Amazon's data centers used 2.5 billion gallons of water last year 亚马逊数据中心去年用水量达25亿加仑

TL;DR

• Amazon disclosed global data center water use for the first time: 2.5 billion gallons in 2025.
• Water use intensity was 0.12 liters per kilowatt-hour, a 2% drop from 2024.
• Disclosure came just after a Seattle data center moratorium pushed by some Amazon employees.
• Amazon claims greater efficiency than some rivals, citing a comparative graphic.

Key Data

Deep Analysis

The timing of Amazon’s first-ever public disclosure on data center water consumption is not accidental; it’s a classic case of strategic preemption. Releasing these figures just after the Seattle moratorium—a local policy push partly fueled by internal employee activism—is a calculated move to reframe the narrative from "problem" to "transparent leader." It’s corporate PR jiu-jitsu: using the momentum of a negative event to demonstrate control and progress. The 2.5 billion gallons figure is staggering in absolute terms, but Amazon immediately pivots to the efficiency metric: 0.12 liters per kilowatt-hour. This is the number they want etched into the public consciousness. By framing the discussion around intensity (water per unit of compute) rather than total volume, they are subtly arguing that their growth is responsible growth.

This reveals a core tension in the sustainability debate for hyperscalers. The efficiency narrative is technically valid—a more efficient data center uses less water per unit of power—but it can obscure the monumental absolute resource demands of AI scaling. A 2% improvement in efficiency, while positive, is dwarfed by the likely double-digit percentage growth in compute capacity to train and serve ever-larger AI models. It’s a classic Jevons paradox in the making: as efficiency improves, total consumption can still soar because the utility of the resource increases, driving more overall use. Amazon’s claim of being more efficient than rivals (hinting at Microsoft or Google in the graphic) also signals the dawn of a new competitive front. The "green" arms race in Big Tech is no longer just about carbon credits or renewable energy PPAs; it’s now about quantifying and optimizing every unit of water, land, and power. Efficiency becomes a key differentiator for attracting ESG-minded investors and appeasing regulators.

The employee activism that pushed for the Seattle moratorium is equally significant. It shows that internal stakeholder pressure is now a material force shaping corporate policy and transparency. Amazon’s disclosure is, in part, an attempt to defuse that internal pressure by providing data and demonstrating proactive management. Looking forward, the data center industry is moving into a phase where its social license to operate will be scrutinized by kilowatt-hour and gallon. The battle will be fought over water-use effectiveness (WUE) metrics, the adoption of closed-loop cooling and water recycling, and the siting of new facilities in water-stressed versus water-abundant regions. Amazon’s report is a salvo in this new war—one where the most compelling case will be built not just on total output, but on the granular, verifiable efficiency of the infrastructure powering the AI revolution.

Industry Insights

1. Efficiency Metrics Will Become the New Public Battleground: Absolute consumption totals will be challenged by intensity-based metrics (like water/kWh), forcing regulators and the public to understand nuanced trade-offs.
2. The "Water Footprint" Will Join the Carbon Footprint: Comprehensive sustainability reporting for cloud providers will inevitably expand to include detailed, audited water usage and impact metrics as standard practice.
3. Geopolitical Siting of Data Centers Will Intensify: Water availability and local regulation will become primary factors in data center location strategy, moving beyond just low energy costs and fiber connectivity.

FAQ

Q: Is 2.5 billion gallons a lot of water for data centers?
A: Yes, it’s an enormous amount, roughly equivalent to the annual water consumption of over 20,000 average U.S. households. This highlights the significant physical resource demands of cloud computing and AI.

Q: How can Amazon use water more efficiently than its rivals?
A: Efficiency (liters per kWh) depends on cooling technology design, climate, and operational optimization. Amazon is claiming superiority in its infrastructure design, likely referring to metrics like Water Usage Effectiveness (WUE).

Q: Does this water get consumed and disappear, or is it recycled?
A: Most data center water is used in evaporative cooling systems. While some is lost to evaporation, a significant portion can be treated and recirculated. The 2.5 billion gallons likely represents total consumption, not just initial intake.

TL;DR

• 亚马逊首次公开披露其全球数据中心在2025年的总耗水量为25亿加仑。
• 其数据中心用水效率为每千瓦时电力耗水0.12升，较2024年下降2%，同时运营规模仍在扩张。
• 亚马逊在报告中声称其用水效率优于部分科技巨头竞争对手。
• 此次披露发生在美国西雅图市对亚马逊等公司实施数据中心建设禁令之后，时机敏感。

核心数据

深度解读

亚马逊这次“首次披露”，时机选得堪称经典。在西雅图——其总部所在地——刚刚因其员工推动而通过了一项针对数据中心的建设禁令后，这份带着具体数字和效率提升报告的出炉，很难不被视为一场精心策划的舆论反击与形象修复。它传递的核心信息很明确：我们不仅在用水，而且比你们想象中（或比同行）更会用水。

报告的核心叙事是“效率提升”：用水量单位产出下降了2%。这是一个聪明的对比角度。在AI算力需求爆炸式增长、数据中心规模不断扩大的背景下，亚马逊确实做到了让每一瓦特电力对应的水耗减少了。这符合企业ESG叙事的所有要素——用“相对值”的改善，来对冲“绝对值”增长的道德压力。然而，绝对耗水量25亿加仑依然是一个惊人的数字，足以填满数万个奥运游泳池。这就像一个油耗更高的汽车制造商，告诉你他们每公里耗油降低了5%，但因为卖出的车更多，总耗油量其实又创新高。效率的“绿袍”之下，包裹的依然是资源消耗的“庞大躯体”。

更值得玩味的是那份隐晦的“同行比较”。报告图表中提到的“M…”，极有可能指向微软。微软因激进采购核电为AI供能而备受争议，但其在水耗问题上的透明度相对较低。亚马逊此举，无疑是在将公众视线引向“水耗效率”这个其自认的比较优势上，从而试图转移外界对“绝对水耗”和“能源结构碳排放”的注意力。这是一种高明的叙事转移：将讨论从“消耗多少”转向“谁消耗得更有效率”。

这起事件揭示了AI时代一个更深层的悖论：构建更“智能”的系统，其物理基础却是极其“原始”和“贪婪”的资源消耗。数据中心不仅是用电黑洞，也是用水巨兽。亚马逊的披露，与其说是主动的环保宣言，不如说是被动监管压力下的一次危机公关。它承认了问题的存在，但给出了一个“我们正在解决，且比对手做得好”的答案。真正的考验在于，这种被动透明度能否演变为行业常态？当西雅图、都柏林、新加坡等地因水电资源紧张而相继出台限制措施时，科技巨头们是会继续这种“挤牙膏”式的信息披露，还是会被迫转向更根本的技术或能源解决方案？这份报告，是这场全球资源博弈中，亚马逊投出的一枚带着计算过的涟漪的石子。

行业启示

1. 透明度将成为被迫的基准：在社区和监管机构的双重压力下，对关键环境数据的被动披露可能转变为行业“新常态”，倒逼所有大型基建运营商（不仅是科技公司）进行数据公开。
2. “效率提升”叙事需结合绝对值审视：行业宣传和ESG报告应警惕仅强调单位效率提升，而忽略因规模扩张导致的资源绝对消耗增长，这关乎真正的可持续性。
3. 能源与水资源安全将重塑科技地理：未来数据中心的选址将更深度绑定区域水电资源的可持续性与政策稳定性，单纯的能源成本优势可能让位于综合资源治理能力。

FAQ

Q: 亚马逊为什么选择在这个时间点披露数据中心用水数据？
A: 这被认为是对西雅图市在员工推动下实施数据中心建设禁令的直接回应，旨在通过主动披露环境数据和强调效率提升，来缓解公众和监管机构对其资源消耗的担忧，是一种危机公关和形象修复策略。

Q: 用水效率提升2%是否意味着亚马逊的数据中心对环境更友好？
A: 需辩证看待。单位用电的耗水量减少是技术进步的体现，但报告同期指出其运营规模仍在扩张，因此总耗水量（25亿加仑）很可能并未减少，甚至可能增加。“更友好”的结论应基于总消耗量是否下降来判断。

Q: 其他科技巨头的数据中心用水情况如何？
A: 亚马逊在报告中暗示其效率优于部分竞争对手，但未直接点名。实际上，谷歌、微软等公司也披露过环境数据，但统计口径、披露范围（是否包括冷却系统水耗）和时间点各不相同，横向直接比较需要谨慎，且各方都在通过技术创新来改善指标。