In an Expert Focus article for WaterBriefing, Jonathan Taylor, UK Managing Director of engineering, technology and consulting services provider Expleo, takes a look at how AI is driving new infrastructure demand and explores how the UK water sector should respond.

Jonathan Taylor: As the UK looks to expand its AI infrastructure, attention is turning to the role of water in new data centres. To date, location, power and computing capacity have dominated the debate. But the role of water availability in determining whether proposed sites can support development has received less attention, partly because there is no single measure of what a data centre will need.

That unknown sits within a much broader supply challenge. The Environment Agency’s National Framework for Water Resources warns that England could face a public water supply deficit of up to five billion litres a day by 2055, while demand beyond the public water system is also expected to rise. Data centres are one of several emerging pressures that planners will need to account for.

How water is shaping data-centre planning

That issue is now beginning to shape policy. In its National Framework for Water Resources, the Environment Agency calls for water availability to be considered early in data-centre planning, with particular attention to site location and the requirements of the chosen cooling technology.

This is further reflected in the Government’s AI Growth Zone application criteria, which treat water as a test of whether a proposed site can support development at the scale envisaged. Applicants seeking to accommodate at least 500MW of AI infrastructure must provide confirmation from the relevant supplier that the necessary volumes are available, identify any constraints in the local network and explain how long additional capacity would take to deliver. They must also account for wastewater discharge, bringing the full water cycle into the assessment before a project moves forward.

That puts water firmly within the feasibility discussion for new AI infrastructure. It is not simply an environmental impact to assess once the main development decisions have already been made.

The requirement for supplier confirmation is more than a procedural hurdle. It forces projects to confront whether the water capacity promised on paper can be delivered on the same timetable as the development itself, before capital is committed and options begin to narrow.

The challenge here is that data centres are not built to the water sector’s timetable. A project may be ready to move long before the supply schemes it relies on, tying investment to capacity that could still be years away. That tension is now feeding into a much wider debate about how England plans and pays for its water infrastructure.

This challenge comes as the Department for Environment, Food and Rural Affairs (Defra) sets out plans to overhaul England's water sector. Its January 2026 white paper, A New Vision for Water, proposes more co-ordinated regional planning and a longer-term approach to infrastructure investment. For water companies, this means taking account of emerging demand from developments such as data centres alongside existing commitments to maintain resilient public water supplies.

The role of data in understanding demand

Planning for this additional demand is difficult without reliable information on water consumption. The Environment Agency points to wide variation in water use across the data centre sector and says greater transparency is needed to improve understanding of demand.

Water companies and regional planners make decisions over decades, while digital infrastructure can move from proposal to operation on a much shorter cycle. If demand is poorly understood when plans are set, capacity assumptions can become outdated before major new water resources are delivered.

The timing is particularly significant for water companies delivering their AMP8 investment programmes, which cover £104 billion of expenditure between 2025 and 2030. At the same time, the Government proposes abolishing Ofwat and creating an integrated regulator for England, bringing economic oversight together with relevant functions currently held by the Environment Agency and Drinking Water Inspectorate. While companies must continue delivering existing AMP8 commitments, the reforms could reshape the next price review, PR29, and how future investment is assessed. Reliable demand forecasts will be essential to support those decisions.

Assurance before scale

UK Government guidance describes AI assurance as a way of establishing and communicating whether AI systems can be trusted to perform as intended. It gives organisations evidence about how a system behaves in practice, where its limitations and risks lie and how those risks are being managed throughout its lifecycle.

For organisations scaling AI, that assurance mindset also encourages closer scrutiny of the conditions needed for a programme to succeed. An AI system may perform as intended, but deployment at scale still depends on the infrastructure around it. Where significant compute capacity is involved, the availability of power and water can affect whether the original assumptions behind a programme remain viable.

This makes assurance particularly valuable as organisations move beyond pilots. Assumptions made during early development should be tested again as workloads increase and operating conditions change. The aim is to identify the critical dependencies and test whether the evidence still supports the assumptions behind them. If conditions change, the programme should be able to respond accordingly.

From data to operational decisions

While AI infrastructure adds to water demand, the technology could also help water companies manage growing pressure on their networks. The proposed reforms also place greater emphasis on understanding the condition of existing infrastructure, with calls for improved asset mapping and stronger resilience standards. AI could support these ambitions by identifying emerging maintenance requirements and improving demand forecasting.

Digital twins and predictive analytics can help operators anticipate how networks will respond as demand and operating conditions change. But their value rests on the information beneath them: incomplete asset records, unreliable forecasts or outdated assumptions can turn a sophisticated model into a source of false confidence. For critical infrastructure, the test is whether that evidence is sound enough to support the decision at hand.

One infrastructure system

Water and digital capacity are part of the same infrastructure problem. Treating them separately risks solving one constraint while creating another elsewhere.

The Government's AI Growth Zone criteria already recognise the importance of water availability in infrastructure planning, while Defra's proposed reforms aim to create a more coordinated approach across the water sector. Yet regulatory reform alone will not resolve the practical challenges of forecasting demand and delivering infrastructure. Water companies and developers need to test these assumptions before investment decisions become difficult to reverse.

The bigger risk is that investment moves faster than the evidence and infrastructure needed to support it. Decisions made on optimistic assumptions about future capacity risk leaving operators managing constraints that could have been identified much earlier.