Perspective · Central Texas Water · 2025 data

AI is thirsty.
Your lawn is thirstier.

Headlines say every AI email "drinks a bottle of water," and Texas data centers are now measured in billions of gallons. Here's the honest 2025 math on where the water really goes — and which part you can actually control.

0.26–0.32 mL
Water to cool the servers behind one AI prompt — a few drops (Google & OpenAI, 2025).
~264 billion gal
Water U.S. AI data centers consumed in 2025 — hundreds of millions of gallons a day.
~1.6 trillion gal
Water wasted on U.S. lawns every year to overspray, runoff, and overwatering.

The number that went viral

Researchers at UC Riverside put it in a form everyone remembers: a single 100-word AI email can evaporate roughly a 500 mL bottle of water — about 519 mL — once you count the power plants behind the servers.1 The image stuck. But that's a full-lifecycle figure.

Measured at the data center itself, the number is far smaller. In 2025, Google reported a median Gemini text prompt uses about 0.26 mL — five drops.2 OpenAI's Sam Altman put a typical ChatGPT query near 0.32 mL, "roughly one fifteenth of a teaspoon."3 Heavier reasoning models climb into the tens of milliliters. Both numbers are real — they just measure different boundaries.

Where the water goes — 2025

United States — billion gallons per year
U.S. landscape irrigation — total~3,285
U.S. landscape irrigation — wasted~1,640
All U.S. AI data centers~264
Google — entire company~11
Landscape: ~9 billion gal/day, ~half wasted (EPA WaterSense; CEOWorld/Irrigreen, Nov 2025).7 AI data centers: ~1 trillion liters in 2025 (Mordor Intelligence / SpaceDaily).5 Google: 2026 environmental report, 2025 figure.2
Americans waste about six times more water on their lawns than every AI data center in the country consumes — cooling, power, and all.

Set against a single hyperscaler, it isn't close: U.S. lawn waste is more than 100× Google's entire 2025 water footprint.

In Texas, this stops being abstract

A 2025 study from the Houston Advanced Research Center and the University of Houston projects Texas data centers will use 49 billion gallons of water in 2025, climbing to as much as 399 billion gallons by 2030.4 Globally, the IEA's 2025 Energy and AI report puts data-center water near 560 billion liters and rising toward 1,200 billion by 2030.6 Meanwhile nearly two-thirds of the U.S. sat in drought through 2025, and UC Riverside's Shaolei Ren notes data centers evaporate about 80% of the water they draw — it doesn't come back.1

That's a real, local strain, and it deserves scrutiny. But hold the scale in view: every AI data center in the entire country used about 264 billion gallons in 20255 — while U.S. households waste roughly 1.6 trillion gallons a year on lawns.7 The thing you can actually change is far bigger than the thing making headlines.

It's not "a bottle per email." It's a river per yard.

Here's the comparison that matters for a homeowner. A single poorly tuned sprinkler system — wrong pressure, misting heads, a stuck valve, watering the driveway — wastes up to 25,000 gallons a year, according to the EPA.8

At roughly 0.32 mL of cooling water per AI prompt, those 25,000 wasted gallons equal the water behind about 300 million AI prompts. One misadjusted system. One yard. One year.

Nationwide it compounds fast: Americans pour about 9 billion gallons a day onto landscapes, and roughly half never reaches a plant — an estimated $200 billion a year in wasted water once treatment and delivery are counted.7

You can't turn down a data center. You can fix your sprinklers.

The water anxiety aimed at AI is, for most Central Texas households, aimed at the wrong target — the biggest lever you actually hold is the system in your own yard.

Stop paying for water your lawn never uses

We measure your system zone by zone, fix the waste, program it to your legal watering day, and file the rebate paperwork. Most Austin homeowners cut outdoor water use 30–50%.

Sources

  1. Shaolei Ren / UC Riverside, reported by Bloomberg (2025) — ~519 mL per 100-word prompt (full lifecycle); data centers evaporate ~80% of the water they draw. bloomberg.com
  2. Google environmental reporting (2025) — median Gemini text prompt ≈ 0.26 mL water; Google used 10.9 billion gallons company-wide in 2025. blog.google
  3. OpenAI (Sam Altman), "The Gentle Singularity" (2025) — ~0.32 mL per query. blog.samaltman.com
  4. Houston Advanced Research Center & University of Houston (2025) — Texas data centers 49 billion gal (2025) → up to 399 billion gal (2030), via Lincoln Institute of Land Policy. lincolninst.edu
  5. Mordor Intelligence / SpaceDaily analysis (2025) — U.S. AI data centers consumed ~1 trillion liters (~264 billion gallons) in 2025. via Barchart
  6. International Energy Agency, "Energy and AI" (2025) — global data-center water ~560 billion L (rising toward ~1,200 billion L by 2030); ~2 million L/day for a 100-MW facility. iea.org
  7. CEOWorld / Irrigreen analysis (Nov 2025) with U.S. EPA WaterSense — ~9 billion gal/day U.S. outdoor irrigation, ~4.5 billion gal/day wasted (~1.6 trillion/yr), ~$200 billion/yr economic toll. ceoworld.biz
  8. U.S. EPA WaterSense (2026) — up to 50% of outdoor water is wasted; a poorly maintained system can waste up to 25,000 gallons/yr; landscape irrigation is ~1/3 of residential use. epa.gov/watersense

All figures are from 2025–2026 sources. AI and data-center water estimates vary widely by methodology and scope (on-site cooling vs. full lifecycle; withdrawal vs. consumption) and are evolving quickly; per-prompt figures reflect on-site cooling, and the ~300 million-prompt comparison uses that same on-site basis against measured landscape waste. Provided for general education by Greyroot Dynamics.