Do you use Chat GPT? Do you talk to Siri on your phone? If so, you’ve helped fuel the rise in data centers.
Now, the energy-hungry, water-thirsty centers are coming to places in the Southwest, including the lands of native peoples.
That was the topic of a panel discussion Friday in Window Rock, Arizona, organized by Diné C.A.R.E., a Navajo environmental organization.
Executive director of Diné C.A.R.E. Robyn Jackson said data centers have become a serious concern for the Navajo Nation. She said five centers have been proposed in and near the nation, three in Arizona and two in New Mexico.
The Arizona centers are in Leupp, Page, and Coalmine Canyon (possibly). The New Mexico ones are at Nenahnezad and Shiprock (a proposed expansion).
Northern Arizona University Professor Karen Jarratt-Snider said artificial intelligence, which is present in almost every aspect of our lives, requires huge amounts of energy.
Even fossil fuels will not be able to keep up much longer with the exponentially growing demand for energy from AI, she said.
Water required to cool the facilities is also enormous. Yet centers are being built in hot, arid states such as Arizona, even as it and six other states wrangle over how to allocate Colorado River Water.
Jarratt-Snider noted that tribal nations have water rights independent of those Colorado River water rights.
“But if there’s not enough water to go around now and we keep inviting data centers in who use more and more water, how long will it be before people come to indigenous nations and say, ‘we want to renegotiate your water rights’?” she asked.
In addition, AI data centers require cooling. Yet many are being built in states such as Texas. Thus they require water for cooling.
That raises critical questions for arid states such as Arizona – one of seven Colorado River Basin states embroiled in difficult negotiations over how to allocate water from that river.
In addition, AI data centers require cooling. Yet many are being built in states such as Texas, which is not generally cool in climate.
That raises critical questions for arid states such as Arizona – one of seven Colorado River Basin states embroiled in difficult negotiations over how to allocate water from that river.
Meeting host Professor Chris Jocks, chair of the Northern Arizona University Department of Applied Indigenous Studies and a Bear Clan Mohawk, said that on indigenous lands in the Southwest, “Local communities are being pressured to consent to projects without full information about what they are for, who will benefit from them and what will be left behind.”
People now can ask AI to do many things for them, he said, including writing papers and giving them relationship advice.
“But I, along with many colleagues and students I know, will continue to consult and rely on a different AI – ancestral intelligence,” he said.
The event was organized by Diné C.A.R.E., which is a community-based Navajo organization working for environmental health and justice; NAU’s Department of Applied Indigenous Studies; and the Southwest Environmental Health Sciences Center at the University of Arizona.
Jackson spoke about the impacts that resource extraction has had on the Navajo Nation through history.
“We have endured more than a century of oil and gas operations, including its associated pollution and waste,” she said.
“We've endured more than six decades of coal-mining and power-plant operations, including their pollution and waste. And we've endured legacy uranium-mining, milling, and waste.”
She said there are hundreds of abandoned and un-reclaimed uranium mines across the Navajo Nation, as well as thousands of abandoned and orphaned oil and gas wells.
“That means our community is left with the bill and cleanup responsibilities,” Jackson said.
Now, data centers have become a new concern because there is not enough energy and water for them, she said.
The biggest proposal, Project Jupiter in New Mexico’s Doña Ana County, could require 5,000 megawatts of power, Jackson said.
“But in New Mexico, the biggest utility company is PNM,” she said. “Their entire energy load is just 2,600 megawatts, so the one proposed project far exceeds what’s even available.”
She said as of 2024, ChatGPT used over half a million kilowatts a day, equivalent to the daily power use of 180,000 U.S. households.
“Even before data-center proposals and AI, we were in a climate crisis. This exacerbates the problem,” Jackson said.
Centers are often built quickly, she said, with little opportunity for communities to be involved in discussions and become aware of concerns.
Professor Karen Jarratt-Snider of NAU’s Department of Applied Indigenous Studies, another panelist, spoke about generative AI.
“Maybe you were looking at something online and you went on to something else,” Jarratt-Snider said, “and ads for the thing you were looking at start to pop up. That’s a low-level example of generative AI. It uses a lot of energy and a lot of water.”
ChatGPT is just one platform, she said. Others are much bigger, “so when you put in a generative AI request, it might take a minimum of ten times as much energy as a simple Google search used to take.”
Society uses generative AI for everything, she said.
“When you make a request of your phone, that’s generative AI. When you call a bank or doctor’s office and you get that automatic system . . that’s generative AI. It’s in almost every aspect of our daily lives unless we’re totally off the grid.”
Precisely how much energy it requires is unknown, she said, “because these are private companies and they don’t have to tell us.”
Jarratt-Snider said Amazon just built an AI warehouse on 1,200 acres of former cornfields in Indiana.
“If we keep replacing ag fields with AI data centers, that becomes a problem down the road for humanity,” she said. “How do we feed ourselves?”
Current energy sources cannot supply the exponentially growing demand of energy from generative AI, she said.
Another panelist, Imran Mithu of the University of Arizona’s Southwest Environmental Health Sciences Center, said writing something with ChatGPT takes about 500 milliliters of water.
“It’s a huge amount of water for one prompt,” he said.
In 2023, AI data centers used 4.4 percent of the total U.S. electricity, he said.
Data centers also emit a tremendous amount of noise from the operation of machinery, the hum of cooling systems, and the rumbling of diesel generators.
That can result in hearing loss, sleep disruption, even cognitive impairment, Mithu said.
Hardware in data centers has to be replaced frequently, creating a great deal of waste that cannot all be safely recycled, he said.
In a sense, data centers are a modern form of colonization, he said, as they are often put in places where people are poor, and only a few billionaires benefit from the centers.
Mithu said he believes better AI centers can be built. He said there should be ways to reduce the water used for cooling and to utilize reusable hardware that doesn’t create e-waste.
Panelist Logan Moya, a student in NAU’s Department of Environmental Sciences, raised the question of why data centers are being located in the Southwest.
“The Southwest is not really known for its water abundance,” she said, “and you’re bringing in these incredibly energy-intensive and water-consumptive data centers. It seems completely backwards to the direction we need to be going.
“We are 26 years into a long-term drought, the driest period this region has seen since 800 CE, which is further exacerbated by climate change,” Moya said.
“This year was the worst snowpack on record. The Colorado River Basin is in the middle of tense, deadlocked negotiations.”
Yet data centers are incentivized to come here, she said, drawn by tax incentives meant to promote business, a lack of labor laws, and cheap land.
Central Arizona Project water is relatively cheap in comparison to water in other states, she added.
But Arizona already faces significant cuts in water use because of the dwindling Colorado River Basin supply.
“Surface water is already over-allocated and extremely strained,” Moya said.
Water from the Colorado River provides only about a third of Arizona’s water supply, she said, but depletion of aquifers is also problematic.
A recent study using NASA satellite imagery indicated that many aquifers are in crisis levels, Moya said.
It makes sense that developers of data centers are turning to reservation lands, she said, because indigenous peoples have senior water rights as well as available lands.
She said the large tech companies try to frame what they’re doing as not environmentally destructive, “but there’s always going to be a trade-off somewhere.”
“If you use less water, you need more energy, and if it’s more energy-efficient, you use more water,” she said.
The panelists urged people to be alert to data-center proposals, to be informed about the projects, and to take action if they have concerns.
Jocks, the meeting host, called for “acknowledging with respect and love this beautiful earth our feet walk upon, our home, and for many of us our first mother, including all of the beings and elements with whom we share existence.”
The presentation can be viewed at https://youtu.be/C33oA3hUstE?si=ttkWr9BQf1lWfjvs
