Data centers use a lot of water

Iowa’s affordable land and affordable electricity rates make it an attractive destination for data centers from the world’s tech giants.

Tech companies also love wind power in Iowa, which has one of the highest rates of renewable energy in the nation. About 60% of Iowa’s electricity comes from renewable energy, allowing tech companies to power Iowa’s data centers while working towards ambitious climate change goals for low-carbon electricity. Masu.

While Iowa has an abundance of green energy, there’s another important resource needed for data centers that Iowa often lacks. It’s water.

While data center operators power the servers that keep the internet running and enable artificial intelligence, they also need large amounts of water to cool the servers so they don’t overheat. Increased water consumption by data centers is a challenge for some host communities.

“There are areas in Iowa that are clearly starting to feel water stress,” said Kelly Johansen, energy program director for the Iowa Environmental Council. newsweek.

Johansen’s group tracks increases in water use by the state’s data centers, most of which are concentrated in the suburbs around Des Moines.

“We have a data center in a suburb called Altoona that uses up to about a fifth of the city’s water, so that’s very important,” she said.

Iowa is in the midst of its longest drought in decades. The National Oceanic and Atmospheric Administration reported that by mid-March, about 85 percent of the state was in drought conditions, and about 56 percent of the state was in extreme or severe drought.

State geologist Keith Schilling warned earlier this year that groundwater levels in some of the state’s aquifers are declining and the state “needs a plan to protect our groundwater reserves.” In addition to traditional water uses the state needs to balance, such as agriculture and housing systems, Schilling also cited “data centers that require a lot of cooling water.”

The boom in AI is increasing demand for water. Recent studies have shown that the massive computing power, larger chips, and additional servers required for AI will not only significantly increase power demands but also make many data centers even thirstier. I am.

Researchers say AI is already likely to cause increased water use by some technology companies, and predictions for the next few years suggest that global AI growth will be accompanied by some small countries consuming more water. It has been shown that more water may be required.

Many large technology companies are in a position to rapidly expand their operations to support AI, but they must also consider growing water demands. Additionally, some data center host communities in water-stressed regions may face difficult choices due to a lack of hard information about how much water to give up.

What makes AI so hungry?

Xiaolei Ren, an associate professor in the Department of Electrical and Computer Engineering at the University of California, Riverside, has been studying water use at large technology companies for about a decade.

“When researching sustainable computing, people were talking about carbon emissions, but I believe water is also a very important metric,” Ren said. newsweek.

Ren’s latest research focuses on exactly how AI is increasing water usage. OpenAI’s favorite requires first training large language models like ChatGPT-3, which is a data- and energy-intensive process and can also increase water usage. there is. Ren discovered that when he trained GPT-3 at Microsoft’s high-end data centers, it could directly evaporate 700,000 liters, or about 185,000 gallons of water.

With AI models, each inference and response to a query also requires energy and cooling, which can be a pain. Ren et al. estimate that GPT-3 would need to “drink” a 16-ounce bottle of water roughly every 10 to 50 responses, and as the model processes billions of queries, that amount will increase. It will be huge.

To fully understand water usage, Ren looked at both the water used directly to cool the servers in the data center and the water used to generate the electricity the data center requires.

“When you use coal-fired power plants and nuclear power plants to generate electricity, you’re actually using a lot of water as well,” he said, adding that AI is significantly increasing the energy needed for data centers. explained.

Add all of that and apply it to the growth in AI that is projected over the next few years, and the sum quickly reaches staggering levels. Ren concluded that global AI demand could generate up to 6.6 billion cubic meters of water withdrawal by 2027, or what he called 8.6 billion cubic yards. To put this into perspective, we compared it to the amount of water some countries use.

“This is roughly equivalent to the water withdrawals of four to six Denmarks,” Ren said. “That’s quite a lot.”

Microsoft AI training

Ren said Microsoft’s total water use increased by 34% in 2022, the most recent year for which data is available, and that increase is likely due, at least in part, to the demand for AI. said. This has big implications for areas like Iowa, where Microsoft uses its data centers to train OpenAI models and where water use is a concern.

Microsoft opened its first data center in Iowa in West Des Moines in 2012 and has built four more data centers in the region by the end of 2023. These facilities include the Azure supercomputer that Microsoft built for OpenAI to train his AI models, forming “America’s Small City.” “Heartland is unlikely to be the epicenter of the AI ​​revolution,” the company said in a press release last year.

A Microsoft spokesperson declined to comment on Ren’s water usage calculations, but the company has set a goal to become “water positive” by 2030, meaning that it uses more water than it uses. He pointed out that the goal is to return more water to the environment. Microsoft said it also achieved an interim goal of reducing water waste by 95 percent.

Microsoft ranked 34th newsweek‘s 2024 list of America’s Most Responsible Companies, ranking No. 5 among software and communications companies.

Bobby Hollis, vice president of energy at Microsoft, said energy and water demands are closely intertwined and must be considered together to optimize data center sustainability performance.

“You have a kind of yin and yang, a balance between the two, and you’re trying to figure out which one to focus more of your efforts on,” Hollis said. newsweek.

When clean energy is plentiful and water is scarce, the company uses more air-cooling methods to reduce the impact of water, and closed-loop systems prevent losses through evaporation, he said.

“Of course, it depends on the geography and what you’re looking at in a place, like how dry it is or how cool it is,” Hollis says.

Iron Mountain’s Deep Thoughts

At its data center in Boyers, Pennsylvania, just north of Pittsburgh, Iron Mountain literally dug deep into its sustainability solutions. The company’s facility is located more than 200 feet underground in a former limestone quarry, naturally cooling its servers and providing a water source that doesn’t draw water from nearby communities’ systems.

“We utilize a large underground lake to cool our data center, and we regularly monitor the water level and temperature of the lake,” said Chris Pennington, director of energy and sustainability at Iron Mountain. Ta. newsweek By exchanging emails. “This is a feature that worked extremely well and helped make this facility the first underground facility to receive Energy Star certification.”

(The U.S. Environmental Protection Agency’s Energy Star certification recognizes facilities that meet energy efficiency standards.)

Pumps lift water from the lake to a heat exchanger, where the cold water carries heat away from a closed-loop cooling system that circulates between servers. That water is then returned to the lake. Because it’s underground, the site is also highly secure, which is a selling point for some Iron Mountain clients who handle sensitive data.

Innovations like this earn Iron Mountain four stars. newsweek‘s 2024 ranking of America’s Greenest Companies.

“However, not every data center can be located underground with its own lake,” Pennington said, adding that the main lesson to be learned from the Iron Mountain underground example is that when building a data center, This may mean thinking deeply about the characteristics of the region.

Equinix is ​​one of the world’s largest data infrastructure providers, occupying slot number 181 out of 600. newsweekRanking of America’s Most Responsible Companies.

Christopher Wellies, Vice President of Global Sustainability at Equinix. newsweekwho happened to be in Scandinavia, provided some regional examples of how the company is deftly adapting its data center operations to local conditions and needs.

“Here in Scandinavia, we have a lot of air cooling at our disposal,” Wellies says. “But it could provide valuable heat that can be used for other purposes.”

Unfortunately, in most environments in the United States, there is no easy way to harness that heat. But in European cities with district heating systems, “waste heat” from data servers can be routed through a central network of pipes and ducts to heat homes and buildings, Wellies said.

“We heat thousands of homes in Helsinki,” he said. “You have to think about these things in terms of managing the entire lifecycle, not just the electrons you’re using and how green or brown they are.”

That overall idea will have its highlight moment in July, when athletes gather in Paris for the Summer Olympics. The new Olympic Aquatic Center pool will be connected to Equinix’s data center, and diving and swimming competitors will be kept comfortable using the extra heat generated by the data servers. This is an interesting example of the possibilities available to us in this new AI era.