Slough, UK, is currently grappling with a water shortage that has forced residents into hosepipe bans. However, the town's 32 data centers continue to draw significant water supplies for AI server cooling without facing similar restrictions.

Advertisement

Thames Water's 74% reservoir level and the hosepipe divide

Residents in Slough are facing a precarious water situation as Thames Water reports that local reservoirs are only 74% full,significantly lower than the typical 87% expected for this season. While the utility company urges households to take shorter showers and avoid outdoor watering, the town's industrial sector operates under a different set of rules.

According to the report, the large trading estate in Slough, which hosts a growing cluster of data centers, is exempt from the hosepipe and sprinkler bans that bind private citizens. This creates a stark disparity where domestic conservation is mandated while industrial consumption remains unrestricted during a declared water-stress period.

The 100,000 cubic metre appetite of UK data hubs

The scale of water usage in these facilities is immense, though preecise local figures for Slough are not publicly available. As reported in the source, TechUK estimates that 4% of UK data centers consume more than 100,000 cubic metres of water annually—a volume capable of filling 40 Olympic-size swimming pools or supplying 3 ,000 households.

This industrial thirst is part of a wider trend where the digital economy's physical footprint clashes with local ecology. However, the full extent of the problem remains obscured; the TechUK survey only covered 72 of the more than 500 data facilities across the United Kingdom, leaving the water footprints of hundreds of sites unknown.

Slough's 65% electricity draw and the cooling challenge

The water crisis is inextricably linked to energy consumption, with Slough's data hubs estimated to use roughly 65% of the town's total electricity. This massive power draw generates intense waste heat, which larger installations must expel using water-based evaporation systems.

The environmental impact extends beyond water volume to local temperature. While a Cambridge University study suggested that large centers could raise local ambient temperatures by up to 2 degrees Celsius, the report notes that Slough's weather station recorded 37.6 degrees Celsius on the year's hottest day, slightly below the 38.1 degrees Celsius recorded in nearby Wisley.

Half a litre of water for 15 AI inference tasks

The rise of generative AI is accelerating these demands. A 2023 analysis from the University of California found that a data center consumes approximately half a litre of water for every 10 to 15 AI inference tasks. When scaled to millions of daily requests,the cumulative water requirement becomes a significant burden on local infrastructure.

Energy demands are similarly spiking. MIT researchers highlighted that a single itinerary-generation request using Meta's Llama 3 model—released in July 2024—consumes as much power as an e-bike would on a six-foot ride. This "twin thirst" for electricity and water suggests that Slough's current infrastructure may soon reach a breaking point.