Britain's AI Boom Needs Water It Doesn't Have, Water Industry Warns

Water UK, a group that speaks for water companies across Britain and Northern Ireland, told parliament on 21 July 2026 that the United Kingdom will not have enough water to serve planned datacentre growth. It called current government forecasts for water demand "fatally flawed" because they "explicitly exclude" datacentres.
A datacentre is a large building full of powerful computers that store information and run online services, including artificial intelligence (AI). These computers generate enormous amounts of heat, so they need to be cooled, and cooling them uses a lot of water.
The warning came in a written briefing to MPs, published on the parliamentary committees evidence portal (The Guardian). Water UK argued that the government "makes not a single mention of water" in three flagship AI policy documents: its AI growth zones plan, its AI for science strategy, and its AI opportunities action plan. The Environment Agency's national framework for water resources, published in June 2025, similarly failed to include an estimate of datacentre water demand, according to the trade body.
The geographic mismatch sharpens the problem. More than three-quarters of UK datacentres are located in the south and east of England, regions already classified as water-stressed. Water UK told MPs that any new datacentre growth in those water-stressed areas "may be impossible" without policy change. The UK's single highest concentration of datacentres is in Slough, served by Affinity Water, which reported that about 125 datacentres are proposed or under construction in its region. In May 2026, Affinity Water told MPs that some datacentres in its area were requesting up to 3 million litres of water per day, equivalent to the peak demand of 3,500 homes (The Guardian). According to Global Action Plan, 84% of water-intensive datacentre developments in the UK are proposed for areas already under water stress (Global Action Plan).
The scale of consumption is significant. A Water UK annex document, citing the engineering firm Arup, states that a small one-megawatt datacentre using traditional cooling methods would consume more than 25 million litres of water annually. Datacentres primarily use water for cooling servers, including through evaporative cooling towers, which remove heat by letting water evaporate. Globally, the average mid-size datacentre uses 1.4 million litres of water daily for cooling (Reuters).
These figures arrive against a backdrop of tightening national supply. In May 2026, a House of Lords report found that England faced a 5 billion litre per day shortfall for public water supplies by 2055. Water UK noted that major reservoirs can take 15 years to plan and build. The trade body also pointed to progress on leakage: water companies have cut leakage by approximately 40% since privatisation and plan a further 17% reduction by 2030 (The Guardian).
The UK government has pledged £68 billion in AI investment since January 2025 (Chatham House), but the tension between that industrial ambition and water infrastructure planning is not confined to Britain. In June 2026, UN researchers projected that AI would double global datacentre power and water consumption by 2030, with datacentre water consumption reaching 9.3 trillion litres and CO2 emissions rising to 399 million tons (Reuters). Almost half of all datacentres globally are located in areas of high or very high water stress, according to S&P Global (Reuters). North American datacentres alone consumed nearly 1 trillion litres in 2025 (Reuters). Global water demand is projected to outstrip supply by 40% by the end of the decade (Reuters).
In April 2026, the environmental charity Global Action Plan published a report titled "Not a Drop to Drink," examining how Britain's AI datacentre surge threatens water security (Global Action Plan). Water UK's own annex documents reinforce the same concern from the supply side, framing the sector's strategic planning regime as inadequate for the scale of demand now entering the pipeline.
The broader context here is a timing problem. AI investment moves in months; water infrastructure moves in decades. Water UK's submission to parliament puts the trade body on the side of those arguing that the gap between the two has not been reckoned with in the government's own planning documents. Whether MPs treat the "fatally flawed" assessment as a technical correction or a fundamental challenge to the UK's AI growth strategy will depend on how seriously the relevant select committees engage with the evidence. What is clear from the data is that the conflict is structural: datacentres are concentrating precisely where water is scarcest, and the planning frameworks that should reconcile those pressures do not yet account for the demand.


