Across the Southwest, residents are taking to the streets with signs that read “Protect our water” and “Water for people, not AI.” In Texas, New Mexico and Arizona, locals are voicing concern over the rapid build‑out of artificial‑intelligence data centers and the water they consume for cooling.
How data centers use water
Most modern data centers rely on water‑based cooling systems to pull heat away from processors that can reach internal temperatures of up to 176 °F. The water absorbs heat and then evaporates, much like sweat cools a human body. While energy use in data centers is well documented, water consumption is less straightforward.
According to a 2023 estimate cited by Knowable Magazine, U.S. data‑center cooling systems used about 66 billion liters of water—less than 1 % of the nation’s total water use. That sounds modest, but the figure could rise sharply as AI‑focused facilities multiply.
Projected growth and regional stress
Researchers project that by 2030 U.S. data centers could consume between 731 billion and 1.125 trillion liters of water annually—roughly the amount New York City drinks in a year. Those numbers include water used not only for direct cooling but also for electricity generation, much of which still relies on coal or natural‑gas plants that need water to condense steam.
In drought‑prone states such as New Mexico and Arizona, even a small increase in demand can exacerbate water stress. Texas, while generally more water‑rich, still faces local shortages during severe heat waves.
Technological fixes and renewable power
Engineers say the situation is not hopeless. Cornell University’s energy‑systems expert Fengqi You notes that newer cooling designs and renewable‑energy sourcing can dramatically lower water footprints. He points to emerging “net‑zero water” concepts that combine efficient cooling with water‑replenishment projects.
Recent advances include:
- Liquid cooling: Pipes filled with water (or specialized fluids) run directly over server racks, creating a closed‑loop system that recirculates the same water without loss.
- Improved evaporative methods: When ambient temperatures are extreme, some facilities mist the air around pipes, but this is used only during peak heat.
- Immersion cooling: Hardware is submerged in a non‑conductive liquid, eliminating the need for external water‑based cooling.
University of Texas at Austin engineer Vaibhav Bahadur estimates that Texas data centers currently use less than 1 % of the state’s water demand, but could rise to 3‑9 % by 2040 if unchecked. He expects efficiency gains to keep the actual figure lower.
Location matters
Strategic siting is another key lever. You recommends locating new AI facilities in states with abundant renewable energy and lower drought risk—such as Montana, Nebraska, parts of Texas and South Dakota—rather than in already stressed regions like Arizona, New Mexico and Southern California. He says proper siting could cut future water use by up to 86 %.
Industry commitments
Major cloud providers are already pledging water‑positive goals. Amazon Web Services says it aims to return more water to communities than its operations consume by 2030 and is already more than halfway there. Google’s Gemini chatbot reportedly uses only five drops of water per median‑length query, and OpenAI has published a roadmap for minimizing water use in its data centers.
While the numbers show that AI data centers consume a fraction of national water use, the rapid expansion of the sector means the issue cannot be ignored—especially in regions already battling drought. Continued innovation, renewable‑energy adoption, and thoughtful site selection will be essential to keep AI’s thirst from drying up local water supplies.
Original reporting: KTBS 3 (Shreveport) — read the source article.
