Water from air machines: what a gallon actually costs to make

There is only so much water up there

Air holds water in proportion to how warm it is. At 30 °C a cubic metre of fully saturated air holds 30.4 grams of water, and at 80 per cent humidity it holds 24.3 grams. At 15 °C and 40 per cent humidity, the kind of afternoon most of the country has for much of the year, the same cubic metre holds 5.1 grams. That is the whole story in two sentences: the machine cannot pull out water that is not in the room.

Saturation figures from the NIST steam property tables, NISTIR 5078, read on 16 September 2026. We checked them against the Buck equation and they agree within 0.3 per cent.

To make 20 litres in a day, the amount a large consumer unit claims, a machine with perfect extraction would have to process about 820 cubic metres of air at 30 °C and 80 per cent humidity. Nothing extracts perfectly, so the real airflow is higher. This is why these machines are the size of a small fridge and why they are loud.

Our water from air calculator does this for your city with NOAA hourly normals, in July and in January, so you can see the two seasons instead of one number.

The rating on the box is a weather report, not a promise

Manufacturers rate these units at conditions that flatter them. Aquaria states it plainly on its own product page, calling 30 °C and 80 per cent humidity "the standard conditions used to compare atmospheric water generators". Watergen rates the GENNY differently, at 26.7 °C and 60 per cent humidity, which is closer to a real living room and worth noticing.

So when a unit says 20 litres a day, read it as 20 litres a day in a warm, humid room. Phoenix in June is hot and dry. Minneapolis in January is neither. The same machine in those two places is two different machines.

Product pages for the Aquaria Hydropack and the Watergen GENNY Home, both read on 16 September 2026.

The number nobody prints: what a gallon costs in electricity

Manufacturers publish watt-hours per litre. Watergen says 400 Wh per litre for the GENNY in its standard conditions. Aquaria says about 240 Wh per litre for its whole-home unit. Those are the brochure figures.

There is one long independent measurement we could find. A commercial unit was run for a full year in Sharjah, a hot and humid climate that should favour it, and the published annual average was 2.25 kilowatt-hours per litre, five to nine times the brochure figures. It was published on 1 November 2022 in Case Studies in Chemical and Environmental Engineering.

Turn that into money. A gallon is 3.785 litres, and the average American household paid 18.34 cents per kilowatt-hour in June 2026, the most recent month the EIA has published.

Our arithmetic, from the sources listed at the bottom. Electricity at the U.S. residential average of 18.34 cents per kWh, EIA Electric Power Monthly Table 5.3, data for June 2026, published 26 August 2026.

Where you live moves that a lot. At the cheapest state price in that same month, Nevada at 13.11 cents, the measured figure works out to $1.12 a gallon. In Hawaii, at 52.72 cents, it is $4.49 a gallon. Forty-five kilowatt-hours a day, by the way, is more electricity than the average American home uses in a day for everything else combined.

None of this makes the technology fake. It makes it expensive, and it tells you exactly which problem it solves: not cheap water, but water where there is no other source.

What the EPA found when it tested one

The EPA ran a commercial unit continuously for three months and published the results in a technical brief in November 2019. The findings are more reassuring than the internet suggests, and more demanding than the brochures admit.

• No element measured above the EPA primary or secondary drinking water standards.
• No Legionella and no mycobacteria detected.
• No fecal coliforms and no E. coli detected.
• Heterotrophic plate counts were high, in the raw water and after treatment. That is a general measure of bacteria happy to grow in standing water.

The brief's own recommendation is the part to remember: bacterial growth "could be addressed by incorporating chlorination or ozonation as an additional disinfection process", and by keeping a disinfectant residual in the tank. Water from air starts almost pure, which sounds good and is actually the problem: it has nothing in it to discourage anything from growing, and it is aggressive toward metal plumbing.

There is also a standard, ASSE 1090-2020, written for drinking water atmospheric water generators, which requires filtration and disinfection controls as part of the device. If a unit does not mention a standard at all, that absence is the answer.

EPA/600/F-19/129, Atmospheric Water Generation Technology, published November 2019. ASSE 1090-2020, American National Standard.

What they cost to buy

Prices we could verify on manufacturer or dealer pages on 16 September 2026, for units sold to households in the United States.

Two of those pages do not state a daily output at all, which tells you something about how the category is sold. We left out units whose seller pages we could not open on the day we checked, rather than repeat a price we did not see.

So who should buy one

If you have a tap and a roof, a rain barrel and a few stored containers give you far more water for far less money, and they do not stop working when the power does. Our rain barrel calculator gives you the gallons your own roof catches per inch of rain, and our water storage guide covers the containers and the rotation.

An atmospheric water generator earns its price in a narrower case: warm humid air, no trustworthy source, and grid power you can count on. That last condition is the one people skip. A machine that needs 45 kilowatt-hours to make five gallons is not an outage plan. It is a plumbing alternative.

Sources, with the date of each

• NIST, NISTIR 5078, saturation property tables. Read 16 September 2026. Cross-checked against the Buck equation.
• EPA, Atmospheric Water Generation Technology, EPA/600/F-19/129, November 2019.
• Ahmad and others, twelve months of operation of a commercial unit in Sharjah, Case Studies in Chemical and Environmental Engineering, published 1 November 2022, doi 10.1016/j.cscee.2022.100270.
• U.S. Energy Information Administration, Electric Power Monthly, Tables 5.3 and 5.6.A. Data for June 2026, published 26 August 2026.
• ASSE 1090-2020, Performance Requirements for Drinking Water Atmospheric Water Generators.
• Manufacturer and dealer pages for Watergen, Aquaria, Kara Water and Atmospheric Water Solutions, read 16 September 2026.

Things we could not confirm, and therefore did not state: whether the NSF P343 protocol is current and which products hold it, the monthly breakdown of the Sharjah measurement, and the often repeated claim that real output is 40 to 70 per cent of the rating, which we could only find on commercial blogs.