Data centers supporting artificial intelligence are under scrutiny in the US for their heavy water and power consumption. Nvidia, Microsoft, Amazon, and Meta have introduced new cooling technologies that reduce water use, aiming to respond to environmental concerns amid escalating demand for AI services.
- Data centers worldwide used 222 billion liters of water in 2025 for cooling.
- Nvidia’s DSX system nearly eliminates water use by employing warmer closed-loop cooling.
- Tech firms report improved water use efficiency but face public scrutiny and cost barriers.
What happened
US-based tech giants such as Nvidia, Microsoft, Amazon Web Services, and Meta have acknowledged the significant water consumption of data centers that power AI applications. In response, some have adopted advanced cooling technologies, notably closed-loop systems that recirculate cooling liquid directly through servers to minimize water waste. Nvidia's recent DSX system reportedly cuts water use almost completely in certain facilities by running cooling liquid at elevated temperatures compared to prior solutions.
Despite these technological strides, total water use in data centers remains substantial and is projected to triple globally by 2030 without intervention. Companies have also reported improvements in water efficiency at existing sites, with Microsoft and AWS noting respective gains of 25 percent and 37 percent in recent years. However, reporting inconsistencies and omitted disclosures by companies like SpaceX highlight challenges in fully assessing industry-wide environmental impacts.
Why it matters
Data center water consumption is drawing growing public anger in the US due to its scale and environmental impacts, particularly given the large volumes of water required for cooling high-performance AI hardware. As water is cheaper than electricity, firms lack strong cost drivers to reduce consumption, creating a reliance on public relations incentives amid environmental criticisms. This dynamic pressures the industry to innovate and improve transparency regarding sustainability efforts.
Furthermore, water use is only one aspect of the environmental footprint of AI data centers. The indirect water needed for power generation and manufacturing of chips and servers may double the total water impact of operations. Addressing water use comprehensively requires advances in cooling technology as well as broader supply chain and energy sourcing reforms.
What to watch next
Industry observers will be monitoring the adoption rate of newer, water-efficient cooling solutions like Nvidia’s DSX and how widely such technologies can be retrofitted into existing data centers. The economic trade-offs between water savings and energy consumption for cooling will also be under scrutiny, especially during extreme weather events or in water-stressed regions.
Regulatory and public pressure may accelerate disclosure standards and ESG reporting uniformity, as inconsistent data makes it difficult to benchmark and compare company progress. Additionally, innovation in reducing the hidden water footprint from electricity generation and hardware production could become a critical part of the conversation as AI expansion continues to demand growing data center capacity.