AMAZON RELEASES WATER USAGE DATA, CLAIMS EFFICIENCY

Amazon disclosed water usage for its global data center operations for the first time on Tuesday, revealing that its facilities consumed 2.5 billion gallons of water in 2025 at a rate of 0.12 liters per kilowatt-hour of electricity. The company said water consumption actually dropped by two percent year over year, and that it uses air cooling "about 90 percent of the time", resorting to evaporative cooling only "the hottest hours of the hottest days".

The disclosure arrives amid intensifying scrutiny of data center water use across the industry, driven largely by the computational demands of artificial intelligence. Google, Microsoft, and Meta have all faced questions about the environmental footprint of their AI infrastructure, and Amazon's move puts it ahead of competitors in terms of voluntary transparency, though the company's numbers are self-reported.

THE EFFICIENCY CLAIM

Amazon says its data centers are seven times more water-efficient than the industry average, a figure the company is using to position itself ahead of Microsoft, Google, and Meta. Among those three, Amazon claims Google uses the most water per kilowatt-hour, though the comparison is complicated by differences in how each company measures and reports.

The 0.12 liters-per-kWh figure refers only to direct water use at Amazon's facilities. It does not account for indirect water consumption at the power plants that generate electricity for Amazon's data centers, a metric known as scope two water use that environmental advocates have increasingly pushed companies to disclose.

WHATS MISSING

The gap in the data is notable. When a data center draws a kilowatt-hour from the grid, that electricity was generated somewhere, and power generation is often water-intensive, especially at natural gas and nuclear plants. Without accounting for that upstream use, the 2.5 billion gallon figure represents only part of Amazon's true water footprint.

Amazon did not disclose which regions or facilities consume the most water, leaving the geographic profile of its data center operations opaque. This matters because water stress varies dramatically by location. A data center in a water-scarce region like the southwestern United States draws on a fundamentally different resource than one in a wet climate like Seattle.

THE AI DRIVER

The timing of the disclosure reflects a broader shift. AI model training and inference require enormous computational resources, and data center operators have been under pressure to quantify the physical infrastructure behind the technology. Amazon's Gemini AI data centers have been a particular focus as the company races to compete with OpenAI and Google's own AI products.

The two percent reduction in water use year over year occurred even as Amazon expanded its data center footprint, suggesting the company has achieved marginal efficiency gains. Whether those gains can keep pace with the accelerating demand for AI compute remains an open question.

WHAT COMES NEXT

Transparency advocates are likely to push Amazon to disclose scope two water use and break out consumption by region. The company's self-reported efficiency claims will face scrutiny from independent researchers, and competitors may respond with their own disclosures. For now, Amazon has opened a window, however incomplete, into a part of its infrastructure that has long been hidden from public view.