The number Amazon hadn't shared before

Amazon's data centers used 2.5 billion gallons of water last year. That figure, which the company disclosed reportedly for the first time, is striking not just for its scale but for how long it went unreported. For an industry that publishes detailed sustainability pledges, the basic resource accounting has often been missing.

The disclosure came shortly after Seattle — where Amazon is headquartered — enacted a one-year moratorium on new data center construction. Some of Amazon's own employees had pushed for that policy, according to reporting by The Verge.

Why data centers use so much water

The mechanism is straightforward: large data centers generate enormous amounts of heat from densely packed servers. The dominant cooling method is evaporative cooling, in which water absorbs heat and is released as vapor. The more compute a facility runs, the more water it moves through that cycle. As AI model training and inference workloads have grown, so has the thermal load — and with it, water demand.

Water consumption is distinct from water withdrawal. Withdrawal measures how much water a facility pulls from a source; consumption measures how much doesn't return. Evaporative cooling is consumptive by design, which is why it draws scrutiny in drought-prone or water-stressed regions.

Disclosure in context

Amazon's 2.5 billion gallon figure is a global aggregate. Without a breakdown by facility or region, it's difficult to assess where the stress falls hardest. A data center drawing from an aquifer in the American Southwest carries different implications than one operating in a water-abundant region of northern Europe. Amazon has not, as of this writing, published that granular data.

The timing of the disclosure is worth noting. Companies sometimes release unflattering data when external pressure makes continued silence more costly than transparency. Seattle's moratorium, backed in part by Amazon's own workforce, created exactly that kind of pressure. Whether this disclosure represents a shift toward systematic reporting or a one-time response to a specific moment is not yet clear.

The broader pattern

Amazon is not alone in facing these questions. Microsoft and Google have both reported water usage figures in recent years, and both have faced criticism for rising consumption as they expand AI infrastructure. The difference is that Amazon's disclosure appears to be a first, which means there's no prior year to compare it against — a limitation that makes trend analysis impossible for now.

For communities near data center campuses, the abstract becomes concrete quickly. Water used for cooling is water not available for agriculture, municipal supply, or ecosystem flow. As data center construction accelerates to meet AI demand, local governments and regulators are increasingly asking for the numbers that companies have been slow to provide.

Amazon's disclosure is a start. Whether it becomes a baseline for accountability or a one-time gesture will depend on what the company reports — and when — next year.