being plumbers, mechanical contractors
and sprinkler fitters. The FACTORY: heat pumps get their own product lines, but the Census collects no unit count
for the air-source kind at all, and withholds the dollar figure for the air-source-only line. The WHOLESALER: an
$8.4bn line called "wholesale sales of heat pumps", also with no unit count. And the AMERICAN HOUSING SURVEY, which
our HVAC page is built on: a heat pump, a gas furnace and a boiler are all job code 27, "added or replaced built-in
heating equipment", so it can price the job and never the machine.
One number did fall out of it. Exactly one heat pump has a published factory-gate price, and it is
the water-source kind: $648,262,000 across 382,455 units, so $1,695 each leaving the factory. A water-source unit
is a commercial water-loop machine, not the air-source box that would hang beside your house, so we are not going
to dress it up as your equipment cost. We mention it because the distance between a number like that and a
five-figure quote is mostly ductwork, electrical, distribution and labour, and not the machine.
Where every number above comes from
-
Energy prices
EIA, Electric Power Monthly, Table 5.6.A: average retail price of electricity by state, residential (April 2026)
eia.gov ↗
-
Energy prices
EIA, price of natural gas delivered to residential consumers by state, 2025 annual
eia.gov ↗
-
US Census
Census, American Housing Survey 2023 (national): how 133 million American homes are heated, by system and by fuel
www2.census.gov ↗
-
DOE / NREL, National Residential Efficiency Measures Database: the MATERIAL price of 109 air-source heat pumps. We take the machine and refuse the install cost in the same file, because its labour is sourced from Homewyse and RSMeans
data.openei.org ↗
-
US Census
Census, 2022 Economic Census, product lines by industry (EC2200NAPCSINDPRD): searched for an installed price and a factory-gate price for a heat pump
www2.census.gov ↗