Water Out of Thin Air, Skywalker-Style (Part 1)

The “moisture vaporators” that Luke Skywalker’s Aunt and Uncle used to harvest water out of dry, desert air are becoming a reality. A slew of hopefuls, including a Nobel Prize winner, can plausibly scale to meet the modest water demands of today’s dry-cooled data centers.

By: Jason Bak

On a summer night in Death Valley, in a national park that averages barely two inches of rain per year, a white box the size of a carry-on suitcase quietly got heavier. Inside was a cartridge of aluminum-based crystal powder called MOF-303, a metal-organic framework engineered so that its internal pore walls grab water molecules out of bone-dry air. When dawn came, the sun heated a sealed chamber to 149°F, causing the trapped vapor to “sweat” back out, then hit a chilled condenser and drip into a beaker. The nighttime relative humidity averaged 14%, the kind of number that makes conventional dehumidifiers useless. The device still produced water: 114 grams per kilogram of sorbent that cycle. In the peer-reviewed results, its harvester delivered up to 210 grams per kilogram of MOF-303 per day at Furnace Creek in Death Valley, and 285 grams per kilogram in cooler, moister Berkeley. It ran on nothing but sunlight.

That 2023 demonstration, published in Nature Water by a UC Berkeley team led by chemist Omar Yaghi, is the origin myth of a technology suddenly flush with money and hype: Atmospheric Water Generation (AWG), machines that make drinking water from air. For most of its history it was a curiosity for preppers, disaster-relief NGOs, and off-grid villas. Then, on October 8, 2025, Yaghi was awarded a share of the Nobel Prize in Chemistry for inventing the material class that makes the Death Valley accomplishment possible.

Why It Works Now: The Machines Aren’t As Thirsty

In my last article, I discussed how today’s GPUs are designed to run hotter than desert air, thereby enabling them to run with less cooling. Chilling needs can therefore be fulfilled by waterless, radiator-style systems. A 100 MW data center cooled this way needs only two truckloads of water for cooling each year, leaving employee breakrooms and restrooms as the only significant water requirement. Those only need about three truckloads of water every week, no different from any other industrial facility.

For conventional evaporative cooling systems (the ones that give data centers their water-guzzling rap), AWG is unthinkable. However, with dry-cooled data centers needing only a few truckloads of water per week, certain branches of AWG suddenly become within reach.

For more information on how waterless cooling works with today’s AI processors, especially when connected to on-site solar, see my recent article at https://www.linkedin.com/feed/update/urn:li:activity:7487148476988878848.

Southwest High Desert: Why Bother Anymore?

Data center companies have been flocking to dry, inhospitable desert climates because the dry air, counterintuitively, makes conventional evaporative cooling systems work better. So if evaporation is no longer required to cool today’s processors, why does a desert environment make sense?

Answer: It doesn’t. Unless the reason for being in the desert is to use on-site solar power.

The desert’s abundant land and low disaster risk still help, but you can find those outside a desert too. The desert-specific advantage left standing is solar.

Bright, solar-friendly expanses are ideal for solar power. Inexpensive, quick-to-install (with certain methods like east-west racking, clean-running at the site, quiet. Excess power supply when the dry coolers need the most energy.

A big win for sustainability, the community and shareholders alike.

In our build economics, that equates to 30–40% more power per dollar invested than the alternatives, with revenue starting in well under half the time. It’s also not subject to the 4-6 year wait for grid access or grid price increases, which are all but certain in the coming years.

See another of my recent articles on rapid solar-deployment methods at https://www.linkedin.com/posts/greendatacenters_cleanenergy-solar-datacenters-activity-7480673735885156354-x00T?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAAIgCcBKEEMpG_cZwjM92JSf0wTBbqCD9Y).

There’s a second kind of line-skipping, too. While on-site power generation sidesteps the 4-6 year grid-interconnection queues, waterless cooling simultaneously sidesteps most of the water-rights wait. In addition to the uncertainty of water rights ever becoming available.

Water rights for new data center facilities are difficult to come by, which is why trucking in water is a viable, albeit less attractive, solution. Even though waterless cooling (like ours) only needs a few truckloads per week for 100 MW, those truckloads need to scale significantly as power generation expands. A 1 GW facility would be more like 20 truckloads per week.

The cost of trucking, while higher than utility, aquifer or other connected source, remains immaterial to a project of this scale. However a scaled trucking process brings a host of problems: Security, outage and accident exposure, haul-route permits, road-damage agreements with a county whose roads weren’t built for heavy use, dust complaints. Not to mention they still often rely heavily on the same municipal bulk-fill stations that can restrict or suspend out-of-district commercial sales without notice. Common during Southwest droughts, and politically volatile. Or when the main breaks or aquifers experience contamination. All compounded by exposure to water price increases over time. And even though the amount of diesel used by the trucks is negligible relative to that which is required by other power-generation methods, the sight of a constant stream of trucks going by does not create the best of optics for a facility that’s designed to be environmentally friendly.

AWG Stands to Close This Last Gap

Generating water on-site removes those headaches and, as the economics below shows, the water can pay for itself via reduced cooling capex.

The question on everyone’s minds is whether AWG can really work at scale. And, if so, at what cost?

Truthfully, no one’s tried it at scale in arid environments. Small-scale trials and detailed in-laboratory modeling suggest it’s possible. The modeled cost is lower than trucking in water once the units are installed, however the amortized capex brings the per-gallon price much higher.

Several factors are promising: The searing, bone-dry stretch is May and June. Those are also the longest, brightest days of the year. On these days, solar panels typically generate more kilowatts than the site’s batteries can absorb, and more than the dry coolers’ fans can consume at full tilt. There’s also not much water in the air during those months worth chasing anyway.

Then the monsoon arrives, mid-July into September. The wettest air of the year. Still-long solar days with excess solar power. Rain to help fill storage tanks for good measure. A timing issue: Solar peaks at noon, but desert moisture peaks between 10 p.m. and 8 a.m. So the midday surplus needs to be banked in the batteries and used to run the harvesters overnight when dew points are highest. The battery round trip costs about 13% of the energy. However, harvesting cool night air instead of 95°F afternoon air cuts the energy per gallon by roughly 40%. The haircut is smaller than the humidity penalty it buys, and there are still plenty of long days with excess battery charging power the next day.

The calendar hands the additional gift of phase matching. The only season we’d ever need extra water is during the searing early summer, which is when the temperature is highest and the coolers work hardest. Thankfully, that is the same season the sky delivers it. Supply and demand share the same season, offset by a few dry weeks. As a result, storage doesn’t have to carry water from September to June; it only bridges the few hot, dry weeks before the monsoon returns. Our sizing math puts that bridge at 25-35% of annual demand. By late September the tanks are full, and the machinery coasts through the mediocre winter months on storage, with capture of monsoon rain as a bonus.

To be sure, entirely waterless cooling works. As I wrote in my last article, the majority of non-discretionary water needs is the water needed for employee breakrooms.

However when excess water is available, it creates a discretionary option to provide some additional cooling the old-fashioned way: Giving a light spray on the radiator coils. This trims a surprising amount of fan draw right when the grid and the coolers are most stressed.

The durable value of AWG is therefore thermal insurance: It creates the ability to build a smaller radiator field to handle typical conditions rather than a massive overbuild for the extra cooling needed for a once-a-decade scorcher. Real capex saved to offset the cost of the water harvesters.

To put a hypothetical number on it: if a light spray lets a site trim its radiator overbuild from, say, 2x to 1.5x of typical-day capacity, the avoided capital would be on the order of $10 million at 100 MW and $100 million at 1 GW, more than the cost of the harvesters needed to keep a spray reservoir full (every site’s actual overbuild is its own).

At the end of the day, AWG’s per-gallon operating cost is expected to run below trucking’s, particularly when free excess power is used, as provided by on-site solar. Still, the cost of both is relatively immaterial for waterless cooling systems. The strongest case is capex savings and risk mitigation.

The Desert Viability Envelope

There is a sturdy viability envelope for AWG in the interior Southwest, roughly 3,500 to 7,000 feet in elevation east of the lower Colorado basin, where nights are cool and nocturnal humidity climbs past 50–60 percent.

Tucson averages around 53 percent humidity in the morning. And from mid-July through September, the North American monsoon pushes dew points past the mid-50s°F. That seasonal moisture is a real, harvestable and renewable resource.

Inside the envelope: Albuquerque, Santa Fe, Las Cruces and Doña Ana County, El Paso, Lubbock and Midland, Flagstaff, Tucson and the Sierra Vista corridor.

In the failure zone: Phoenix, where afternoon humidity sits around 22–28 percent; Yuma; and above all Las Vegas, with an annual average dew point near 30°F and a weak monsoon, close to the worst case in the continental US.

The Texas Data Corridor

Our project site in Haskell, Texas is semi-arid and transitional, receiving roughly 24-26 inches of rain per year, primarily during a May-June peak and a secondary bump in October. Using Abilene as a climate proxy, July humidity runs around 72% at 6am and drops to about 38% by mid-afternoon. That diurnal swing is the whole story: a machine rated for 24/7 nameplate output here would disappoint, but a unit tuned to harvest at night and store water in closed tanks fits the climate’s natural rhythm.

The hyperscalers building out the corridor, which include OpenAI, Oracle, Microsoft, NVIDIA and Crusoe, haven’t detailed how they’ll source water, though one announcement stated that Texas facilities “will use advanced air-cooling technology, limiting water use to site operations like kitchens” (implying the breakroom assertion above). Notably, the newest campuses are being co-located with utility-scale solar and battery storage, designed around the same closed-loop philosophy this series describes. That pairing is precisely the profile where night-harvesting AWG earns its keep: Abundant daytime power, cool nights to condense, and a water demand already trimmed to site operations. If air-to-water fits anywhere in Texas, it fits here.

Does the Desert Still Beat America’s Temperate Climates?

Here’s the inconvenient geography: Most existing US data-center capacity doesn’t sit in a desert. It sits in humid and temperate regions, mostly favorable to AWG technology.

Northern Virginia’s “Data Center Alley,” centered on Loudoun County, is the largest data-center market on Earth, with roughly 370 facilities in operation or under construction. Data centers in Loudoun County used on the order of 900 million gallons of potable water in 2023 (up 250% from 2019); statewide, Virginia data centers used about 2.1 billion gallons that year. Like the desert, water use peaks in the summer, exactly when drought risk is highest. During the 2024 drought, several Loudoun towns imposed mandatory water-use restrictions, while Loudoun Water, which supplies the data centers, issued voluntary conservation requests. The Potomac River supplies about 75 percent of the water for five million people in the region, and data center consumption continues to accelerate.

The scrutiny has reached the desert too. In late August 2026, New Mexico’s Supreme Court stayed a well authorization for the 2.5 GW Project Jupiter campus in Doña Ana County amid groundwater challenges; notably, the project’s own answer, a closed-loop system filled once with about 11 million gallons and topped off at roughly 4,000 gallons per year on purchased farm water rights (per project statements), is exactly the water-light architecture this article argues for.

Thankfully, the physics for Northern Virginia is condenser-friendly. Summer dew points in the Mid-Atlantic routinely sit in the 60s and low 70s°F from June through September. This is humid enough for a plain refrigerated AWG unit to run near its rated output, albeit with less “free” excess solar power to run them. Atlanta is similar. Columbus and New Albany, Ohio (where Intel, AWS, Google, and Meta are building fast), as well as the Dallas-Fort Worth and Chicago corridors, all share warm, humid summers.

The catch is the calendar. Shoulder seasons are moderate. Winter collapses output, because condenser AWG is effectively dead below about 40–50°F and low absolute humidity, requiring freeze protection or indoor siting. Modeling from the seasonal-curve logic, a defensible estimate is that a well-sited temperate unit could reach an annual capacity factor materially higher than a high-desert one, plausibly in the 25-40% range.

Unsurprisingly, the humid East out-harvests the desert. 25-40% in the East versus 5-25% in the desert. However, the East sits on cheap municipal taps, which turns AWG into an insurance product rather than a massive capex offset.

The Physics, in Plain Language

Every method of making water from air bumps into the same physics. It takes energy to boil water into vapor, and that same energy comes back out into the air when the vapor condenses (it’s why steam feels hot). Any apparatus that turns water vapor from the air into liquid form has to carry all of that latent heat away. That heat equates to about 0.63 kilowatt-hours for every kilogram of water.

Here’s the reprieve: A refrigerator doesn’t burn a kilowatt-hour of electricity for every kilowatt-hour of heat it moves. Rather, it pumps several units of heat for each unit of electricity it uses. So the actual power bill lands below the latent-heat number. The best machines, running in warm, humid air, get down to roughly 0.25-0.35 kWh per liter. One recent field study of a commercial unit in Abu Dhabi matched that band in its best months, though most field results land 2-10x higher.

The catch is that everything about this gets harder as ambient air dries out. The single most important fact about AWG is that its appetite for power swings wildly with humidity. Not with the season, but with how much moisture is actually in the air. The same machine that sips energy in a Gulf Coast summer where water is easier to pull out of the humid air becomes an electricity glutton in dry desert air.

This is why AWG’s nameplate ratings are, like solar panels, subject to wild swings based on local weather conditions. As a result, few data centers consistently experience the warm, humid sweet spot the nameplate ratings assume. An AWG machine rated at 1,000 gallons per day, therefore, might deliver a small fraction of that in Phoenix in June. The only number that matters to a buyer is the site-specific, hour-by-hour output curve for their actual location and their actual seasons.

Three Ways to Harvest Water From Air

1. Refrigeration Condensers

A classic dehumidifier, industrial scale. Fans pull ambient air across coils that are chilled below the dew point. Vapor condenses, drips into a tank and gets filtered, UV-treated and mineralized. Cheap and mature, however with significant limitations. Output tracks dew point, not relative humidity, and when the coil has to run near freezing, frost chokes the fins and output collapses toward zero. Below 40-50°F, or below 30–35% humidity, a bare condenser is mostly a fan. Who ships these (or hybrids): Watergen Ltd, Aquaria, GENAQ, Airowater Global Water Technology and Altitude Water.

2. Solid sorbents

A sponge, enhanced. Certain materials are riddled with pores so fine and so numerous that a single gram has the internal surface area of a field of tennis courts. Imagine a sponge that large.

Silica gel (the little packets in shoeboxes and Tylenol containers) and zeolites are the familiar examples. They pull water straight out of the air by adsorption: Water molecules stick to those large internal surfaces on contact with no chilling required, even at humidities that would defeat a condenser. Then a gentle heat is all it takes to shake the water loose into a small, sealed chamber. Since that chamber is tiny and now packed with vapor, the water condenses easily inside it.

Capture is nearly free, just some fan power, so most of the energy is spent on the heat used for regeneration.

That split is the superpower for solar power: Capture at 4 a.m. when the air is willing, and regenerate at 1 p.m. when there’s ample solar power to generate the heat.

The sorbent is effectively a moisture battery, storing the newly-extracted water until the sun comes up. Something a condenser simply can’t do.

Old desiccants needed 200–400°F to release their water; the new engineered ones let go at around 115°F, which is low enough that mild solar heat, or warm waste heat, can get most of the way there, leaving less work for supplementary heating systems. Some newer sorbents also achieve some water release from natural atmospheric swings, alleviating a level of reliance on heat.

Sorbent-based approaches have faced challenges commercializing at scale, largely due to material degradation, equipment cost and, in some cases, health risks. Some frameworks slowly break down in the very water they capture, letting trace metals or linker molecules leach into the product water. Damp internal surfaces can also host bacteria and biofilm, which clog pores and cut water yield over time.

Who ships (or nearly ships) it: Wahaso - Water Harvesting Solutions, Inc., AirJoule and Uravu Labs ; Atoco is pre-order. SOURCE Global’s hydropanels were the passive solar version until they ceased operations in 2025.

3. Liquid desiccants

A brine that drinks the sky. Concentrated salt solutions such as lithium chloride, lithium bromide, calcium chloride and/or triethylene glycol are hygroscopic, meaning the dissolved salt is so “thirsty” that it pulls water vapor out of the air on contact. The brine gets more diluted as it absorbs. Heat then boils the water back off, re-concentrating the brine to do it again. Industrial dehumidification has run on this chemistry for a century.

The AWG twist is using the brine as a concentrator stage: Sponge moisture from a big, lazy airflow, hand it to a small internal stream, and condense that instead.

Boveda Inc., the popular cigar humidor packets, use a saturated salt solution and the same hygroscopic principle. They hold humidity at a fixed set-point by releasing or absorbing water rather than harvesting it, thereby providing consistent humidity that’s optimal for humidors.

Who ships it: In its pure form, not many at scale. Sweden’s Drupps specializes in vapor streams and waste heat from industrial processes. They remain active in that niche, most recently commissioning a commercial plant at a Mexican tile maker in 2025, though they harvest industrial vapor rather than ambient air. Genesis Systems® (USA) - Water Your Way patents describe a hybrid of this methodology and condensation.

Jason Bak's feature on AWG continues later this week with an update on the Nobel Prize winner, the US Army’s adoption of AWG and more!