The Cloud's Water Bill Is 12 Times Bigger Than Reported
Author: Protik Ganguly
Every AI query, every cloud backup, every streamed video runs through a data centre somewhere. A single large one, 100 megawatts, uses as much electricity as 100,000 households. The cloud runs on a scale of water and power that most people never see — and the most commonly cited number for it is roughly twelve times too small.
Start with electricity, since it explains the rest. US data centres consumed 176 terawatt-hours in 2023, about 4.4% of national electricity, projected to reach 12% by 2030 (Lawrence Berkeley National Laboratory, 2026). Virginia's data centres alone now account for almost 40% of the state's total consumption. Meta's Hyperion project in Louisiana will require at least 5 gigawatts on its own — three times what New Orleans uses (Consumer Reports, 2026).
The water figure usually quoted, 17 billion gallons a year for direct cooling, is the smallest part of the real number. That's the water a facility uses on-site. It excludes the water consumed by the power plants generating its electricity in the first place — and the same federal research that produces the 17-billion figure puts that indirect total at roughly 211 billion gallons a year (EESI, 2026; Lawrence Berkeley National Laboratory, 2026). Coal plants alone consume close to 19,185 gallons of water per megawatt-hour produced. Indirect consumption outweighs direct consumption by more than 12 to 1 — meaning the number that gets repeated in most coverage captures less than 8% of the actual footprint.
The concentration is local, not just national. Nationally, data centre water use is roughly 0.3% of public supply — small. But nearly half of US capacity clusters in five regions, several of them already managing drought. In The Dalles, Oregon, one company's water use roughly tripled over five years while the town's population grew 12%, and reports put its share of the town's total water consumption near 30%.
The industry's own counter-evidence is real and worth weighing fairly: Amazon has cut water use per unit of server capacity by 40% since 2021, and Google now replenishes most of what it withdraws. Efficiency is genuinely improving per query. The catch is the same one that shows up everywhere technology gets more efficient — total use keeps climbing anyway, because falling cost per use multiplies how often it gets used. The infrastructure is being built faster than the accounting for it.
References
Congressional Research Service. (2026, May 12). Data centers and their energy consumption: Frequently asked questions. https://www.congress.gov/crs-product/R48646
EESI. (2026). Data centers and water consumption. https://www.eesi.org/articles/view/data-centers-and-water-consumption
Lawrence Berkeley National Laboratory. (2024). 2024 United States data center energy usage report. https://doi.org/10.71468/P1WC7Q
MOST Policy Initiative. (2026, April 8). Data center water use. https://mostpolicyinitiative.org/science-note/data-center-water-use/
Related Articles
Inflation Is Running Hot. Here Is Your Strategy.
• High-yield savings accounts and money market funds offer 4 to 5% APY, which is comparable to or slightly below current inflation rates, making them a suitable option for liquid funds. These accounts are a better choice than standard bank accounts earning 0.1% APY. • Series I Savings Bonds issued by the US Treasury automatically adjust their rate every six months based on CPI, making them a direct inflation hedge for money that won't be needed for at least a year. • Historically, equities and real estate have outpaced inflation over long periods, with the S P 500 returning approximately 10% annually on average over the past century, or roughly 7% after inflation.
GDP Is Growing. But Does It Reach Ordinary People?
• Gross Domestic Product (GDP) measures the total monetary value of all final goods and services produced within a country, but it does not necessarily reflect the living standards of its citizens. The US has the largest GDP in the world, at approximately $30 trillion in 2026. • GDP per capita, which divides total GDP by population, is a rough proxy for average income and living standard, but it has limitations, such as being influenced by high earners and non-resident multinationals. US GDP per capita in 2026 is approximately $85,000. • The difference between GDP per capita and median household income can indicate income inequality, with the US being a clear outlier due to a significant divergence between GDP per capita growth and median income growth between 1979 and 2013.
The AI Digital Divide Is Widening. Most Coverage Isn't Tracking It.
• The AI digital divide is widening, with developing nations facing significant barriers to accessing AI technology due to limited internet connectivity and infrastructure. Mobile data in Sub-Saharan Africa costs fourteen times more than in Europe. • In the United States, 19 million Americans lack broadband internet, with a disproportionate concentration in rural states, while connectivity gaps persist across rural Europe. This limits access to AI tools requiring reliable internet and cloud access. • The global distribution of AI benefits is skewed, with only 3% of projected $19.9 trillion in benefits expected to go to Latin America and a mere 8% to the combined populations of Africa, Oceania, and other Asian markets excluding China.