BY MICHELLE TALSMA EVERSON
Artificial intelligence may feel invisible, but in Arizona, its footprint is anything but. Behind every AI search, chatbot prompt, or image generator is a network of massive data centers — physical facilities that require land, energy, and often significant amounts of water to operate. As Arizona continues to grow as a major U.S. market for data centers, the conversation is shifting from expansion to resources.
Some estimates suggest that a single large data center can use up to two million liters (over half a million gallons) of water per day, primarily for cooling. In a desert state already navigating long-term water challenges, that figure is drawing increased attention.
Still, experts say the issue is more nuanced than it may appear.
“Most people experience AI as something abstract — a chatbot, an image generator, or a search tool,” said Chris Deaton, executive director of the Responsible Innovation Lab. “But behind those interactions is a very physical infrastructure. AI is infrastructure now. It is not something we can choose to do without.”
That shift from software to infrastructure changes the conversation. Like electricity or telecommunications, AI requires long-term planning, resource management, and public accountability.
A GROWING FOOTPRINT IN A WATER-STRESSED REGION
Arizona’s appeal to data center developers is no accident. The state offers available land, a relatively low risk of natural disasters, and a growing technology ecosystem anchored by semiconductor manufacturing and cloud computing.
But growth has raised questions at the local level. In recent years, communities such as Chandler have engaged in public debate over proposed data centers, with residents raising concerns about water use and long-term sustainability.
According to Sarah Porter, director of the Kyl Center for Water Policy at Arizona State University, data centers are one piece of a much larger water landscape.
“In spite of the region’s significant urban growth and industrial development over the last few decades, agriculture still uses more water than any other sector,” Porter said, referencing a recent Kyl Center report examining large water users in central Arizona.
The report underscores that while data centers are drawing scrutiny, agriculture remains the dominant water user in the region. Called “From Copper, Cattle and Cotton to Chips and Cloud Computing,” the report found that while comprehensive data on water use by data centers is still limited, most facilities in central Arizona are served by municipal providers with a Designation of Assured Water Supply (DAWS) from the Arizona Department of Water Resources. Under this designation, providers must demonstrate they can meet projected water demand for decades before approving new development, helping protect existing users as new high-volume customers come online.
“The DAWS process protects existing water users from impacts of new high-volume users and forces providers to consider what types of water demand they want to take on,” Porter said.
In some cases, large water users may rely on unreplenished groundwater, a scenario that warrants careful scrutiny from policymakers given Arizona’s broader water challenges.
NOT JUST WATER, BUT ENERGY
Water is only part of the equation. Data centers also require large amounts of electricity, and energy production can carry its own water demands depending on the source.
“Indirect water consumption for data center power needs is also an important consideration,” Porter said.
Arizona utilities have improved water efficiency as they incorporate more renewable energy and natural gas, though future demand remains uncertain.
For Deaton, this highlights the need for a more grounded conversation.
“The key question isn’t whether these systems use resources. That’s an established given,” he said. “All infrastructure does. The real question is how responsibly they are designed and managed.”
Responsible innovation frameworks push that thinking further, encouraging companies to ask who benefits from new technologies, who may bear unintended costs, and how risks are addressed before systems scale.
That means shifting the focus toward transparency, planning, and engineering decisions rather than treating AI as purely digital.
TECHNOLOGY IS EVOLVING, AND SO ARE EXPECTATIONS
Not all data centers use water in the same way. Cooling systems, which drive most water use, range from water-intensive evaporative cooling or closed-loop systems that reduce consumption, to newer air-cooled or immersion technologies that can significantly limit or eliminate water use.
“We’ve found that data centers use a range of cooling technologies,” Porter said. “Communities should insist that data centers they serve have a water use plan that meets the community’s needs.”
To reduce water consumption, some companies are exploring alternatives such as liquid immersion cooling and air-cooled systems. Others are using recycled or non-potable water where available.
“There is definitely progress happening,” Deaton said. “We’re seeing experimentation with zero-water cooling systems, renewable-powered data centers, and AI-optimized energy efficiency.”
Those innovations are driven by both environmental concerns and cost. More efficient systems are often less expensive to operate over time. Still, transparency varies.
“Communities often don’t have clear visibility into resource use or long-term infrastructure planning,” Deaton said. He emphasized that the conversation should move away from whether to build and toward how to build responsibly through collaboration between industry, policymakers, and communities.
LOCAL CONTROL AND POLICY TOOLS
In Arizona, water policy decisions largely happen at the municipal level. Cities manage water delivery and establish policies for large users, which means approaches can differ across the region.
A recent Kyl Center report found that 10 municipal water providers, serving more than four million Arizonans, have adopted ordinances or other measures to regulate large-volume water users. Chandler was the first city in the United States to implement such an ordinance.
These policies help define expectations between communities and developers, including how water use is managed and what benefits projects bring.
“They capture the community’s expectations in terms of the return on investment from a high-volume user,” Porter said.
For residents, that means local engagement can influence how projects move forward.
WHAT COMES NEXT
As AI continues to expand, so will the infrastructure behind it. For Arizona, the challenge is balancing economic opportunity with environmental limits.
Deaton said the focus should not be on slowing innovation but guiding it responsibly.
“Responsible growth means recognizing two realities at the same time: AI infrastructure is becoming economically essential, and the Southwest has real environmental constraints,” he said.
That means improving efficiency, increasing transparency, and strengthening long-term planning around water and energy use.
“The goal isn’t to slow innovation,” Deaton said. “It’s to ensure the systems we build today remain viable as we manage this transition.”
For Arizona, that balance will shape not just the future of technology, but the sustainability of the communities supporting it.
