“We are now working to address the issue that renewable energy is not available all the time and everywhere by aiming to run on carbon-free energy 24/7 and to achieve net-zero emissions across all of our operations and value chain by 2030,” Learning & Sustainability Senior Vice President Ben Gomes and Chief Sustainability Officer Kate Brandt wrote in a Google blog.
In addition, it explained that it also evaluates and takes into account local water stress when deciding where to locate its facilities, how to design them, and how to operate them — from water systems in its offices to cooling systems in its data centers.
“Our goal is even with water growth, we will have a net positive impact,” he observed.
But scarcity is just one point of concern, he added. “If the data center uses air-side economizers and the outside ambient air is entering the data center, the water is treated for minerals or hard water, and bacteria treatment, as it will be inhaled by employees,” he told TechNewsWorld.
“What do they mean by replenish?” he asked.
Amazon Web Services also considers water stress levels when locating its data centers. “In certain places, we might look at the community and decide it’s not the right thing for us to use water to cool our data centers,” explained AWS Global Lead for Water Will Hewes.
Swenson asserted that it is likely that AI models will impact water unless data centers change their methods. “AI and new inferred compute loads with a multiplier effect will likely cause data centers to be built 10 times current levels, and possibly 20 to 25 times, over just the next five years,” he predicted.
Offsetting Demand With Efficiency
“There’s no water going back to those sources,” he continued, “and even if it did, it would kill the fish living there.”
Wes Swenson, a founder, investor, and CEO of Novva Data Centers, a privately-held data center company based in West Jordan, Utah explained that if water is a finite resource in any given area, and data centers enter the community, and they use water for evaporative chilling, then they can increase the pressure on water inventory.
“There’s also liquid immersion cooling, where you submerge the equipment in some kind of non-conducting liquid,” he added. “That has challenges, too. It’s newer. It’s more expensive, and it has maintenance issues because you can’t upgrade the equipment while it’s submerged.”
AI Impact on Data Center Water Demand
Google directs most of its water consumption towards cooling its global server network. Water cooling nets the company two benefits, according to the report. “We’ve found that our water-cooled data centers use about 10% less energy and emit roughly 10% less carbon emissions than our air-cooled data centers,” it explained.
“Clients of data centers could also be more wise in terms of cooling,” he added. “Raising operating temperatures would help, for instance, moving the thermostat from 70 to 80 degrees.”
If Google’s and everyone else’s data centers are thirsty for water now, that thirst could worsen as the demand to run artificial intelligence models increases. “AI will increase the demand for water by data centers significantly,” maintained Rob Enderle, president and principal analyst at the Enderle Group, an advisory services firm in Bend, Ore.
Google will have to hustle to meet that goal, however, since that 271 million gallons is only 6% of the company’s freshwater consumption in 2022.
Thirsty Data Centers
John DeVoe, a senior fundraiser and advisor of WaterWatch of Oregon, a water quality monitoring organization in Portland, Ore., cautioned that corporate promises about water usage have to be carefully scrutinized.
In 2022, Google reported, the total water consumption by its data centers and offices was 5.6 billion gallons — the equivalent of what it takes to irrigate 37 golf courses annually, on average, in the southwestern United States. Google was careful to note, however, wherever feasible, it tries to use non-potable sources of freshwater and alternatives to freshwater
“Water that is not lost to evaporation is flushed into the city’s wastewater system, where it must either be retreated or shifted to irrigation uses,” he continued. “It takes massive energy use to clean it and reclean it.”
Simmons pointed out that water-free systems have been around for years. “The argument against them is that they use more energy to run the systems than water systems, which is a myth,” he said.
“The problem is air is a less efficient cooling mechanism than water,” he told TechNewsWorld. “Air also doesn’t work in hot environments, so it’s not going to work as well in Arizona and Texas as it might in Canada or Northern Europe.”
He noted that in 2021, AWS data centers used about 0.25 liter of water per kilowatt hour. In 2022, it lowered that to 0.19 liter per kilowatt hour.
For the sixth year in a row, the company matched 100% of its annual global electricity use with renewable energy purchases, according to the 104-page report.