The Economics
of Waste Heat

The recovered-heat line is the most commonly inflated number in on-site generation economics: rated thermal output, priced as if every unit of it finds a use, in every hour, forever. Recovered heat is real money only when the machine produces it at a temperature an actual load can accept, in the hours that load exists, displacing fuel that would otherwise have been burned.

Section 01What a unit of recovered heat is worth

Start with the only valuation rule that survives scrutiny: a unit of recovered heat is worth what the site would otherwise have paid to produce that same unit, delivered to the same load, in the same hour. For most industrial and institutional sites, the thing otherwise producing the heat is a natural gas boiler. So the ceiling value of a delivered million Btu of recovered heat is the price of gas divided by the boiler's efficiency, because the boiler burns more than one unit of fuel to deliver one unit of heat.

The inputs are checkable. The U.S. Energy Information Administration publishes California's industrial natural gas price monthly; over the twelve months of published data through May 2026 the monthly averages ran between roughly $9.50 and $15.35 per thousand cubic feet, with May 2026 itself at $9.49.⁵ A thousand cubic feet of pipeline gas contains close to 1.04 million Btu, so that span translates to approximately the same range per million Btu. Assume, illustratively, gas at $10 per million Btu and a boiler running at 80 percent efficiency, a defensible mid-case for serviceable existing equipment. Each delivered million Btu of recovered heat then displaces 1.25 million Btu of purchased fuel, worth $12.50. Every number in that sentence is either sourced or labeled as an assumption, which is how the heat line in any study should read.

Now note what recovered heat is not worth. It is not worth the retail electric rate, unless it genuinely displaces electrically produced heat, which is the rarer case and should be priced only when the electric heater actually exists on the site today. And it is not worth the headline efficiency of combined heat and power transmuted directly into dollars. The U.S. Environmental Protection Agency reports that combined heat and power can exceed 80 percent total fuel efficiency, against roughly 50 percent for conventional separate generation and on-site boilers.¹ That is a true and important statement about fuel. It is not a price. Converting it into dollars requires the displaced-fuel arithmetic above, applied to the heat actually captured, and seller materials routinely skip that step.

Section 02Grade before quantity: where each machine's heat lives

Two machines burning identical fuel can offer entirely different thermal products. Electrical efficiency fixes how much heat remains after generation; the machine's architecture fixes the temperature at which that heat leaves; and temperature fixes what the heat can serve. A load that needs 350°F cannot be fed with 200°F water, no matter how many million Btu of it the brochure counts. Grade comes before quantity, and the honest inventory by technology class looks like this.

Reciprocating engines reject heat in two separate streams. Roughly half or more of the recoverable total arrives as low-grade heat from the jacket water and lube oil circuits at around 200°F, and the balance arrives in exhaust gas that EPA's example systems show at about 720°F to 1,000°F.² The exhaust stream can raise low-pressure steam, in EPA's characterization up to about 400 psig from the exhaust alone, while the jacket stream is hot water and nothing else.² The case for engines is that the recoverable total per unit of power is large and the packages are mature. The honest limit is that a site whose thermal demand is higher-pressure steam can use only the exhaust fraction, and a site with no use for warm water strands half the thermal product.

Combustion turbines put essentially all of their rejected heat into a single exhaust stream, which EPA characterizes at up to 800°F to 900°F for smaller industrial machines and up to about 1,100°F for aeroderivative designs, capable of raising process steam at conditions as high as 1,200 psig and 900°F through a recovery boiler.³ That single high-grade stream is the strongest thermal product in the field, and it is why turbine-class machines have historically anchored large steam hosts. The honest limits run the other way: simple-cycle electrical efficiency at modest scale is lower than the best engine-class machines, efficiency falls at part load,³ and paying for steam-grade recovery hardware at a site with only warm-water loads is buying quality the load cannot use.

Microturbines produce one moderate-temperature exhaust stream suited to hot water or modest low-pressure steam duty. The case for them is compactness and a single simple recovery interface; the case against is modest electrical efficiency and a thermal grade that will not serve demanding process loads.

Fuel cells convert more of the fuel to electricity than combustion classes, which is precisely why they offer less heat to recover per unit of power. Low-temperature classes yield mostly warm water; high-temperature classes exhaust hot enough for steam-capable recovery. The case for them is that a machine already efficient on the electric side needs less heat credit to justify itself, along with non-combustion air permitting treatment in strict districts. The honest limit is the mirror of the virtue: a business case that needs a large heat credit to close cannot get it from the class designed to minimize rejected heat, and capital cost remains the other side of that ledger.

None of these profiles is a defect. They are design choices, and each class wins somewhere. The discipline is matching the temperature of what the machine rejects to the temperature of what the site actually consumes, before a single dollar of heat credit enters the model.

Section 03Rated output is not captured output

The specification sheet states what the machine can reject as recoverable heat at full load. The site determines what fraction of that ever becomes displaced fuel. Four discounts stand between the two numbers, and every one of them is knowable before purchase.

• Coincidence. Heat is perishable. It counts only in the hours when the thermal load and the machine's operation overlap. A generator running around the clock against a thermal load that lives in two shifts captures at most what those shifts can absorb.
• Season. Space heating loads collapse in summer; some process loads collapse in winter shutdowns. An annual average built from a January site visit is not an annual average.
• Distance. Heat travels badly. Piping runs between the machine and the load lose energy continuously and cost real capital and pumping power. A thermal host two buildings away is not the same host as one across the wall.
• Turndown. The machine follows one master, electric or thermal. Whichever side it does not follow spills. A unit dispatched to electric demand sheds recoverable heat whenever the thermal load is not there to take it, and the reverse discipline sacrifices electric value instead.

The product of those four discounts is the capture fraction, and it is where honest and promotional economics part company. An illustration, with every input labeled: take a machine rated to reject 4.0 million Btu per hour of recoverable heat, running near continuously. Price delivered heat at the $12.50 per million Btu derived in Section 01. The brochure arithmetic values the thermal side at 4.0 times 8,760 hours times $12.50, about $438,000 per year. Now suppose the measured overlay of machine operation against the site's actual thermal loads supports an annual average capture of 45 percent, a common outcome once coincidence, season, and turndown are counted. The honest figure is about $197,000 per year. Same machine, same site, same gas price; 45 cents on the brochure dollar, and the difference is not pessimism. It is measurement.

Rated thermal output is a property of the machine. Captured thermal output is a property of the site. Only the second one pays.

The capture fraction is discoverable in advance. Twelve months of fuel bills establish the thermal baseline; boiler logs or metered steam and hot water flows establish its shape; an hourly overlay of expected machine operation against that shape yields the fraction. A proposal that presents a heat credit without showing this overlay is asserting the single most decision-relevant number in the file rather than deriving it.

Section 04Absorption chilling: buying cold with heat

Sites long on heat and short on thermal load in summer have one more sink available: absorption chillers, which produce chilled water from heat input rather than compressor work. The U.S. Department of Energy's technology fact sheet characterizes the two commercial families. Single-effect machines are driven by hot water at roughly 200°F to 240°F or low-pressure steam near 15 psig, with representative full-load coefficients of performance of 0.70 to 0.79, meaning each million Btu of driving heat yields 0.70 to 0.79 million Btu of cooling. Double-effect machines reach 1.35 to 1.42 at full load but demand hot water near 350°F or steam near 115 psig.⁴

Read those driving temperatures against Section 02 and the quality-matching logic repeats. Engine jacket water can drive a single-effect machine and nothing more. Double-effect performance is available only to sites with turbine-class exhaust or genuine high-pressure steam. A model that pairs a low-grade heat source with a double-effect coefficient of performance has made a physics error, and it is a common one.

The honest burdens belong in the model too. The published coefficients exclude the electricity consumed by solution pumps, tower fans, and other ancillaries.⁴ Because the cycle rejects both the driving heat and the cooling load, an absorption machine requires a larger cooling tower than an electric chiller of the same duty,⁴ with the water use and fan power that follow. Maintaining cycle vacuum and purging non-condensables is real ongoing work, and the machines are physically large.

What is the heat worth when cooling is the sink? Illustratively: one million Btu of driving heat into a single-effect machine at a coefficient of performance of 0.74 yields 740,000 Btu of cooling, about 62 ton-hours. If that cooling would otherwise come from an efficient water-cooled electric chiller drawing an assumed 0.6 kilowatts per ton, the displaced electricity is about 37 kilowatt-hours. At an illustrative blended avoided rate of 20 cents per kilowatt-hour, the driving heat earned about $7.40 per million Btu, before the parasitic loads above. At 30 cents, about $11.10. Two conclusions follow, both honest. First, at moderate electric rates, heat-to-cooling conversion values a unit of heat below direct boiler displacement, so it is the second-best sink where a fuel-displacing load exists. Second, at California commercial rates, and at sites where summer heat would otherwise spill entirely, absorption chilling can raise the annual capture fraction materially, which is often worth more than the per-unit arithmetic suggests. Both effects belong in the hourly overlay, not in a footnote.

Section 05When heat flips the decision, and when it rounds to zero

The heat line matters most when it is large enough to reorder the technology ranking. Here is the mechanism, with every input labeled illustrative. Consider two machine classes at the same site. Machine A converts 44 percent of fuel energy to electricity, higher heating value basis, and offers little usable heat. Machine B converts 33 percent but rejects a usable thermal yield near 3,500 Btu per kilowatt-hour, a figure inside the range EPA tabulates for engine-class example systems.² A kilowatt-hour requires 3,412 Btu of electricity-equivalent energy, so Machine A burns about 7,800 Btu of fuel per kilowatt-hour and Machine B about 10,300. At $10 per million Btu, that is roughly 7.8 cents versus 10.3 cents of fuel per kilowatt-hour: a 2.6-cent penalty for Machine B before heat enters.

Now credit the heat at the $12.50 per million Btu delivered value from Section 01. At full capture, Machine B's 3,500 Btu per kilowatt-hour of usable heat is worth about 4.4 cents, more than erasing the penalty; the less electrically efficient machine wins on total energy cost. At 50 percent capture the credit is about 2.2 cents and the two machines are close to even. At 20 percent capture the credit is under a cent and Machine B loses decisively. In this illustration the decision flips near 60 percent capture. That is the entire economics of waste heat in one sentence: the technology ranking at a site can hinge on a capture fraction that most proposals never measure.

This comparison deliberately ignores capital cost, maintenance, air permitting, footprint, and noise, all of which differ by class and frequently decide the answer on their own. The point is narrower: the heat credit is the swing line, and its size is set by the site, not the machine. From it follow the honest boundary cases. Waste heat tends to flip decisions where thermal demand is large, steady, temperature-matched, and close, the classic continuous steam and hot water hosts, which is exactly the population where combined heat and power earned its record.¹ It tends toward a rounding error where the thermal load is seasonal or distant, where grade is mismatched, or where a high-electrical-efficiency machine already wins without it. And where there is no thermal load at all, the honest heat value is zero, and the right machine is simply the one that makes the most electricity per unit of fuel within permit, footprint, and capital constraints.

One second-order effect deserves a sentence. Because the heat credit is indexed to the same gas price as the generation fuel, a gas price increase raises both the cost of running the machine and the value of the heat it recovers. High-capture configurations therefore carry a partial internal hedge against fuel prices that heat-blind arithmetic misses entirely. Partial is the operative word; the hedge dampens exposure and does not remove it.

Section 06The discipline that keeps the heat line honest

Everything above compresses into a short procedure, and any owner can hold any analysis, including ours, to it.

• Measure the thermal baseline. Twelve months of fuel bills at minimum; metered steam, hot water, and chilled water where they exist. The baseline is what the heat credit displaces; without it the credit is fiction.
• Audit temperatures, not just quantities. For each thermal load, record what temperature the process actually requires, not what the header happens to run at. Loads are frequently served hotter than they need, and the audit sometimes reveals recoverable-grade demand that was invisible.
• Match grade before counting quantity. Map each candidate machine's thermal product, stream by stream, against the loads it can physically serve. Heat that cannot reach a matched load at temperature is not output; it is rejection.
• Build the hourly overlay and state the capture fraction. Machine operating profile against thermal load shape, hour by hour, season by season, with distance losses and turndown behavior included. The capture fraction goes in the report as a named number with its derivation shown.
• Price captured heat at displaced fuel. Sourced gas price, stated boiler efficiency, dated inputs. Where the sink is absorption cooling, price it against the chiller it displaces, parasitics included. Label every estimate as an estimate.
• Run the decision with and without the credit. If the technology ranking changes, the study should say so in plain language, because that sentence tells the owner exactly how much weight the thermal assumptions are carrying.

On incentives, one note within a study's competence and one outside it. Under current federal law as of this writing, the investment tax credit for qualifying energy property is 30 percent, with statutory adders that exist but must be individually qualified rather than assumed.⁶ Whether specific heat-recovery equipment at a specific site qualifies is a determination for qualified tax counsel, and a study should say that rather than book it.

The heat line is where seller optimism hides, because it is the hardest line for a buyer to check from a proposal alone. It compounds a physics question, a temperature-matching question, and an hourly-coincidence question into a single dollar figure, and every party compensated by the equipment sale is paid more when that figure is generous. The correction is not skepticism about waste heat, which is real, measurable, and at the right sites decisive. The correction is measurement standing where marketing wants to stand: a capture fraction derived from the site's own data, a value anchored to the fuel actually displaced, and a decision shown both with and without the credit. If a project only closes on a generous heat line, the owner deserves to read that finding in writing before capital moves.