Technology

AI Is Thirsty. This Data Center Developer Thinks It Found A Fix

Andreas Lischka, CC0, via Wikimedia Commons

A Utah developer building what it says will be the world’s largest data center campus is betting on an old scientific principle to tackle one of artificial intelligence’s most pressing problems: water.

Creekstone Energy announced Wednesday that it is seeking to patent GIGAWATER, a technology the company says can capture water vapor produced by natural gas generation and turn it into usable water, according to the Utah News Dispatch.

The company has dubbed the novel water stewardship effort “Project Oasis.” Creekstone is developing Delta Gigasite in Millard County, Utah, where it plans more than 10 gigawatts of electrical capacity at full buildout, more than twice the electricity currently consumed by the entire state of Utah, per the Dispatch.

Creekstone Energy did not immediately respond to The Daily Caller News Foundation’s request for comment.

The planned data center is set to use dry cooling and consume conventional water primarily for drinking and bathrooms, according to the Dispatch. If GIGAWATER performs as projected, the company says the campus would produce “vastly more” water than it consumes.

“We at Creekstone Energy, we don’t believe Utah should have to choose between economic growth and responsible water stewardship,” Creekstone founder and CEO Ray Conley told reporters during an embargoed briefing, the outlet reported.

The announcement comes as the AI data center boom increasingly collides with concerns over the industry’s voracious demand for electricity and water. Communities across the country have pushed back against proposed data centers over concerns ranging from increased utility bills to water and electricity demand.

Creekstone expects about 2.1 gigawatts of its planned power supply to come from natural gas, along with six gigawatts of solar and 1.8 gigawatts from the nearby Intermountain Power Project, per the Dispatch.

The developer also projects GIGAWATER could recover approximately 4,000 acre-feet of water per year from its natural gas generation — roughly 1.3 billion gallons — according to the outlet.

“Now, that’s just a projection,” Conley told the Dispatch, adding the figure is “based on the chemistry and the physics of what we’ve done.”

New Technology, Old Science

The alleged underlying chemistry is much older than the AI boom. English scientist Henry Cavendish found in the 18th century that the “dew” produced when what was then called “inflammable air” was burned was actually water, according to the American Chemical Society (ACS). French chemist Antoine Lavoisier subsequently reacted oxygen with the gas now known as hydrogen in 1783 and correctly concluded that water was a compound of hydrogen and oxygen.

Lavoisier went further and actually collected the water produced by combustion. In his 1789 Elements of Chemistry, Lavoisier described water depositing on the inside of his experimental vessel, accumulating into large drops and running to the bottom where researchers could measure the amount produced. Lavoisier and fellow scientist Jean-Baptiste Meusnier used the experiment to determine the proportions of oxygen and hydrogen that form water.

More than two centuries later, scientists have explored whether the same basic principle could be harnessed to recover water from natural gas power generation.

Researchers Erfan Ghamati and Javad Mirrezaie Roudaki proposed a theoretical system in a 2022 study that would recover water from the exhaust of a 320-megawatt natural gas-fired unit at Iran’s Bandar Abbas power plant. The proposed system used four condensing heat exchangers to cool exhaust below the dew point of its water vapor, causing the vapor to condense, according to the study.

Researchers estimated the proposed system could recover 52.8 tons of water per hour and potentially reduce the power plant’s water use by nearly 14%, according to the study.

But the researchers were note reporting results from a working system. Their paper explicitly described a “theoretical design” and modeled an “assumed condensing unit,” meaning the research demonstrated the potential on paper rather than documenting a commercial installation.

Potential Problems

Creekstone is now trying to prove this approach can work at a much larger scale. GIGAWATER would capture exhaust from the company’s natural gas generators and cool it using what Creekstone calls a “regenerative rock-bed system,” converting water vapor into liquid water that could be collected on site, according to the Dispatch.

Even if Project Oasis works as planned, the technology could create an unprecedented question for Utah lawmakers over who controls the water it produces. Creekstone expects the system to eventually generate more water than the data center consumes, potentially leaving the company with excess water that could be used for agriculture, irrigation or other purposes, according to the Dispatch.

Conley acknowledged there is no clear precedent governing what would happen to that water, including whether Creekstone would own or be able to sell it. The answer could affect the economics of Project Oasis, with Conley telling the outlet that a policy decision depriving the water of its economic value would factor into whether pursuing the technology is ultimately worth the investment.

“That’s actually condensed water in the tailpipe from the combustion process,” Conley told the Dispatch. “What we’ve done is figured out how to do that same process, only at a very large scale in the desert, to capture the water byproduct of combustion when we’re burning natural gas to make electricity.”

Whether Creekstone’s particular system can actually perform at the enormous scale the company projects remains to be demonstrated. Creekstone signed a research agreement with the University of Utah’s Energy & Geoscience Institute to independently evaluate the technology and determine whether its assumptions hold at scale, according to KSL News.

Creekstone hopes to build a working pilot near its construction site to test the technology, with university researchers expected to produce results next summer. The company expects to know within roughly two years whether the system can be scaled commercially, Conley told Utah News Dispatch.