Can the UK’s Water System Cope with the AI Data Centre Boom?

Rising Demand for Water for Cooling Data Centres

A recent UK government–commissioned report on water use in data centres and AI has sounded a clear warning: England could face a daily water shortfall of nearly 5 billion litres by the mid-2050s, and current water-resource plans do not fully account for the extra demand from digital infrastructure.

The Environment Agency’s National Framework for Water Resources projects that, by around 2050–2055, England may be short of about 5 billion litres per day for public water supply, roughly one-third of today’s public consumption, with additional billions of litres needed for industry and the wider economy. AI growth zones, hyperscale data centres and cloud/AI hubs are largely absent from these forecasts, despite being on track for rapid expansion.

AI, data centres and water demand

Large data centres, especially those hosting AI workloads, influence water resources in two main ways:

• Direct use on site: Many facilities rely on evaporative or adiabatic cooling, which can consume substantial volumes of water to reject heat. In hot or dry periods, this can amount to millions of litres per site over relatively short timescales.

• Indirect use through electricity generation: Data centres draw huge amounts of power, much of which is still produced by thermoelectric plants (gas, biomass, nuclear) that themselves use cooling water. As AI pushes electricity demand higher, water use rises both on-site and across the power system that feeds these facilities.

The report stresses that overlooking this combined demand risks worsening water stress in already vulnerable catchments and undermining long-term planning for both people and industry.

Cooling choices and the energy–water trade-off

The UK wants to be a global hub for AI and digital infrastructure while also remaining climate-resilient and water-secure. Cooling technology sits right at the intersection of those goals.

Broadly, operators face a balancing act:

• Evaporative / adiabatic systems

  • Very efficient at removing large heat loads

  • Can significantly cut electricity use compared with purely air-based approaches

  • But depend on continuous water consumption, which is problematic in water-scarce regions

• Air-only and closed-loop systems

  • Use far less water or virtually none

  • Typically require more electrical power for fans, chillers or dry coolers

  • May limit how far rack densities can be pushed without redesign

As AI density increases, these trade-offs become sharper. A design that looks attractive on energy grounds may impose a heavy water footprint, while a water-lean solution may increase power demand and emissions if not carefully optimised.

Pathways towards lower water footprints

The government report calls for better monitoring and disclosure of data centre water use, and for future infrastructure planning to reflect digital demand more accurately. On the technical side, several cooling directions are attracting attention:

• Direct-to-chip liquid cooling to reduce reliance on large air flows and enable higher densities with less total cooling energy.

• Two-phase and immersion cooling that use boiling and condensation to move heat more efficiently, potentially allowing smaller, more efficient plant and less dependence on evaporative towers.

• Improved cold-plate, micro-channel and heat-exchanger design, bringing cooling closer to the silicon and reducing thermal resistance.

• Waste-heat recovery, for example feeding low-grade heat into district-heating schemes or nearby industrial processes, improving overall system efficiency.

Each approach brings its own profile for water use, electricity demand, complexity and cost. Robust design and simulation are essential to compare options on a like-for-like basis.

Implications for planning and infrastructure

Projected water deficits are driven by a combination of climate change, population growth, environmental protection and economic activity. With AI and data centres added to that list, infrastructure planning needs to view digital projects as major water and energy users, not just as buildings with large electricity bills.

This has several consequences:

• Location decisions need to consider local water availability and catchment stress, not just land price and grid connection.

• Cooling architectures should be chosen with explicit awareness of long-term water constraints as well as energy efficiency.

• Integrated modelling of water, energy and cooling is increasingly important at the early concept stage to avoid costly redesigns later.

High-quality thermal–fluid modelling, techno-economic analysis and scenario testing can help quantify the trade-offs between heat rejection, water consumption and efficiency, and show how designs perform under future climate and demand conditions rather than just today’s averages.

Looking ahead

The core message from the report is stark:

England could be billions of litres short of water every day by mid-century, and current planning does not fully include the demands of AI and data-centre growth.

As AI infrastructure scales up, its cooling strategies and siting choices will play a visible role in regional water security. The way new data centres are designed – their cooling technologies, their relationship to local water and energy systems, and the degree to which their heat can be reused, will help determine whether the UK can expand digital capacity without deepening long-term resource stress.

Source: https://shorturl.at/Kg513