Getting your
Trinity Audioplayer ready...Long before data centers became fodder for national discourse and the future of artificial intelligence, Adam Wierman was studying them.
Wierman, a computing and mathematical sciences professor at Caltech in Pasadena, began his research in the early 2010s when the need was for cloud computing. The data centers at that time were smaller, yet he helped them use electrical energy more efficiently, as well as less water for cooling down banks of computers.
More recently, he began tackling the issues swirling around the mega, hyperscale data centers the size of several football fields bombarding cities and states that Big Tech says are needed to support artificial intelligence. That has culminated into more recent findings, including a Caltech Science Exchange report released on July 2.
Without ignoring the environmental and economic impacts, Wierman found that there may be a middle ground between data center critics and their proponents — something that could not come too soon.
At his office last week, he ran down the list of environmental and economic impacts that large, hyperscale data centers would have on local communities.
Air pollution from diesel generators, constant noise from humming equipment, plus water and electrical power demands that are off the charts and strain limited resources, resulting in higher consumer utility bills.
These issues land amid an acute national discussion over data centers that has hit home Southern California. In a matter of months, local city councils have found themselves faced with a data center future — or not in their town.
Cases in point:
In Monterey Park voters recently became the first city in California to outright ban them, and where others appear to be following suit.
In El Segundo, a developer pulled the plug on a plan to convert a hotel into a data center after community backlash in the South Bay.
In Chino, a planning board on Wednesday narrowly rejected a plan to develop a so-called battery energy storage system, which many critics see as a precursor to data centers. And in Alhambra, on Tuesday, the City Council took a page out of Monterey Park’s book, looking to voters to decide the fate of data centers in that city.
At their core, such data centers seek to house the computing power capacity that fuel the future of artificial intelligence, needing much more capacity of computing, and therefore using more electricity and cooling water for massive strings of computers than what was needed in the 2010s.
The energy and water needed, plus concerns about air pollution, has unleashed concerns among communities about the availability and potential costs, not just to their pocketbook but also to the environment.
It’s nothing new for Wierman, who’s research is at the center of understanding the technology, its effects and how to mitigate the latter.
Even as local municipalities, interest groups and developers talk it out, there is the possibility of an elusive “win-win” scenario when it comes to the giant hubs, he said.
Can data centers be ‘done right’?
While laying out the detrimental effects, his research, and also his suggestions to cities targeted for these data centers driving artificial intelligence is to make deals on the developer’s dime that limit the damage and include building infrastructure like reservoirs and electric generating plants. “When done right, the data center is on the hook. When done wrong, you underestimate the cost and the ratepayers pay the bill,” Wierman said.
To get to what he called a “win-win” game, the municipality or county must “be aware of the negatives,” which are often left hidden by developers in a lack of transparency.
Next, both sides should negotiate agreements requiring the developer to build their own power generators on site and make them solar-powered. Some data centers can use up to 5 million gallons of water a day, so to facilitate massive water usage, the developer can build a new reservoir or pay for the cost of generating water, so the burden doesn’t fall on residents.
He suggests the developer invest in local schools as a condition of approval, or pay for other community needs.
But managing these effects as trade-offs is not always possible. Or desirable.
Water is an issue
Take water, which is needed in large amounts each day to cool down always operating computers.
About 80% of data centers built last year were in drought areas, where water supplies are an issue, he said. No amount of money can make it rain. So far, Virginia is No. 1, with 566 data centers, followed by 391 in Texas and 287 in California, according to the World Resources Institute.
Southern California is reliant upon Northern California snowpack, which was only at 18% of its historical average on April 1. And another imported water source, the Colorado River system, is in an extreme drought. So water stored in underground aquifers or above-ground reservoirs are major sources, but when those run low, Southern California has been placed on mandatory water restrictions.
Air pollution in Southern California
The south coast air basin also has the poorest air quality in the nation. This will only get worse — especially in local hot spots — if the alternate generators that run the computers during brown outs or other power shortages are burning diesel or other fossil fuels that produce particulate pollution.
Particulates, especially PM2.5, get into the lungs and blood stream causing lung and cardiovascular disease. They are already a difficult pollutant to control in Southern California due to tailpipe emissions.
“In Memphis, these generators (at data centers) led to black smoke from diesel-fuel burning and that leads to horrible air pollution,” Wierman said.
Community revolt, bans
Locally, any kind of negotiated community benefit agreement was dismissed out of hand in Monterey Park. The ballot measure banning data centers in the west San Gabriel Valley city passed with an overwhelming 86% of the vote. It was the first city in the United States to permanently ban data centers via a public vote.
That was in part caused by the proposed use of 24 diesel generators, each one with a tank holding 7,500 gallons, a potential for creating bad air quality about 500 feet from homes. Also, residents who worried about rising water and electricity rates organized against the center, forcing the hand of the City Council. After a moratorium, the Council put the measure on the ballot in June.
“We would’ve seen a dramatic impact to our water use and to the cost,” said City Councilmember Jose Sanchez. “So the residents were extremely concerned.”
Sanchez was quite aware that turning down the data center would cost the city revenue. But he also was aware of other nearby cities — San Gabriel, Montebello and Alhambra — looking to see what Monterey Park did.
“We share the same air, the same electricity,” he said.
In fact, Wierman’s making-deals-with-developers approach took another blow this week. On Tuesday, July 14, the Alhambra City Council voted to place a ballot measure banning data centers. El Monte could be next, since it passed a temporary moratorium earlier this year.
The San Gabriel Progressive Action group organized communities against data centers, including pushing for the ballot measure to ban them in Alhambra, said Amy Wong, co-founder. If approved in November, the ban would be in place forever and could only be changed by another vote of the people, she explained.
Of great concern by residents in Alhambra was noise and air pollution. Many working-class residents already struggle with inflation at the supermarket and at the gas pump, so the prospect of higher water and electricity bills was also a major threat, she said.
The community rallies and successful ballot measure in Monterey Park set off a ripple effect, now felt in Alhambra and spreading, Wong said.
In the community of Imperial, 115 miles east of San Diego, a not-in-my-backyard movement has sprung up in the small city. Residents are concerned about a data center that would take up 950,000 square feet, the equivalent of 17 football fields of land if completed by 2028, the newsletter reported.
If built, as the developer expects, the two-story data center could be the largest operating statewide, taking up 17 football fields’ worth of land, according to a June 16 report by Inside Climate News, a nonprofit, non-partisan news gathering group covering climate, energy and the environment. Inside Climate News reported the $10 billion, 330-megawatt data center would use 750,000 gallons of water a day to operate, according to developer Sebastian Rucci, who said electricity and water costs won’t rise due to the data center.
However, the newsletter said the project is facing litigation. And that impacts on water supplies and cost of water and electricity are to be determined.
Is AI worth the cost?
Wierman and colleague Shaolei Ren, an associate professor of electrical and computer engineering at UC Riverside, have studied health costs and water demands of large data centers. Ren told Inside Climate news that 90% of data centers in the U.S. get the majority of their cooling water from municipal systems.
During the hottest summer days, a large 100-megawatt facility can use about 1 million gallons of water for evaporative cooling. That amount is the same as about 10,000 people’s daily water use at home, Ren told Inside Climate News.
Wierman was asked whether these environmental downsides are worth the cost of extenuating AI in the United States so that programs such as ChatGPT and Google’s Gemini can continue to operate. AI data centers have been fast-tracked by state and federal directives, as a way to maintain Big Tech’s competitiveness with China.
He said AI is leading to advances in agriculture, healthcare and medicine.
“Scientists say there is a benefit to AI, which you can debate. But they say it is revolutionary in the way they can do science,” he said.
Also, he surmised that a lot of the opposition to data centers is really based on existential objections to AI, which is seen as using advanced computing that takes jobs away from people.
“Fresh drinking water should be used by humans, not by machines,” Sanchez said.