The AI Boom in Malaysia Has a Water Bill. The People Downstream Didn't Agree to Pay It.
The data centers powering Malaysia's AI ambitions use up to 50 million liters of water a day.
In Johor, residents are finding out what that means in practice.
In February 2026, more than 50 residents from four housing estates in Gelang Patah, Johor, protested outside a Zdata Technologies data center construction site less than a kilometer from their homes. Their concern wasn't abstract. In November 2025, pollution in Sungai Johor had forced four water treatment plants offline, leaving 1.8 million residents without water for up to 12 hours. Then a data center designed to draw tens of millions of liters a day started going up next door.
Johor stopped approving Tier 1 and Tier 2 data centers last year due to high water usage. That restriction didn't stop projects already in pipeline. Malaysia is now Southeast Asia's fastest-growing data center hub — Johor's aggregate capacity nearly doubled in a single year to 5.8 gigawatts by mid-2025. That growth doesn't run on goodwill. It runs on water and coal-fired electricity.
Who this really matters to:
→ Property developers and industrial park operators in Johor — community opposition to data centers is becoming a permitting and reputational risk, not just a footnote in an EIA report → Malaysian businesses selling to MNC clients with ESG requirements — if your supply chain runs through Johor data centers burning coal and drawing local water tables, your scope 3 emissions disclosure needs to account for it → SME manufacturers in Johor Bahru and Pasir Gudang industrial zones — data centers and factories compete for the same grid capacity and water pressure; rising AI infrastructure demand directly increases your operational costs → Technology procurement teams at GLCs evaluating data center vendors — water usage per megawatt, source, and recycling rates are standard due diligence in Singapore and Australia; Malaysian enterprise clients are starting to ask too
MULTIPLE PERSPECTIVES
The government's response is real but limited. Malaysia restricted non-AI data center approvals to reduce grid and water pressure — faster regulatory instinct than most countries in Southeast Asia managed. But restricting new approvals doesn't reverse the pipeline of gigawatts already committed and under construction. Water and grid pressure in Johor will worsen over the next 24 months regardless.
The second-order effect that gets almost no coverage: the communities protesting in Gelang Patah aren't anti-AI. They're asking a straightforward infrastructure question: whose water supply is being drawn down to power cloud services that mostly benefit businesses in KL and Singapore? The residents of Gelang Patah are not seeing cheaper internet or lower electricity bills. They're seeing construction dust and dry taps. That's not a fringe grievance — it's the kind of tension that produces policy overcorrection when it compounds.
The counterintuitive read: the most important beneficiary of Malaysia's data center boom is not Malaysian businesses. It's Singaporean businesses using Malaysian cloud infrastructure at rates they couldn't afford on Singapore land. The communities hosting these facilities are absorbing the environmental externality without the pricing benefit. The infrastructure investment that Malaysia needs for its AI ambitions is being paid for, in part, by Johorean residents' water security.
If Malaysia's AI infrastructure boom is mostly benefiting businesses in Singapore and KL — who should be compensating the communities in Johor that are hosting it?
If you're procuring from or operating data center infrastructure in Johor, the ESG accounting and community impact question is already reaching the boardroom. The regulatory framework to price these externalities is coming.
If you're not in that sector, watch the corridors taking shape in Kulim and Cyberjaya. The same pattern is being built there.
The benefits of AI infrastructure should reach the communities that host it. Right now, mostly they don't.

— Tony
Sharing what I learn building real things with AI.