Global Data Center Water Use May Triple by 2030 Amid AI Infrastructure Boom
Surging artificial intelligence infrastructure threatens to push global data center water consumption to 644 billion liters annually by 2030, introducing material operational and regulatory risks for technology investors and operators.
Global data center water consumption could nearly triple from 222 billion liters in 2025 to 644 billion liters annually by 2030, according to Rystad Energy. This surge is driven by the rapid expansion of artificial intelligence computing capacity. Active mitigation strategies could limit this demand to 388 billion liters in an aggressive water-saving scenario.
For investors and operators, managing this footprint presents complex engineering and financial trade-offs. While rack-level liquid cooling and technologies proposed for platforms like NVIDIA’s Vera Rubin can reduce facility-level heat burdens, they do not eliminate water dependency. Dry cooling systems, for instance, save approximately 2.15 liters of water per kilowatt-hour of IT load but require an additional 0.30 to 0.74 kWh of electricity.
This energy trade-off highlights the importance of the indirect water footprint embedded in electricity generation. In the United States, indirect water consumption for data centers can exceed direct on-site usage by more than double. Consequently, a facility’s overall water intensity depends as much on the local power grid as on its internal cooling architecture.
Current water-use effectiveness varies dramatically across major operators. In recent reporting periods, Amazon Web Services recorded an average direct site water-use effectiveness of 0.12 liters per kWh, while Meta reported 0.19 liters per kWh. Microsoft, Digital Realty, and Equinix posted figures of 0.27, 0.59, and 0.91 liters per kWh, respectively.
These aggregate metrics mask severe regional disparities driven by climate and local infrastructure. AWS regional data shows a spread from 0.02 liters per kWh in Stockholm to 2.85 liters per kWh in Jakarta. Such variance underscores why geography remains a primary determinant of operational risk and cooling costs.
Regulatory scrutiny is intensifying in response to these pressures. The European Commission is developing performance standards and labeling for data center energy efficiency, while Singapore has targeted a maximum of two liters per kWh. In the United States, Texas recently halted new data center approvals pending audits of local water and power usage.
Water stress will increasingly dictate where and how these facilities can operate. Regions facing high water stress are projected to account for 34 percent of global direct data center water consumption by 2030. Mandating the least water-intensive cooling technologies in highly exposed areas, such as Reeves County in Texas or Jamnagar and Thane in India, could reduce regional consumption by 45 percent.
Major hyperscalers including AWS, Google, Microsoft, and Meta have pledged to become water positive by 2030. These commitments involve replenishing more water than they consume through freshwater restoration projects. However, the actual impact on local watersheds will remain uneven, leaving operators exposed to localized regulatory and reputational risks.
