Residents in Slough are currently enduring a hosepipe ban during a period of extreme drought, yet 32 local data centres continue to draw massive quantities of water to cool AI servers. This disparity has sparked scrutiny over how industrial cooling needs are prioritized over residential water security in a designated water-stressed region.
Thames Water's 74% reservoir capacity and the Slough hosepipe ban
The tension in Slough is driven by a stark contrast in water access. While local citizens are urged to take shorter showers and avoid watering lawns, commercial premises in the Slough Trading Estate are not subject to the current hosepipe and sprinkler restrictions. This occurs as Thames Water reservoirs sit at 74 per cent capacity, significantly lower than the 87 per cent usually expected for this time of year.
The town is officially classified as being in a water-stressed area, meaning demand frequently rivals or exceeds the available supply. As reported, the lack of restrictions on industrial users means that while residential gardens turn brown, the cooling towers of the data centre hub continue to operate without the limits imposed on private citizens.
32 data centres consuming 65% of Slough's electricity
The scale of the digital infrastructure in Slough is immense, with 32 data centres estimated to consume 65 per cent of the town's total electricity. This massive energy draw generates significant waste heat, which larger facilities manage through water-based evaporation cooling systems. According to the report, these systems become most demanding during summer heatwaves,exactly when the local water supply is most precarious.
There are also concerns regarding the "heat island" effet created by these facilities. a Cambridge University study suggests a large data centre can theoretically raise the temperature of its immediate surroundings by up to 2C. While Slough's weather station recorded 37.6C on August 13—slightly lower than the 38.1C recorded in nearby Wisley, Surrey—the localized heat radiating from these buildings remains a point of contention for residents.
The half-litre cost of 10 to 15 AI tasks
The surge in artificial intelligence is exponentially increasing the water requirements of these facilities. A University of California study found that a data centre consumes roughly a half-litre bottle of water for every 10 to 15 tasks an AI performs. Given that large-scale facilities process millions of requests daily, the cumulative strain on the water table is substantial.
The energy intensity of these models is similarly staggering. A study by the Massachusetts Institute of Technology noted that Meta's Llama 3, released in July 2024, requires as much electricity to generate a simple travel itinerary for Istanbul as it takes to ride an e-bike for six feet. This combination of high energy use and subsequent cooling needs creates a feedback loop that threatens ageing infrastructure.
The missing data on 400-plus British data centres
A critical gap in oversight exists regarding exactly how much water these facilities use. While the trade body TechUK reports that 1 in 25 data centres use more than 100,000 cubic metres of water annually—equivalent to the consumption of 3,000 households—this data is based on a survey of only 72 facilities. this leaves the water usage of over 400 other data centres across Britain entirely unknown.
The lack of transparency makes it difficult to assess the true environmental cost of the AI boom. Without comprehensive reporting on the remaining 400-odd facilities, policymakers cannot accurately determine if the current water infrastructure can sustain the projected growth of the digital economy in water-stressed regions like Slough.
Comments 0