After Amazon passed a one-year "emergency moratorium" on new data center construction in Seattle, USA, it systematically announced the annual water usage of its global data centers for the first time, triggering a new round of attention from the outside world on the consumption of infrastructure resources in the era of artificial intelligence. This data from the company's official sustainability report states that Amazon's global data centers will consume a total of approximately 2.5 billion gallons of water in 2025, equivalent to approximately 9.5 billion liters, and as the scale of data centers continues to expand, total water consumption will still drop by 2% compared to 2024.
According to data disclosed by Amazon, the water intensity of its data center in 2025 will be approximately 0.12 liters of water per kilowatt hour of electricity consumption. This is the first time that specific water use indicators based on electricity consumption have been given to the outside world. In the context of the demand for large-scale computing power driven by generative artificial intelligence and the increasing controversy over data center location and approval, water and energy consumption issues have become important variables for local governments and communities to evaluate data center projects.
The disclosure comes shortly after Seattle officially passed a one-year moratorium on data center construction, driven in part by Amazon's own employees, who called on the company to be more transparent about environmental impacts and reexamine the pace of data center expansion. In this newly released water resource report, Amazon attempts to respond to public opinion with quantitative data, emphasizing that the water efficiency of its data centers is relatively leading in the industry.
In the report, Amazon cited public data from other large technology companies such as Microsoft, Google and Meta, and created a comparison chart, saying that the "water consumption per kilowatt-hour" indicator of these companies in the past few years was generally higher than Amazon's. The report shows that in terms of water intensity per unit of energy consumption, Google has the highest water use level in the quoted range. However, the relevant data is mainly concentrated on its specific AI data centers used to support artificial intelligence services such as Gemini, while Amazon makes statistics based on the overall data center business. The two calibers are not completely consistent.
Despite its emphasis on relative efficiency advantages, the figures disclosed by Amazon still do not include the indirect water consumption behind its data center operations, such as the cooling water used in the power generation process of the power plants that power the data centers, and the water consumption involved in the construction of new data centers. Environmental groups and industry observers point out that this type of indirect water use is often difficult to quantify, but it may constitute a significant part of the overall water footprint of a data center, and therefore has become the focus of further information disclosure requirements from companies in the future.
In terms of specific cooling strategies, Amazon said that its data centers rely on air cooling about 90% of the time, and only activate the evaporative water cooling system "during the hottest hours of the year" to reduce water demand while maintaining equipment stability. The company also said its servers' tolerance range for high temperatures has been moderately increased, allowing data centers to operate at higher ambient temperatures and thus reduce reliance on intensive cooling and associated water consumption.
Amazon claims that based on the calculation method it uses, its data centers are about seven times more water efficient than the industry average. This statement is based in part on a peer-reviewed research paper published last year that measured global data center power and water usage patterns, which Amazon adjusted to industry averages to arrive at what it calls a "seven-fold efficiency" conclusion.
As iterations of generative AI models accelerate and computing power density and energy consumption continue to rise, large technology companies are accelerating the construction of new generation AI data centers around the world, and therefore face more intensive regulatory scrutiny and community resistance. From the United States to Europe, policy debates over the impact of data centers on grid loads, groundwater and urban water systems are escalating, and Seattle's one-year moratorium is seen as one of the latest examples of this trend.
Amazon’s latest water resource data disclosure responds to the outside world’s doubts about the transparency of its environmental information to a certain extent, and also provides a new reference for comparing the resource use of different technology giants. However, on the premise that issues such as indirect water use, long-term regional water security impacts, and the balance of interests with local communities have not yet been fully quantified and disclosed, the resource disputes surrounding AI data centers will still be difficult to eliminate in the short term. In the future, how to strike a balance between industrial development and water resources carrying capacity when approving new projects will continue to test the decision-making capabilities of regulators and enterprises.