Home › Generators › Fuel & runtime guide

Generator fuel consumption & runtime

A generator is only as good as its fuel supply. This guide turns manufacturer consumption figures into hours of runtime — by tank size, by load level — so a multi-day outage is a plan, not a surprise.

How fuel consumption is rated

Manufacturers rate standby generators at two load points: 50% load and 100% load, meaning the engine is producing half or all of its rated electrical output. Both figures appear on the spec sheet for every standby model we track, in gallons per hour for propane (LP) and cubic feet per hour for natural gas (NG).

The 100% number matters for worst-case minutes — everything in the house running at once. The 50% number is the one that governs your outage, because real loads cycle: the furnace blower runs in bursts, the refrigerator's compressor is off half the time, and the well pump fills its pressure tank and rests. Over a full outage day, most homes average closer to half of their calculated peak than to the peak itself, which is why this guide plans around the 50% column and treats 100% as the ceiling.

One surprise in the data: halving the load does not halve the burn. Across the models in our dataset, fuel use at 50% load runs about 55–80% of the full-load figure — an engine has fixed overhead (turning itself, cooling, controls) no matter how little electricity it makes. That is also why a much larger generator than you need is quietly expensive to feed: its half-load burn can exceed a smaller unit's near-full-load burn.

Note that the same model burns different fuels at different rates and makes different power on each. Most standby units produce less electricity on natural gas than on propane — the table below shows both ratings — while consuming gas by the cubic foot rather than the gallon. Comparing models on the fuel you will actually use is the only honest comparison.

Standby fuel burn at 50% load, model by model

Compiled from manufacturer specification sheets — the same dataset behind the model database, which adds 100%-load figures, noise ratings, and prices. Kohler and Cummins publish propane use in cubic feet per hour; the gallon equivalents (at 36.4 cu ft per gallon) are shown in parentheses in our dataset and used here.

Two patterns worth noticing. Burn does not scale neatly with size class — the Guardian 26 kW sips less at half load (188 cu ft/hr NG) than the Guardian 22 kW (228), because the engines and tuning differ. And the spread between brands at the same size is real money over a long outage: at 14 kW, half-load propane burn ranges from 1.6 to 1.81 gal/hr, roughly a 13% difference in how long a tank lasts.

Propane runtime: tank size is runtime

Propane tanks are never filled to the nameplate number. To leave room for the liquid to expand with temperature, fills stop at about 80% of nominal capacity — so a 120-gallon tank holds about 96 usable gallons, a 250 holds 200, and a 500 holds 400. Runtime is then one division:

Runtime (hours) = usable gallons ÷ burn rate (gal/hr)

Worked once, in full: a Generac Guardian 14 kW burning 1.81 gal/hr at half load, on a 250-gallon tank — 250 × 0.80 = 200 usable gallons; 200 ÷ 1.81 ≈ 110 hours, and 110 ÷ 24 ≈ 4.6 days of continuous half-load running. Swap in your own model's gal/hr from the table above and the same two steps give your number.

These are continuous-running figures at a steady half load — the conservative planning case. Homes that shed discretionary loads (next section) routinely do better. Note the inverse relationship buyers underestimate: stepping from a 14 kW to a 22 kW unit on the same 250-gallon tank cuts runtime from about 4.6 days to about 3.3, before the bigger unit has powered a single extra appliance. Size to the load with the sizing calculator first; then read runtime off your actual size class.

Natural gas: the tank you never fill

On natural gas there is no runtime table, because there is no tank: while utility gas keeps flowing, the generator runs indefinitely. Two practical caveats replace the runtime math. First, your meter, regulator, and line size cap how much gas the generator can draw — a large standby at full load wants 250–330 cu ft/hr, and an undersized service can starve it exactly when the load peaks; installers check this against the spec-sheet figures in the table above. Second, output drops on gas: most models in our dataset make 1–3.5 kW less on natural gas than on propane, which is why the model database rates every model on both fuels.

For cost purposes, utilities bill natural gas in therms (or hundred-cubic-foot units that convert to them). One therm is 100,000 Btu, and typical pipeline gas carries roughly 1,000 Btu per cubic foot — so 1 therm ≈ 100 cu ft is the working conversion. A model burning 195 cu ft/hr is drawing about 1.95 therms per hour, or about 47 therms across a full day of half-load running. Your bill's per-therm rate turns that into dollars, which the next section does with stated price assumptions.

What a day of backup power costs

Fuel is the operating cost nobody quotes at purchase. The table below runs three representative models at 50% load for 24 continuous hours, on both fuels, at these stated price assumptions (see Sources & notes): propane at $2.60 per gallon and natural gas at $1.50 per therm. Your local prices will differ — propane especially — so treat the gallon and therm columns as the durable numbers and re-price them with your own rates.

Three takeaways. Natural gas is the cheaper fuel to run — roughly half the daily cost of propane at these assumptions — on top of never needing a delivery. Propane's cost is the price of independence from the gas grid, and it argues for the smallest generator that honestly covers your loads. And a continuous half-load day is the expensive scenario: a real outage day with cycling loads and a few shed circuits lands below these figures, which is what the planning habits below are for.

Portables and inverters: hours, not days

Portable and inverter generators carry their fuel with them, so the same division applies with smaller numbers: onboard tank gallons ÷ burn at your load. Manufacturers of portables usually publish runtime at 50% load rather than a burn rate; the gallons-per-hour figures in our dataset for these models are arithmetic from those published tank-and-runtime pairs, and are labeled "derived" there. A few from the model database:

In practice, portable runtime is a refueling schedule: a standard 5-gallon can adds roughly 5 hours to a large dual-fuel portable like the XP13000EH (5 ÷ 1.04 ≈ 4.8 hr), about 8 hours to the XP9000iH inverter, and over 13 hours to a small inverter like the iGen4500DFc. Dual-fuel models can also draw from a propane tank instead of the onboard gasoline tank, converting hours into days the same way standby runtime works. If you are matching a portable to a load list, start with the sizing calculator totals, then divide your stored fuel by the burn rate — and remember these runtimes assume a steady half load, so heavy continuous loads shorten them.

Stretching the fuel you have

• Shed the discretionary loads. Electric water heaters, dryers, ovens, and space heaters are fuel budgets disguised as appliances. The sizing calculator wattage table shows what each one costs to keep; dropping the big resistive loads during an outage can move a home from the 50% column's appetite toward something noticeably smaller.
• Load management beats tank size. A load-management module that staggers motor starts and pauses the AC while the well pump runs flattens peaks — and since burn follows load, a flatter day burns less. It can also keep a smaller, thriftier generator viable in the first place.
• Exercise cycles are cheap insurance. Standby units self-test weekly for a few minutes at low speed. That sip of fuel is negligible against an outage budget; don't disable the exercise cycle to save it — a generator that fails to start has an effective runtime of zero.
• Cold weather shrinks propane supply. Propane tanks vaporize liquid into gas more slowly as temperatures fall, and a small, nearly empty tank in deep cold may not keep up with a large generator's draw even with gallons remaining. Bigger tanks and keeping the fill level up ahead of winter are the standard mitigations; treat nameplate runtime as a warm-weather ceiling.
• Long runs need a service pause. During multi-day continuous running, manufacturers call for periodic oil checks and eventual oil changes at intervals stated in the owner's manual. Budget a brief shutdown for that rather than discovering it — it is the one planned pause in an otherwise continuous outage plan.

Sources & notes

• Consumption figures are manufacturer-published ratings compiled into our generator dataset (the same data as the model database), cross-checked against retailer listings; portable gallons-per-hour values are arithmetic derivations from published tank size and 50%-load runtime and are labeled "derived" in the dataset. Always confirm against your specific model's current spec sheet before planning around a number.
• Propane price assumption — $2.60/gal. U.S. Energy Information Administration (EIA) residential propane averaged $2.598/gal in the week of October 5, 2026; state averages in the same survey ran from about $1.63 to $4.86, so re-price the tables with your delivered rate. Source: EIA, Heating Oil and Propane Update, eia.gov/petroleum/heatingoilpropane/ (accessed October 2026).
• Natural gas price assumption — $1.50/therm. EIA's Short-Term Energy Outlook puts recent U.S. residential natural gas around $13–$18 per thousand cubic feet (Mcf); at roughly 10 therms per Mcf that is about $1.30–$1.80 per therm before state-by-state variation. We use $1.50 as the midpoint assumption. Source: EIA, Short-Term Energy Outlook, eia.gov/outlooks/steo (accessed October 2026).
• Conversions used: propane tanks filled to ~80% of nominal capacity; 1 therm = 100,000 Btu with typical pipeline gas at ≈1,000 Btu per cu ft, so 1 therm ≈ 100 cu ft; propane 36.4 cu ft of vapor per liquid gallon (used for the Kohler/Cummins gallon equivalents published by those manufacturers). All runtime and cost figures in this guide are arithmetic from those inputs and the dataset above — no consumption figure was estimated where a manufacturer value was unavailable.