AI, Water, Need

My post content

Mesa County faces a new danger, how will the County Commisioners decide what to do when confronted with the problems listed below?

The time to find out the answer to that question is NOW, not when the signatures are done.

AI data centers are projected to consume as much water annually as 1.3 billion people by 2030, creating severe local water stress in drought-prone regions like Arizona, Texas, and parts of Sub-Saharan Africa. While global data center water use remains a small fraction of total consumption, the local impact is intense, with large facilities using up to 5 million gallons per day—equivalent to the needs of a small town—and 80% of that water lost to evaporation rather than being returned to the source.

This resource drain disproportionately affects low-income and minority communities, a trend critics label digital redlining, where tech giants build resource-intensive facilities in areas with the least political power to resist. The diversion of potable water for server cooling has led to community protests, higher utility bills, and health risks, including contaminated water reports in areas like Memphis, Tennessee, and Newton County, Georgia.

To mitigate these impacts, experts recommend responsible AI ecosystems that prioritize water recycling, closed-loop cooling systems, and transparent community consultation during the siting process. Without such guardrails, the expansion of AI infrastructure threatens to exacerbate existing water inequalities and public health crises for the communities that host the technology.

The Scale of the Problem

The numbers have grown dramatically. By 2025, U.S. data centers were consuming nearly one trillion liters (~264 billion gallons) of water annually, driven largely by AI workloads. Google alone consumed 10.9 billion gallons in 2025, up 34% year-over-year and more than double its 2021 level, according to its 2026 Environmental Report. In Texas, data centers used roughly 25 billion gallons in 2025 (direct and indirect), with projections reaching 399 billion gallons by 2030 per a Houston Advanced Research Center study. A UN University investigation warned that by 2030, AI's water use will match the needs of 1.3 billion people.

The water is consumed at three levels: on-site cooling (evaporative systems use 70–90% of withdrawn water, versus 5–15% for residential users), power plant operations that generate the electricity to run the servers, and semiconductor fabrication, where a typical chip factory uses about 10 million gallons of ultrapure water per day. A single AI text response from a large language model consumes roughly 10–50 milliliters of water; generating an AI image uses about 23 milliliters.

Why It Hits Certain Communities Harder

The core issue is local concentration, not national volume. A single large AI data center can use up to 5 million gallons per day—equivalent to the needs of 10,000–50,000 people. The World Economic Forum notes that roughly 40% of the world's data centers are clustered in areas of high or extremely high water stress, and demand peaks in summer when communities and farmers already face shortages.

Specific cases illustrate the pattern:

• The Dalles, Oregon (pop. ~16,000): Google's data centers consumed 550 million gallons in 2025, close to 40% of the city's total water use.

• Newton County, Georgia: A Meta data center uses 500,000 gallons per day—10% of the entire county's consumption. Residents report disrupted private wells, sediment-clogged appliances, and one resident told the BBC: "I can't drink the water."

• Memphis, Tennessee: xAI's facility raised alarm over daily withdrawals from aging public water infrastructure, while predominantly Black neighborhoods face additional pollution from gas-powered turbines operating without permits.

• Phoenix, Arizona: Data centers already use 385 million gallons per year for cooling, projected to skyrocket to 3.7 billion gallons as planned facilities come online—enough to supply 34,000 homes.

• Fayetteville, Georgia: A Quality Technology Services data center campus quietly consumed 30 million gallons from the local water supply before receiving a utility bill.

• Tremonton, Utah: Protesters gathered ahead of a county commission meeting to oppose a large data center, with USU research showing 68% of Utahns oppose data centers within 3 miles of their homes, driven primarily by water concerns.

The concept of "digital redlining" describes the pattern where tech companies place their most resource-intensive facilities in Black and Brown neighborhoods and rural communities with the least political power to resist, while building cleaner, more efficient centers in predominantly white areas.

Community Backlash and Political Consequences

Resistance has escalated from local grievances to a national movement. Protest signs reading "Water for people not AI" and "Don't mess with our water" have appeared from Texas to New Mexico to Arizona. Key developments:

• $98 billion in data center investments have been blocked as communities organize to demand stricter regulations, according to Data Center Watch.

• In Tucson, Arizona, the City Council unanimously rejected the "Project Blue" data center complex in August 2025, citing the region's 7–10 inches of annual rainfall.

• In Saline Township, Michigan, residents oppose a 2.2-million-square-foot Oracle/OpenAI data center that would become the largest single water user in the Great Lakes region.

• In Chile, public protests led to the temporary revocation of Google's authorization for a planned $200 million facility.

• A CNN report (published today, October 12) documents how the backlash has turned violent in some cases, with death threats and an alleged plot to attack data centers in California.

• Gallup found that excessive water usage is one of the top reasons Americans oppose local data center construction, with about two-thirds of Americans opposing it.

Regulatory and Industry Responses

States are beginning to act. Virginia and Minnesota have drafted legislation requiring strict water-use reporting from data centers. Loudoun Water in Northern Virginia reported that data center potable water use grew by more than 250% between 2019 and 2023. The International Energy Agency estimates an average 100-megawatt data center consumes about 2 million liters of water per day.

On the industry side, newer facilities are shifting toward closed-loop liquid cooling and hybrid air/water systems that can sharply reduce fresh-water use. Some developers, like Fermi America in Texas, are investigating recycled water. However, critics note that companies have little incentive to adopt alternatives when water is cheap, unregulated, and invisible on balance sheets—and that many facilities operate under non-disclosure agreements that hide water usage data from communities.

The Counter-Argument

It's worth noting that some analysts push back. A The Atlantic piece argued the "data-center panic is overblown," noting that national-level water use by data centers remains under 1% of total U.S. consumption, and that utility billing structures typically separate residential from commercial rates, meaning data centers mainly threaten to raise prices for other businesses rather than households. The Genetic Literacy Project acknowledged that while the problem is "genuine and growing," the per-query water figures often cited in media are based on approximations, and the industry's overall water footprint, while growing, is still small relative to agriculture, thermoelectric generation, and other major uses.

The most widely accepted framing, however, is that AI water stress is a community-scale utility burden problem, not a national volumetric problem—the aggregate may be small, but the local impact on specific towns and watersheds can be severe, especially where data center burden spans 0.2% to 134% of host utility capacity across ten U.S. sites studied in a 2026 research review.