Every time you type a prompt into ChatGPT, servers thousands of miles away heat up, cooling systems kick in, and real water drawn from local supplies evaporates into the air. The scale of this hidden cost is only now becoming clear, with major tech companies consuming billions of gallons annually to keep AI infrastructure running.

How Much Water Do AI Data Centers Actually Use?

A typical 100-megawatt AI data center consumes between 1.5 to 3.0 million cubic meters of water per year for evaporative cooling, the dominant method used to prevent servers from overheating. To put that in perspective, the average American uses about 80 to 100 gallons per day. A single 100-megawatt facility can match the residential water use of an entire city of 50,000 people.

The numbers are accelerating rapidly. Direct US data center water use currently sits around 17 billion gallons annually. According to the Lawrence Berkeley National Laboratory's 2024 federal report, that figure will reach 38 to 73 billion gallons by 2028, driven primarily by AI training. But the direct water use tells only part of the story. The same report estimated that an additional 211 billion gallons were consumed indirectly through the electricity required to power those same data centers in 2023, making the indirect figure approximately 12 times larger than the direct one.

By 2030, AI data centers alone could drain between 731 and 1,125 million cubic meters of water annually, equivalent to the household water usage of 6 to 10 million Americans.

Why Is AI More Thirsty Than Traditional Computing?

The difference comes down to power density and intensity. A standard Google Search query uses roughly 0.3 watt-hours of energy. Per-query energy estimates for ChatGPT-class queries range from approximately 0.3 to 3 watt-hours, roughly 3 to 10 times more than a Google search, depending on model size and serving infrastructure. More energy means more heat, which requires more cooling, which demands more water.

The power densities for advanced AI racks are scaling dramatically. In 2023, the average was between 25 kilowatts and 40 kilowatts per rack. By 2025, that figure is expected to exceed 200 kilowatts, with densities of 1 megawatt per rack projected in the near future. This structural shift in thermal load per building is fundamentally different from what came before.

What Are the Real Numbers Behind Individual AI Queries?

The water footprint of a single AI query depends entirely on what you measure. The most widely cited figure comes from a research paper by Pengfei Li and colleagues at UC Riverside and UT Arlington, titled "Making AI Less Thirsty." The figure for generating a 100-word email with ChatGPT is approximately 519 milliliters of water, close to a standard bottle.

However, OpenAI CEO Sam Altman pushed back publicly, stating the average query uses just 0.3 milliliters, about one-fifteenth of a teaspoon. His figure accounts only for direct operational water at the data center level, not the water used in electricity generation or chip manufacturing. The gap between 0.3 milliliters and 519 milliliters is not a measurement error; it is a scope decision. The 519-milliliter estimate includes indirect water use, the water consumed by power plants generating electricity for the data center. The 0.3-milliliter figure only counts water physically flowing through pipes at the facility itself.

In 2025, Google disclosed that each query to its Gemini assistant consumes 0.26 milliliters of water on-site, a direct facility-level figure similar to Altman's claim for ChatGPT. For most users, the practical takeaway is that a single query is negligible. However, ChatGPT has over 100 million active users. Multiply any per-query number by billions of daily queries, and the real story emerges.

Which Companies Are Using the Most Water?

Microsoft and Google lead in absolute water consumption. A training run for GPT-4 in West Des Moines, Iowa, consumed 11.5 million gallons of water in July 2022 and 13.4 million gallons in August 2022. Across its five facilities in that city, Microsoft used 68.5 million gallons of water in 2024, making it the region's largest user.

Google's data center water consumption grew from 4.3 billion gallons in 2021 to 7.7 billion gallons in 2024, an increase of nearly 80 percent in three years. Microsoft consumed approximately 1.69 billion gallons in fiscal year 2024, up from 1.26 billion gallons in fiscal year 2022, a rise of around 34 percent, even as its per-unit efficiency improved.

How Are Regulators Responding to Water and Power Demands?

New York has become the first state to impose a targeted moratorium on large data center expansion. Lawmakers including Senator Kristen Gonzalez and Assemblymember Didi Barrett shepherded legislation through both chambers that imposes a one-year moratorium on permits for any data center drawing 20 megawatts or more. The measure mandates a statewide environmental review evaluating energy use, water demand, emissions, and electronic waste.

The regulatory pressure reflects unprecedented grid strain. The New York Independent System Operator reports an unprecedented queue of large loads, with about 11.9 gigawatts of potential demand waiting for interconnection studies. Approximately 8.3 gigawatts joined the list during 2025 alone, figures that dwarf earlier projections and complicate data center regulation efforts. This queue size is equivalent to 10 average nuclear plants.

Global data center electricity consumption reached 415 terawatt-hours in 2024, with projections showing it could reach 945 terawatt-hours by 2030. Utilities must now decide who pays for new substations and transmission lines, while water agencies must evaluate cooling demands as climate stresses intensify.

Steps to Understanding Data Center Water Impact

• Direct vs. Indirect Water Use: Direct water use occurs at the facility itself through cooling systems, while indirect water use happens at power plants generating electricity. Both matter, but indirect use is typically 12 times larger than direct use.
• Per-Query Measurement Scope: When evaluating AI water footprints, clarify whether figures include only facility-level cooling water or also account for power generation. A single query may use 0.3 milliliters directly but 519 milliliters when including electricity production.
• Regional Grid Constraints: Water availability, electricity generation capacity, and transmission infrastructure vary by region. Some areas face water scarcity as the gating factor for new data center approvals, while others face power grid limitations.
• Cumulative Impact Tracking: Individual queries appear negligible, but multiplying per-query consumption by billions of daily queries reveals the true scale. Monitor aggregate usage across all users and all facilities to understand real environmental impact.

The debate over data center expansion reflects competing economic narratives. Industry groups such as the Data Center Coalition argue that regulatory pauses signal the state is closed for business. However, advocacy networks such as NY Renews celebrate the moratorium, calling it a victory for communities facing rising energy use and land pressures. Governor Hochul remains undecided, preferring to pursue parallel infrastructure policy measures through the Public Service Commission.

Other jurisdictions are watching closely. States across the Midwest and West are tracking New York's experiment. Legislators in Oregon, Virginia, and Georgia have floated bills echoing the New York moratorium template. Data center investors now assess policy risk alongside land and power prices.

The underlying tension is clear: the AI infrastructure boom is real and accelerating, but the physical constraints of water, electricity, and grid capacity are becoming the actual bottleneck. As companies like Microsoft and Google continue to scale their AI operations, the water crisis will likely intensify unless cooling technologies improve or demand patterns shift.