As multinational corporations expand their footprint across Southeast Asia, they frequently deploy a standardized environmental defense. When questioned about the ecological impact of their rapidly constructing data centers, industry giants like Microsoft, Google, and AWS invariably cite advanced cooling technologies, renewable energy offsets, and community engagement initiatives.
Adding to the skepticism, major technology firms have recently claimed to reporters that their facilities are becoming ‘less thirsty’ than in the past. This assertion forms part of a broader industry campaign to preempt growing public outrage in the United States regarding the massive water consumption of AI infrastructure.
However, this narrative has yet to resonate in Gelang Patah, Malaysia. Earlier this year, local residents gathered outside a data center construction site to voice a blunt grievance: there would not be enough water remaining for their own communities. While this protest may seem localized, it underscores a significant systemic issue that advocates for the region’s AI growth prefer to ignore.
Johor as the Testing Ground, Already Under Severe Strain
Johor has emerged as Southeast Asia’s most rapidly expanding data center hub, driven not by Malaysia’s domestic digital ambitions but by regional displacement. When Singapore halted new data center approvals between 2019 and 2022 to preserve its scarce land and water, global operators simply relocated their plans across the causeway. Consequently, Johor’s aggregate capacity has surged toward 5.8 gigawatts, with the state now hosting dozens of operational facilities that rely on a water and power infrastructure never designed for such immense industrial demand.
The figures are no longer theoretical. A single 100-megawatt facility can consume approximately 4 million liters of water daily for cooling alone. Malaysian officials have noted that hyperscale sites extract roughly 200 times more water than a typical industrial consumer.
In response, regulators have rejected a significant portion of new applications, increased industrial water tariffs, and instructed the largest prospective tenants to wait until at least mid-2027 for guaranteed water and power connections. Estimates indicate that committed demand from the state’s data center pipeline exceeds 800 million liters per day, compared to the roughly 140 million liters of infrastructure currently capable of delivering it.
The Malaysian federal government has effectively conceded that the current model is unsustainable. Prime Minister Anwar Ibrahim recently informed parliament that Malaysia has ceased approving new data centers lacking an AI connection, operating on the assumption that AI-branded projects will still secure approval. This policy persists even as the sector’s electricity demand is projected to surge to nearly a third of the nation’s total power supply by 2035, up from approximately seven percent today.
Malaysia Is Not an Exception, But a Preview
Johor should be viewed as a warning rather than an isolated incident, as the same mathematical reality is unfolding across the region with considerably less media attention. Indonesia’s Java-Bali grid, which supports the majority of the country’s data center load, was already operating near capacity before any new AI facility came online, and it remains approximately 60 to 65 percent reliant on coal—a contradiction that sits uneasily with hyperscalers’ net-zero commitments. Meanwhile, Thailand’s data center power demand reportedly quadrupled between 2020 and 2024, while generation capacity increased by less than a tenth of that rate.
Conversely, Vietnam continues to attract hyperscale investment based on its affordable land and labor, despite parts of the nation experiencing weekly power outages during summer peaks.
A recent analysis by Bain and Standard Chartered articulated this constraint plainly: the binding limitation on Southeast Asia’s AI-driven growth is not capital or ambition, but the grid itself. The region’s transmission and distribution networks have failed to keep pace with the concentrated, high-value demand that data centers represent. By 2030, hubs including Singapore, Johor, Bangkok, Greater Jakarta, Manila, and Batam are expected to require an additional 35 to 45 terawatt-hours of power, landing on infrastructure originally designed for a slower, more distributed growth pattern.
How Speculative Capacity Exacerbates the Crisis
The resource strain becomes even harder to justify when considering that a significant portion of the driving demand may not actually exist. Malaysia’s Energy Commission discovered that as of mid-2025, data centers were consuming less than half of their declared maximum electricity demand. Officials have since flagged the prevalence of speculative applications—developers reserving grid capacity and water allocations long before securing actual tenant commitments, effectively jumping the queue for scarce resources against uncertain future needs.
This reflects a familiar pattern from past infrastructure bubbles, meaning some of the water and power tensions currently experienced by communities are being generated by capacity that may never be fully utilized.
This is precisely why Johor’s response, despite being belated, warrants serious consideration as a template rather than being dismissed as mere friction. The state has rejected nearly a third of data center applications due to sustainability shortfalls, mandated a transition to reclaimed wastewater instead of municipal supply for new approvals, and established a dedicated water reuse program specifically targeted at the industry.
While none of these measures have fully closed the gap between committed capacity and available infrastructure, they represent a far more honest starting point than the alternative the region has largely defaulted to: approve first, measure the damage later.
Southeast Asia Should Not Import an Avoidable Problem
The uncomfortable truth is that Southeast Asia possesses a genuine opportunity that much of the world lacks. Its AI data center boom is still nascent enough that grid interconnection, water accounting standards, and siting regulations can be deliberately constructed, rather than retrofitted after the fact as the United States is currently attempting to do. While the ASEAN Power Grid interconnection and philanthropic clean-energy pledges for surrounding communities represent positive steps, they remain medium-term solutions for a strain that is already manifesting in increased tariffs, deferred approvals, and community protests today.
Governments across the region would be wise to stop treating data center investment announcements as unambiguous economic victories and begin asking the difficult question that Johor is now forced to confront: what does this facility actually cost the people living adjacent to it, in terms of water, power, and a grid upon which other industries and households also depend?
The alternative is a region that spends its AI infrastructure boom exporting the same environmental trade-offs that Silicon Valley is only now beginning to confront—except this time, with far less capacity to refuse.
