The Hidden Cost of AI: How Data Centers Are Draining Water Resources and What It Means for Investors

How many prompts have you fired off to ChatGPT or Midjourney this week—10, 20, hundreds?

You may not realize it, but each volley of text may have quietly used up a significant supply of fresh water from a data center. Multiply that by billions of daily queries, along with training runs that guzzle upward of 185K gallons, and the link between AI’s expansion and water scarcity problems could create significant problems for these companies and the communities where their data centers are located.

Key Takeaways

• Training a single large-language model such as ChatGPT can consume hundreds of thousands of liters of fresh water.
• Data-center electricity demand is expected to surge 16% by 2030, amplifying water-cooling needs.

Water: AI’s Silent Thirst

AI chips run hot. Most commercial-scale facilities rely on evaporative cooling towers that “drink” clean water, then vent it as steam. Researchers estimate ChatGPT’s training alone vaporizes about 185K gallons and accounts for about 6% of the local utility’s entire supply during peak months, while a typical user session (10 to 50 prompts) uses about half a liter.

With Goldman Sachs (GS) forecasting a 165% jump in data-center power capacity by 2030, the vicious cycle among AI’s energy demands, heat generation, and water needs is expected to intensify.

Why It’s an Environmental Concern

Fresh, clean water is already one of the earth’s most precious resources, and about a fifth of data centers are located in water-stressed regions, where they compete with drinking supplies and agriculture. In Phoenix, Arizona, for instance, data centers’ daily cooling demand can top 170 million gallons, exacerbating ongoing regional water shortages.

Heavy water use lowers aquifers, while discharging warmer effluent can alter river temperatures and degrade ecosystems. Climate change compounds the threat: hotter summers raise cooling loads just as droughts shrink reserves.

Note

Is the answer to AI data center water usage to be found in pig poop ponds? The companies behind high-tech systems for filtering various contaminants, including pig sewage near massive pork farms, are pitching AI data center firms on repurposing waste or low-quality water to reduce their reliance on fresh groundwater.

How AI’s Water Use Stacks Up

Global AI demand is estimated to consume 1.1 trillion to 1.7 trillion gallons of freshwater annually by 2027. That rivals the annual household water use of the entire state of California and is rising faster than any single sector outside agriculture.

For comparison, semiconductor fabrication plants, which are notoriously thirsty, might use up to 10 million gallons a day, equal to the needs of a midsize U.S. city. Hyperscale data centers are catching up fast: some now top 5 million gallons daily, rivaling towns of 50,000 residents.

Agriculture still dominates global water use, accounting for about 70% of annual groundwater use worldwide, yet in drought-prone, high-income regions, the marginal gallon from AI directly competes with farms, households, and legacy manufacturers, heightening the odds of usage caps or perhaps taxes or even charges.

In addition to water, electricity demands from the AI sector may more than double this decade, forcing utilities to restart shuttered plants or import pricier renewables—costs that eventually flow through to customers.

What Can Be Done Before the Well Runs Dry?

Water-intensive AI firms face scrutiny from regulators and environmentally conscious shareholders. However, venture and infrastructure capital are flooding into projects for efficient immersion cooling, membrane recycling, and leak-detection platforms for data centers. Those wishing to invest in such projects can look to established cooling-tower manufacturers or water-themed ETFs like Invesco’s Water Resources ETF (PHO) or First Trust’s Water ETF (FIW).

When considering AI companies, due diligence should weigh specific metrics, including a company’s water-use efficiency, the hydrological risk of its data-center footprint, and progress toward “water-positive” pledges, right alongside the usual AI growth metrics.

The Bottom Line

The race to dominate generative AI is becoming inseparable from a mounting water bill. If unchecked, the clash between AI and water could dent margins, invite regulatory and stakeholder backlash, reshape site-selection considerations, and damage fragile water ecosystems worldwide.

Investors who look beyond headline revenue to the hidden hydrological balance sheet—and back companies that curb, recycle, and monetize every drop—will be better positioned when this form of “liquidity scarcity” shifts from headline warnings to cash-flow reality.