Indonesian netizen unpacks AI data center water use debate

Indonesian user @AryHHAry offered a systematic response across multiple posts to various criticisms of "AI data center water consumption," with the core argument being: the key issue isn't whether Earth's total water volume is conserved, but who gets first claim on local freshwater and whether the timescale of the water cycle can keep up.

Confirmed

• The author acknowledges that Earth's total water is nearly conserved and evaporative cooling doesn't destroy water molecules, but points out that cooling merely transfers water from municipal pipes into the atmosphere; the water cycle takes time, and local residents can't wait. What's truly scarce is local freshwater, not the global total.
• On cooling principles, the author explains that closed-loop systems only solve water circulation, not heat dissipation: in tropical climates, the most energy-efficient way to expel heat remains evaporative cooling—the pipes can be closed, but the water vapor is open.
• Regarding the two water-use pathways for data centers: closed-loop circulation consumes very little; evaporative cooling turns water into vapor to carry away heat, and the leftover concentrated wastewater (blowdown) is hard to reuse simply—this answers the question "why not do water recycling."
• Against claims citing global impacts from data centers (Amazon droughts, Indian Ocean storms, Japan floods), the author argues for distinguishing scales: evaporative cooling alters the seasonal water balance of the local watershed (DAS), and water vapor from one basin doesn't directly become heavy rain in another region.
• On the comparison that "bottled water plants consume more water," the author finds it inapt: bottled water plants extract water as a commodity and ship it elsewhere, while data centers evaporate water into vapor, converting it to latent heat that can't be recovered; moreover, the "several times higher" usage figures are macro-level totals, which is a different matter from whether local residents' wells run dry.

Why it matters

This set of posts offers a more precise middle-ground perspective than either "total conservation means no water use" or "AI is draining the planet": the real cost of AI data center water use is concentrated in seasonal freshwater allocation within the local watershed and the time lag of the water cycle. The discussion needs to separate spatial and temporal scales to avoid macro-level figures masking real impacts on local residents.

2026-09-29 ~ 2026-09-30 · 5 related posts

• [source] Data center water use isn't about total volume, it's who runs out of local freshwater first — AryHHAry · 2026-09-29
• [source] AI datacenters can recycle cooling water loops, but evaporated steam is gone for good — AryHHAry · 2026-09-30
• [source] Why data centers still need evaporative cooling: closed loops handle flow, not heat — AryHHAry · 2026-09-30
• Data center cooling water debate: evaporative cooling hits local basins, not global climate — AryHHAry · 2026-09-30
• Bottled water plants and data centers consume water differently: one sells it, one evaporates it — AryHHAry · 2026-09-30