Heat Pump vs Gas Furnace: Find Your Breakeven COP

Natural gas delivered to U.S. homes averaged $26.24 per thousand cubic feet in August 2025. Seven months earlier, in January, the same series read $12.44 (EIA, U.S. price of natural gas delivered to residential consumers, read 19 August 2026). Pick the wrong one of those to compare a gas furnace against a heat pump and the answer can change sign.

Gas did not cost twice as much in August. Almost nobody burns any in August, and a summer gas bill is mostly the fixed monthly customer charge — the same dollars whether you use one therm or ninety. Spread that charge over a handful of cubic feet and the dollars-per-unit figure falling out the other end is enormous, with the commodity having barely moved.

That matters because the rate you feed into a heat pump comparison decides the answer, and the obvious rate is the wrong one. The 2025 average of that column sits 50% above January's figure, and January is when the furnace is actually running.

Get both fuels into the same unit first

Gas is billed in therms or ccf, electricity in kilowatt-hours, and neither unit is heat delivered into the house.

Three conversions, all from EIA's energy conversion calculators (read 19 August 2026):

• 1 kWh = 3,412 Btu (standard conversion factor)
• 1 cubic foot of natural gas = 1,036 Btu (U.S. average for gas delivered to consumers in 2026, estimate)
• 1 therm = 100,000 Btu, by definition

So 1 therm carries 100,000 / 3,412 = 29.31 kWh worth of heat, and 1 ccf is 1.036 therms.

Use that ccf figure only if you have to. Your utility measures the heat content of the gas going through your meter and prints its own factor on the bill — look for a line labelled therm factor, Btu factor, or heat content, usually beside the ccf reading. It moves with the pipeline and typically sits somewhere near 1.02 to 1.06. Yours is a measurement. EIA's is a national estimate.

The one formula

To deliver 1 MMBtu of heat into the house:

Gas furnace:  10 / AFUE therms   ->  cost = 10 x G / AFUE
Heat pump:    293.1 / COP kWh    ->  cost = 293.1 x E / COP

where G = $/therm (marginal), E = $/kWh (marginal)

Set the two costs equal, solve for COP:

Breakeven COP = 29.31 x AFUE x E / G

That is the whole thing. If your heat pump's seasonal COP lands above that number, it delivers heat for less money. Below it, the furnace is cheaper to run.

AFUE comes off the furnace nameplate. If yours was manufactured on or after 19 November 2015, the federal minimum puts it at 80.0% or better for a non-weatherized gas furnace; the 95.0% minimum does not bind until units manufactured on and after 18 December 2028 (10 CFR 430.32(e)(1), pulled from the eCFR versioner API on 19 August 2026). An old furnace with an illegible plate is the one input worth asking a technician about, because the answer swings the result by nearly a fifth.

Seasonal COP comes from HSPF2 divided by 3.412, since HSPF2 is Btu delivered per watt-hour consumed. The federal minimum for split-system heat pumps manufactured on or after 1 January 2023 is HSPF2 7.5 alongside SEER2 14.3; single-package units are 6.7 and 13.4 (same section, paragraph (c)(5)). ENERGY STAR sets its bar at 7.8 HSPF2 / 15.2 SEER2 / 11.0 EER2 for split systems (ENERGY STAR key product criteria, read 19 August 2026).

Same formula, three states, one yes

Here is why this site will not hand you a national verdict. Below are EIA figures: January 2026 residential gas prices converted at 10.36 therms per Mcf, and May 2026 residential electricity prices from Electric Power Monthly Table 5.6.A (released 23 July 2026, read 19 August 2026). AFUE assumed at 0.80.

The $/therm column is EIA's January 2026 Mcf price — Washington $17.38, Minnesota $11.17, California $22.96, U.S. $13.96 — divided by 10.36 and rounded to the cent, and each breakeven figure is worked from that rounded cent. Redo any row as printed and you get the printed answer.

Washington: 29.31 x 0.80 x 0.1495 / 1.68 = 2.09. A heat pump scraping the federal minimum clears that with room to spare.

Minnesota: 29.31 x 0.80 x 0.1695 / 1.08 = 3.68. Nothing sold as a residential air-source heat pump carries an HSPF2 of 12.6. Against an 80% furnace on those rates, the furnace is the cheaper machine to run, and it is not close.

California has expensive gas and much more expensive electricity, and the ratio parks it beside Minnesota for an entirely different reason. Swap in a 95% condensing furnace and every breakeven figure rises by 18.75% — Washington to 2.48, Minnesota to 4.37 — because a better furnace is a higher bar for the heat pump to clear, not a lower one.

Two warnings about that table. These are averages across all residential customers in a state, so they are neither your rates nor marginal rates; they show that the ranking flips, nothing more. And stripping a fixed gas customer charge out of the $/therm side pushes the breakeven COP up, against the heat pump. That is the realistic case if you are keeping the gas connection for water heating or cooking, because the customer charge outlives the furnace either way.

One more trap, and it is the tempting one, because the figure sits right there in the same column. Average EIA's twelve monthly U.S. prices for 2025 and you get $18.63/Mcf, or $1.80/therm; January 2025 alone is $12.44/Mcf, or $1.20. Hold electricity at 18.44 cents and run the formula both ways: $1.80/therm gives a breakeven COP of 2.40, which most new equipment clears, and $1.20 gives 3.60, which nothing clears. One answer is two-thirds of the other and nothing about the house changed. A column of monthly averages is also not EIA's volume-weighted annual price, and neither of them is the marginal rate you pay for one more therm in February.

Pulling the marginal numbers means reading the tariff, not the summary box on page one. The rate schedule and interval data that decide how many kWh of battery a house actually needs are the same documents where the winter tier boundaries and time-of-use windows live.

The HSPF2 on the label was rated for a climate that is probably not yours

This is the part that gets skipped, and it is written into the test method in plain language.

DOE's Appendix M1 governs how SEER2 and HSPF2 are measured, incorporating AHRI 210/240-2024 by reference. It is the procedure in force for the numbers on today's labels: representations have had to be based on it since 7 July 2025, and it hands over to Appendix M2 only when amended standards written in the new SCORE and SHORE metrics take effect. Section 5.1(b) says that for certifying compliance with the standards, manufacturers must use "the fractional heating bin hours specified for Region IV in table 16 of AHRI 210/240-2024" to determine HSPF2. Section 5.2 then makes representations using any other region's heating hours optional (Appendix M1 to Subpart B of Part 430, 90 FR 1283, pulled from the versioner API on 19 August 2026).

So the HSPF2 you just divided by 3.412 is a bin-weighted average over one designated climate region's modelled heating season. It is not a property of the machine alone, and it is certainly not a measurement of your winter. Two households buying the identical outdoor unit, one in Seattle and one in Duluth, get the same certified number and will not get the same seasonal COP.

Same failure mode as a solar proposal quoting production out of a default weather file: the number is real, the climate behind it belongs to somebody else. Manufacturers may publish HSPF2 for other regions, but section 5.2 means they are not obliged to, so the absence of one tells you nothing.

What neither rating counts

AFUE and HSPF2 each stop short of the meter, and they do it in opposite directions.

The furnace blower is not inside AFUE. It is regulated separately: for a non-weatherized condensing gas furnace fan manufactured on or after 3 July 2019, the fan energy rating ceiling is FER = 0.044 x Qmax + 195 watts per 1,000 cfm (10 CFR 430.32(y), Table 1). At Qmax = 1,200 cfm the ceiling works out to 247.8 W per 1,000 cfm, so no more than about 297 W at full airflow. Run that for 1,000 hours across a season — substitute your own runtime — and it is 297 kWh, or $54.77 at 18.44 cents.

Read that as a worst permitted case rather than a measurement. The rule says a compliant fan "meets or is less than" the number, and a variable-speed ECM blower in a modulating furnace can sit well under it; the honest version is a range with $54.77 at the top. What the figure does establish is that the gas option carries an electric bill, and no AFUE percentage anywhere on the quote will show it to you.

Going the other way, a seasonal COP hides a curve. ENERGY STAR's cold climate designation requires COP at 5°F of at least 1.75 and heating capacity at 5°F of at least 70% of the 47°F capacity, measured per the Appendix M1 H4 test. A seasonal COP of 2.29 does not mean 2.29 on the coldest night. It is an average over a modelled season, and it includes mild hours when the machine did much better than 2.29 along with cold hours when it did much worse.

Precisely how that calculation treats the resistance backup that runs once capacity falls short of load is set out in AHRI 210/240-2024, which DOE incorporates by reference but which is not published free, so this article does not claim to know it. If the backup coil's share of your winter is going to decide the purchase, that is a question for the AHRI certificate and the installer's load calculation, not for one number on a label.

Quotes that still subtract $2,000

The Energy Efficient Home Improvement Credit under section 25C was worth up to $2,000 a year for a heat pump meeting the CEE highest efficiency tier. Qualifying property had to be placed in service on or after 1 January 2023, and the IRS page for the credit now states plainly that you can claim it "for improvements made through December 31, 2025" (IRS, Energy Efficient Home Improvement Credit, read 19 August 2026). Nothing installed in 2026 qualifies. The section 25D credit that covered solar and battery storage closed on the same date (IRS, Residential Clean Energy Credit, read 19 August 2026).

So a quote dated this year that still nets out $2,000 at the bottom is telling you when its template was last revised, which is worth knowing about the rest of the sheet too. State and utility programs are a separate question and some are worth real money. Each one needs its funding confirmed before it goes into the arithmetic, because programs of that kind close mid-year when the money runs out rather than on a published date.

Run it yourself before anyone visits

Two rates and two efficiencies, and only the rates are any trouble.

For G, take the coldest month you have and add up the volumetric charges on it — commodity plus delivery, per therm, converted with the therm factor printed on that same bill. Not the bill total divided by the therms. That total contains the customer charge, and if the water heater or the range is staying on gas you will go on paying that charge after the furnace is gone. Leaving it out is the honest comparison, and it moves the answer against the heat pump.

For E, you want what the next winter kWh costs, which is not your average. It can land either side of that average — a large fixed monthly charge pulls the average up above the marginal rate, an inclining block pushes the marginal rate up past the average — and that is the whole reason the average is no substitute.

AFUE is on the furnace nameplate. If the plate is unreadable, the gap between 80% and 95% moves the breakeven by 18.75%, which is worth a technician's two minutes. HSPF2 is on the AHRI certificate for the specific outdoor and indoor combination being quoted, not on the brochure, and the model numbers on the certificate should be the model numbers on the quote.

Then it is one number against one number: 29.31 x AFUE x E / G on one side, HSPF2 / 3.412 on the other, and work you can put in front of whoever wrote the quote. If they land close together, the equipment is not really the decision. Cutting the heating load changes both sides of the fraction at once, and the audit that ranks those measures runs on the same twelve months of bills you just pulled.