# Amazon’s Data Centers Used 2.5 Billion Gallons of Water Last Year > Source: > Published: 2026-06-11 18:39:20+00:00 Transparency in tech operates like Instagram stories: you only see what companies want you to see. [Amazon](https://www.aboutamazon.com/news/sustainability/amazon-data-center-water-usage) just dropped its first-ever absolute water usage number for [AWS data centers](https://aws.amazon.com/sustainability/data-centers/)—**2.5 billion gallons** annually—while claiming superior efficiency compared to Microsoft, Google, and Meta. The timing feels less like environmental leadership and more like damage control as cities impose [data center](https://www.gadgetreview.com/the-real-questions-you-should-be-asking-about-ai-data-centers-in-your-neighborhood) moratoria. ## The Efficiency Olympics Get Messy *Amazon’s water metrics look impressive until you examine what’s being measured.* [Amazon reports](https://www.aboutamazon.com/news/sustainability/amazon-data-centers-electricity-bills-water-use) using just **0.12-0.15 liters** per kilowatt-hour compared to an industry average of **1.8 L/kWh**. That’s genuinely impressive—if you’re comparing apples to apples. The problem? Amazon is showcasing fleet-wide efficiency across all AWS facilities while comparing against [Google’s AI](https://www.gadgetreview.com/subsidizing-the-trillionaires-why-your-utility-bill-is-surging-to-pay-for-googles-ai-data-centers)-specific Gemini data centers, which run hotter and demand more cooling. This selective benchmarking resembles comparing your average daily screen time to someone’s weekend Netflix binge. [Google’s Iowa facility](https://sustainability.google/google-2025-environmental-report/) alone consumed **1 billion gallons** in 2024—equivalent to five days of all residential water use statewide. Meanwhile, Amazon operates “about 90 percent of the time” on air cooling, using evaporative water systems only during peak heat. The company’s global Power Usage Effectiveness (PUE) of **1.15** in 2024 beats the public cloud average of **1.25**, demonstrating energy efficiency that indirectly reduces upstream water consumption at power plants. ## The 60 Percent Problem *Amazon’s numbers exclude the biggest slice of data center water impact.* Here’s what [Amazon](https://www.gadgetreview.com/amazon-kitchen-gadgets-you-didnt-know-you-were-missing) isn’t counting: indirect water consumption at power plants, which the International Energy Agency estimates represents **60 percent** of total data center water impact. While Amazon optimizes its direct cooling systems, the electricity powering those servers requires massive water volumes at generation facilities. Meta’s 2024 indirect water consumption alone hit **72.2 billion liters** for **18.4 TWh** of electricity—roughly **3.92 L/kWh**, orders of magnitude above Amazon’s direct measurements. Leaked internal Amazon documents from 2022 projected **7.7 billion gallons** annually by 2030, suggesting the company has been tracking much larger numbers privately while emphasizing efficiency publicly. Amazon’s methodology document explicitly focuses on “on-site withdrawals” rather than upstream water embedded in electricity generation. This approach aligns with what company sustainability leads call focusing on their “direct water footprint.” ## When Communities Say No *Local resistance forces tech giants toward selective transparency.* Seattle’s data center moratorium reflects growing community pushback against AI infrastructure expansion. As individual facilities consume billions of gallons annually, water-stressed regions are questioning whether [AI training](https://www.gadgetreview.com/eufy-paid-users-to-fake-thefts-for-ai-training) and cloud services justify such resource allocation. Amazon’s response includes expanding recycled water use to over **120 U.S. facilities**, preserving **530 million gallons** of potable water annually. The company claims to be **53 percent** toward “water positive” by 2030, planning to return more water to communities than it consumes directly through watershed restoration and infrastructure projects. The IEA warns global data center water consumption could double to [1.2 trillion liters](https://www.reuters.com/business/energy/ai-double-data-centre-power-water-consumption-by-2030-un-researchers-say-2026-06-03/) by 2030, driven largely by AI workloads. A typical **100-MW facility** consumes around **2 million liters** daily, making water availability a key constraint for future AI campus development. Yet the fundamental question remains: should water efficiency be measured only at the data center fence line? As AI workloads intensify and communities face drought, expect pressure for standardized reporting that includes the full water footprint—not just the photogenic portions companies prefer to highlight.