AI threatens Southeast Asia’s water supplies

Artificial intelligence is adding a new strain to Southeast Asia’s water resources as data centers expand to support the region’s digital economy. The facilities require substantial electricity, but their servers also need water for cooling.

Malaysia has positioned itself as the initiator and lead developer of ASEAN guidance on sustainable data-center construction. The proposal became a work priority at the fifth ASEAN Digital Ministers’ Meeting in Bangkok in January 2025. During Malaysia’s 2025 ASEAN chairmanship, it led the drafting of the guidance, whose adoption was welcomed at the sixth ministerial meeting in Hanoi in January 2026.

The water demand can be substantial. According to the ASEAN guidance, a one-megawatt data center using evaporative cooling can consume more than 25 million liters of water a year. As AI workloads grow, they generate more heat and increase the need for cooling.

The guidelines therefore include a dedicated section on water management and cooling innovation. They encourage operators to measure Water Usage Effectiveness, or WUE, assess water stress before choosing a site and reduce reliance on treated water needed by local residents. Reclaimed and non-potable water, along with more efficient cooling systems, are also encouraged.

For data centers in areas facing high water stress, developers are urged to prepare water-management plans and consider alternative sources where available. The guidance also recommends aligning data-center locations with national water-resource planning. It sets a regional average WUE benchmark of about 1.8 to 2.0 by Jan. 1, 2030, but does not impose a uniform mandatory rule. Countries can adapt it to their climate, water resources and technology. Malaysia introduced domestic sustainable data-center guidelines in December 2024, recommending a design WUE of 2.2 cubic meters per megawatt-hour or lower, with a target of 2.0 or lower within 10 years.