AI Datacenter Water Usage: A Choice, Not a Tech Problem
alex_verem · x · 2026-07-20
AI datacenter water consumption is surging. Global use is projected at 264 billion gallons in 2025 (including electricity generation), enough to supply 1.8 million Americans for a year.
However, this is a solvable problem. Sweden uses closed-loop cooling, consuming almost no freshwater, and pipes waste heat into municipal grids: Stockholm heats 30,000 apartments, Microsoft's Finnish datacenter will heat 250,000 people, and Google covers 80% of Hamina's heating demand for free. Germany now legally requires new datacenters to recover at least 10% of waste heat.
In contrast, most U.S. datacenters still use evaporative cooling to save upfront costs, doing so in severe drought zones like Arizona and Texas. The author argues the issue is not AI itself, but the industry's choice to prioritize cheap cooling methods in drought-prone areas while calling it 'innovation.'
Related event: AI Data Center Water Consumption Sparks Debate Over Tech Choices(2 posts)→
More from Infra
- Your p99 latency benchmark may be lying: a deep dive into coordinated omission — Franc0Fernand0 · 2026-09-11
- Running MiniMax H3 on 12GB VRAM: quantization, Turbo LoRAs and attention backends compared — Possible_Mood676 · 2026-09-11
- Spomin: live KV cache compaction squeezes 500k tokens of context into 180k resident — wgaca2 · 2026-09-11
- PiPNN nearest-neighbor search wins three awards, up to 78x faster index building — khademinori · 2026-09-11
- M.2-Oculink eGPU Link Silently Downgrades to PCIe Gen1 — Here's How to Check — El_90 · 2026-09-11
- DeepSeek launches V4.1-Flash with 1M-token context and 4x smaller KV-cache — matlabulous · 2026-09-11