Why Our Heat-Pump Model Needs 7,000 Hours to Beat Gas

$146,667 for the heat pump; $112,395 for the gas boiler. Both deliver 4,000 MWh of annual heat in our illustrative factory, equivalent to 4,000 full-load hours, yet the more energy-efficient equipment has the higher bill because electricity carries a separate charge for peak demand.

The peak alone costs $80,000 a year. Our calculation needs about 7,000 equivalent full-load hours to reach energy-bill parity, roughly 80% of a year’s full-output capacity.

Installation costs remain outside that comparison, so the threshold cannot tell a factory whether its proposed investment pays back.

An IEA Heat Pumping Technologies report from June 2026 describes a 12.5-megawatt steam heat pump at Delfort’s Finnish paper mill, but technical capability leaves the bill unresolved: a 2025 Minnesota study estimated 32% technical potential for replacing Minnesota’s industrial gas consumption while assigning only 4.7% economic potential. That smaller figure uses a judgment-based factor, not a statewide count of profitable projects.

Our model asks a narrower question, without reproducing that study: how much added peak demand can cheaper heat afford?

Cheap heat, expensive peaks

A heat pump moves energy from a cooler source to a hotter process. Its heating coefficient of performance, or COP, measures delivered heat divided by electricity consumed.

We give our hypothetical factory one megawatt of useful heat output, a heat pump with heating COP 3 including auxiliaries, free available waste heat and an 85%-efficient alternative boiler, then price incremental electricity at $0.05/kWh including every per-kWh rider but excluding demand fees, against $7 per million Btu of avoidable delivered gas. These are teaching inputs, not a tariff, vendor quote or Minnesota averages.

Keep electricity-volume and demand charges separate: dividing a total bill by kilowatt-hours and then adding demand fees would count them twice.

At COP 3, one electrical kilowatt-hour delivers three kilowatt-hours of heat, drawing the extra energy from the source. Useful heat therefore costs $50 ÷ 3 = $16.67 per megawatt-hour.

Gas costs $28.10 per useful MWh, using EIA’s 3,412 Btu per kWh conversion: $7 ÷ 293.083 ÷ 0.85 × 1,000.

The pump wins on energy volume.

Then the peak arrives on the bill. Xcel’s Minnesota bill guide describes charging commercial or industrial customers for their highest 15-minute average power; it supports the mechanism, not our assumed rate of $20/kW-month.

Delivering one megawatt of heat at COP 3 draws 333.33 electrical kilowatts, so adding that full draw to the factory’s billed peak in each of twelve months costs 333.33 × $20 × 12 ≈ $80,000 annually, however little heat the pump produces between those peaks.

That fixed charge makes annual output decisive: spreading the same $80,000 over more useful heat lowers the cost of each megawatt-hour, which is why we measure utilization in equivalent full-load hours, dividing annual useful heat by rated output rather than counting every hour in which the machine is merely switched on. Two running hours at half output count as one.

At 4,000 hours, $66,667 for electricity volume plus $80,000 for demand makes the opening $146,667 bill. At 8,000 hours, the modeled bill per useful MWh finally falls about 5% below gas.

How much peak can the savings buy?

At 4,000 hours, the savings can buy about 190.5 kW of added monthly peak. To find that allowance, subtract electricity-volume costs from avoided gas costs using unrounded values, leaving $45,728.63 for annual demand charges, then divide by $20 × 12 for a constant monthly addition.

Equality occurs at approximately 190.536 kW; rounding up to 191 would erase the savings.

A 333-kW machine need not add a 333-kW peak. If it stays off during the factory’s old 1,000-kW peak, runs when other loads peak at about 833 kW, and creates no higher total in any other interval, the new factory maximum is about 1,167 kW: only half the pump’s draw has become added demand, because the other half fits beneath a peak the factory already pays for.

The schedule still has to deliver the heat. Pump-on intervals need enough coincident heat demand and available source heat. Continuous full output with unchanged other loads instead adds full demand at the old peak.

For other schedules, let H denote equivalent full-load hours and f denote added billed demand divided by rated electrical draw, held constant across twelve bills.

Heat-pump cost per useful kWh = ($0.05 + 12 × $20 × f ÷ H) ÷ 3.
Energy-bill break-even H = 12 × $20 × f ÷ (3 × $0.028099 − $0.05).

Full added peak requires 6,998 hours; half added peak requires 3,499. At full peak, demand rates of $10 or $30/kW-month instead require roughly 3,500 or over 10,400 hours.

Changing efficiency moves both parts of the bill: with other inputs fixed, COP 4 cuts electricity consumption and full-load draw enough to lower break-even to 3,846 hours, whereas COP 2.5 raises it to 11,854; GEA’s supplier estimates show actual performance varying with source and delivery temperatures. These are scenarios, not selectable machine settings.

Raising electricity’s volume rate from five to six cents pushes break-even to 9,878 hours, beyond our modeled year’s 8,760. If that rate equals or exceeds COP times gas-heat cost, no utilization creates savings.

The strongest case for the pump

For a factory that needs heating and cooling at the same time, the pump can do two purchased jobs with one electrical input, recovering heat that displaces boiler output while avoiding electricity a chiller would otherwise consume; a heating-only comparison like ours misses that second saving, which belongs in the decision whenever the cooling is genuinely needed rather than merely available. GEA’s supplier factsheet illustrates heating COP 3 alongside cooling COP 2. Those are five units of combined service, not five of heat. These are supplier estimates, not independently validated measurements.

Credit only cooling otherwise purchased, counting avoided chiller energy and demand once; a demand credit requires an actual decrease in the facility’s billed peak. With no added billed demand, our pump’s modeled energy cost is 40.7% below gas at any positive output.

Limitations

We have not tested whether a real factory can supply the heat at the assumed operating hours and peaks. Constant COP, prices and monthly peaks simplify the calculation. For varying conditions, annual COP is total delivered heat divided by total electricity, not an arithmetic average of interval COPs; demand charges still require performance during the whole-facility billing peak. Both options supply the same heat block, so unchanged remaining boiler heat cancels while changed backup costs must be added.

We exclude capital, service upgrades, financing, maintenance, retained gas charges, storage losses and avoided boiler auxiliary electricity. That last omission favors gas.

The Bottom Line

Before evaluating a quote, collect a year of electricity readings at the tariff’s billing interval, match them to heat demand and source temperatures, and request a before-and-after facility bill that separates electricity volume, added peaks and avoided cooling while accounting for seasonal rates and minimum charges tied to earlier peaks. Then commission a separate investment appraisal. A brochure’s COP cannot price your next demand peak.