Up against a power crunch and growing concerns about water scarcity, data center developers in the artificial intelligence race are seeking novel solutions, from nuclear power to exotic cooling technologies to buying the entire output of solar and wind farms.
Neither water nor power should be a problem for a massive project underway in central Utah. Mark McDougal, cofounder of developer Joule Capital Partners, is converting parts of his family’s 4,000-acre alfalfa farm to build multiple AI data centers. The farm has historical water rights to a groundwater aquifer; project engineers estimate that the data centers will use 75 percent less water than if the land were farmed each year, partly due to efficient cooling systems.
McDougal also plans to build an array of gas-fired generators along with battery storage. That means no lengthy wait times to hook up to the local utility and no impact on residents’ electricity prices.
The Joule campus is a unique, not easily replicable spin on the dilemma of water- and power-hungry data centers. Some developers are building their own power plants, often gas-fueled, to skirt the overburdened, outdated U.S. grid. Some are seeking private water rights — in McDougal’s case, family-owned. Others are trying to keep their water consumption on the down low: Earlier this year, judges in Colorado ordered the Denver Water Board and the city of Colorado Springs to turn over information on data-center water usage to Business Insider, after the officials claimed the information was confidential.
Water is increasingly critical in Southwestern states like Utah, which are engulfed in a decades-long drought. Reliable power, and plenty of it, can present another challenge.
“We started working on this a decade ago, and realized that the public utility was quickly becoming outdated and overburdened,” McDougal said. “There’s all this demand for electricity from electric cars, our homes, and now, of course, data centers. To solve for that, it became clear that we’d need to build our own power grid.”
Water/power dilemma
For companies seeking to align their digital footprints and sustainability goals, the energy and water use of AI data centers requires a complicated balancing act that few have mastered. Only 21 percent of the more than 1,5000 companies surveyed by S&P Global in May said they quantify the impact of their AI initiatives on sustainability goals.
Big Tech companies, including Google, Amazon, and Meta, have pledged to be “water positive” by 2030 — meaning they will replenish more water in localities than they consume — and to achieve net-zero emissions by that same date. Microsoft has the same water goal and aims to become “carbon negative” by 2030, eliminating more carbon from the atmosphere than it emits. But the rapid buildout of data centers is undercutting those goals.
“Before companies site data centers, they need to think about water stress in the region,” said Kirsten James, senior program director for water at advocacy and data nonprofit Ceres.