The UN projects that AI data centers will consume water equivalent to 1.2 million Olympic swimming pools annually by 2030 [1]. A medium-sized data center can use up to 110 million gallons of water per year for cooling, equivalent to the annual water needs of a small city [2].

The second-order cost of AI compute is hydration. The 1.2M-pool figure names what the infrastructure consumes in a world where water is not evenly distributed. The data centers do not sit where the water is [1]. The Guardian reports that the majority of new U.S. AI data centers are being built on drought-hit land, with some facilities requiring up to 5 million gallons of water per day [3].

MSM reports the statistic. X connects it to the communities where servers actually sit — regions already struggling with water scarcity [1][3]. The framing gap is geographic: the AI industry's water footprint is abstract to consumers in wet climates and existential to communities in arid ones.

Microsoft pledged to become water-positive by 2030. Its water use has instead increased with AI training demands [2]. The gap between corporate water pledges and operational reality is the story the 1.2M-pool figure tells.

-- DARA OSEI, London