SEA’s AI boom has a water problem it cannot offset away

The rapid expansion of data centers across Southeast Asia has led to a growing concern about the region's ability to meet the water demands of these facilities. While hyperscalers emphasize the use of efficient cooling methods, renewable energy sources, and community engagement, one Southeast Asian town highlights the real-life strain of this boom.

Gelang Patah, a town in Johor, Malaysia, has seen protests as residents expressed worry about insufficient water resources for both themselves and the data centers. This small protest signifies a larger issue: Johor is experiencing a surge in data center capacity, with the state hosting numerous operational facilities that strain its already limited water and power infrastructure.

Some sources estimate that the committed demand from the state's data center pipeline may exceed 800 million liters of water per day, while only around 140 million liters are currently available for delivery. The Malaysian government has acknowledged the unsustainability of this model, and it is expected that new non-AI-linked data centers will not be approved.

This situation is not unique to Malaysia; similar trends are observed in other Southeast Asian countries like Indonesia, Thailand, and Vietnam. The region's transmission and distribution networks have not kept pace with the concentrated, high-value demand that data centers represent, leading to potential speculative capacity problems.

Some experts argue that a significant portion of the demand may not even be real yet, as developers reserve grid capacity and water allocations ahead of actual tenant commitments. Some Southeast Asian countries are taking steps to address the issue, such as Malaysia's rejection of a third of data center applications based on sustainability concerns, and the implementation of a water reuse program.

However, these measures have not fully addressed the gap between committed capacity and available infrastructure. Southeast Asia may have the opportunity to design its AI data center boom with deliberate grid interconnection, water accounting standards, and siting rules, rather than dealing with the consequences of an uncontrolled expansion.