Summary

• Water UK has warned Parliament that datacentre water demand is not properly reflected in UK AI planning.
• Its evidence says current water forecasts generally exclude datacentres because they rely on local authority plans that often omit them.
• Water is joining power, grid connection, land, cooling, and planning as a constraint on AI infrastructure.

Water UK has warned Parliament that the government’s AI infrastructure ambitions could be constrained by water supply, arguing that datacentre demand has not been properly reflected in national planning.

In written evidence to the Environmental Audit Committee, the water industry body said the UK’s AI strategy depends on the ability to build datacentres, while several key government documents on AI Growth Zones and AI opportunities give little or no attention to the water required by many facilities for cooling. It also said official forecasts for future water needs in England explicitly exclude datacentres.

The evidence places a physical constraint inside a policy agenda often discussed through compute capacity, grid connections, and investment. Water UK estimated that UK datacentres currently use around 6.6m litres of water a day, with that figure potentially rising to 19.8m litres a day if capacity trebles. Peak demand could rise from about 14m litres a day to 42m litres a day under the same scenario.

Those figures are not large enough to explain the UK’s water problems on their own, but they expose a planning gap. Water companies build long term investment plans around forecasts, local authority plans, abstraction limits, and regulatory assumptions. If large datacentre developments are absent from those forecasts, utilities may lack the authorised headroom to serve new sites, particularly in already stressed regions.

The issue also complicates the government’s decision to classify datacentres as critical national infrastructure. That status can help protect digital services and accelerate attention to the sector, but it does not create water, grid capacity, or local consent. A facility that is nationally desirable can still become locally contentious if households and businesses are facing restrictions while new compute sites require reliable supply.

Cooling choices add another layer. Some datacentres use less water and more electricity; others use evaporative cooling that can reduce power consumption but increase water demand. That trade-off means policy cannot treat all facilities as equivalent. Site assessment, planning rules, environmental scrutiny, and developer obligations need to distinguish between cooling architecture, local water stress, grid availability, and peak demand.

Water UK has called for earlier disclosure of water requirements, changes that allow developers to fund new water supplies, minimum water efficiency standards, and greater use of non-potable water where possible. It also argued that the UK should learn from overseas regimes where datacentre water or energy reporting requirements are becoming more common.

Datacentre developers already treat power availability and grid connection timelines as central to site selection. Water now belongs in the same diligence process, alongside wastewater capacity, cooling design, drought risk, planning consent, and local infrastructure constraints. Sites that look attractive through the lens of land or power may prove harder to justify if water supply is fragile.

The UK’s AI infrastructure agenda will ultimately be paced by physical systems. More compute cannot be separated from utilities, environmental limits, and local resilience. Water UK’s intervention does not argue against datacentres; it argues that AI Growth Zones have to be planned as infrastructure projects in the real economy, where water, power, land, and public consent decide how much digital capacity can actually be built.