How do you stay cool in a summer power outage?
Scenarios Updated 2026-08-08
Move air, do not cool it. A box fan draws 80 W on its top speed, and the machines here that can print an hour figure run one for 2.5 h to 18 h.
A window air conditioner is a different category of question. The record here draws 833 W and asks 6,900 W at the compressor start, and 11 of 12 published US machines here are refused on it — and the one that is not still measures its run in hours rather than nights.
What each way of getting cooler costs
Four cooling loads, from the cheapest to the most expensive
The spread across this table is the whole answer. Two of these are inside what a portable battery does well; two are not, and the reason is the compressor start rather than the running watts.
The maker states a band across three speeds rather than one figure per speed, so this is the top of it.
The bottom of the same band. This is the setting that makes a night’s use realistic on a mid-size battery.
Small enough that the interesting question stops being the battery and starts being whether the air is worth moving.
The record’s low-speed figure times an eight-hour night — the same basis the record’s own test conditions state. The maker published the watts; the multiplication is ours.
The empty row is why every verdict on this load is a downgrade, and the reason has nothing to do with the wattage above it.
A shaded-pole induction motor does draw an inrush. We have found no published figure for it from any fan maker or standards body, and this library declines to invent a multiple — so the engine applies its missing-surge policy instead, which caps the verdict rather than assuming the load is harmless.
Ten times the fan, and still not the number that decides the verdict.
See the record →The compressor’s locked-rotor demand. Eleven of the twelve published US machines here are stopped on this figure, and the twelfth clears it on a start-up ceiling twice the size of any other in the library.
The hose-out-the-window kind. Higher than the window unit, because it is throwing away conditioned room air to cool its own condenser.
See the record →The variable-speed kind, which starts by ramping instead of by slamming a compressor onto the line. It is the only air conditioner here that most of the library carries.
See the record →Official spec + Derived from official coefficients + Derived by WattPairverified 2026-08-01appliance records · cooling
The same battery, four different summers
Live runs across the cooling loads
Two machines, four loads, every row computed on this build. The pattern to read is that the verdict and the hours are answering different questions, and on this scenario they disagree.
Downgraded rather than refused, and not on account of its watts: the load ratio is a few per cent of continuous output. What caps it is a missing inrush figure — the engine will not certify an inrush this library holds no measurement of, so it holds the verdict one grade below a clean pass.
Show the working →The smallest machine here, on the same load, with the same downgrade for the same missing figure — and a night’s worth of fan is still most of what this battery holds.
Show the working →Refused on the start. The running draw is under a third of continuous output, and this record states no start-up ceiling of its own, so the engine holds a compressor’s locked-rotor demand against the continuous rating — which it is several times above.
Show the working →One of the largest machines here, refused on the same clause. The hours beside it are worth reading anyway: even if the compressor started, this is an evening rather than a night, and it is measured against a duty cycle the maker publishes.
Show the working →The variable-speed unit, and the machine clears the run comfortably. It prints no hours: this load record carries no duty cycle, so there is nothing to divide a battery by — which is exactly the gap that keeps this page from being a ranked list.
Show the working →Set the two runtime columns side by side and the practical answer falls out. Across the machines whose records carry a usable-capacity figure, a fan runs for 2.5 h to 18 h, and a window air conditioner — on the machines that would refuse to start it anyway — for 0.3 h to 2.1 h. That is not a difference in degree. One of them is a way to get through a night and the other is a way to get through the worst two hours of an afternoon.
What to do with a battery and a hot house
In the order that buys the most comfort per watt-hour
Each of these is an electrical statement. None of them is a health judgement.
The low speed on this record is not much more than half the top speed, and it is the difference between a few hours and a whole night on the same battery.
Open the source →A fridge and a freezer are the loads that run for the whole outage rather than the evening, and they compete with the fan for the same watt-hours.
It is the only air conditioner in this library that most of these machines carry, because it ramps up instead of slamming a fixed-speed compressor onto the line.
Open the source →How hot is too hot, and for whom, is not an electrical question and this site publishes no figure for it. The US public health and weather services do, and they are where that half belongs.
One line of this page is not ours to write. Nothing here says what temperature is unsafe, for how long, or for which people — that is a health question, this library holds no measurement of it, and the federal heat-illness guidance is maintained by people who do. Wording of the routing sentence above checked against the CDC and National Weather Service heat pages on 2026-08-08.