The race to scale AI infrastructure is accelerating. In the UK, billions are flowing into new data center capacity as operators respond to surging demand for high-performance compute. Much of the discussion has centered on power grid capacity, generation, and how quickly new supply can be brought online.

Power is critical. But it is not the only factor that will determine whether this growth is sustainable. Water is becoming just as important, and in many cases far less well planned for.

England is forecast to face a shortfall of around five billion liters of water per day by 2055. When set against warnings that some of the world’s largest data centers can consume up to 19 million liters of water a day – equivalent to supplying around 50,000 homes – it is clear that AI-driven growth will intensify pressure on already stretched resources. If data centers are to operate reliably, water cannot sit at the end of the planning checklist; it has to be designed in from the outset.

AI is changing the cooling equation

AI workloads have fundamentally altered how data centers generate and manage heat. High-density compute and GPU clusters operate at much higher thermal loads than traditional server environments, increasing the need for advanced cooling.

While many facilities still rely on air-based or hybrid systems, the shift towards liquid cooling is accelerating as operators seek greater efficiency and performance for AI. That shift brings clear benefits, but it also increases dependence on reliable water supplies. According to the World Economic Forum, a 1MW data center can use up to 25.5 million liters of water per year for cooling alone – equivalent to the daily consumption of hundreds of thousands of people.

Even where greywater or alternative sources are used, quality thresholds remain high. Water containing significant waste or particulates cannot be introduced into cooling systems without specialist treatment, otherwise performance and equipment lifespan are compromised. As AI capacity scales, water demand will continue to rise accordingly.

Water and power are now inseparable

As data centers become more energy-intensive, power strategy and water strategy are increasingly linked. In the UK, small modular reactors (SMRs) are being explored as a future power source for large AI facilities. But nuclear generation depends on substantial volumes of water for cooling. Rather than removing water from the equation, it reinforces its importance.

Recent European heatwaves offer a warning. Nuclear plants in France and Switzerland were forced to reduce output or temporarily shut down when river water temperatures exceeded safe cooling thresholds. In warmer conditions, facilities may need to chill intake water before use, adding both complexity and even more water demand.

For data center operators, this highlights the extend that cooling resilience depends more than just access to water. Temperature, seasonality, and long-term availability need to be factored in.

Climate pressure and competing demand

Water stress is no longer a future risk. UK regulators are warning of potential widespread drought as early as this year, and research suggests that growing scarcity could leave some regions facing industrial shortfalls, threaten net-zero targets, and increase the risk of drinking-water shortages without new reservoirs. Rising temperatures compound the issue. Just one degree increase significantly increases evaporation and reduces stored water across catchments. Heat domes, now common across Europe and the US, create prolonged dry spells where demand peaks just as supply drops, directly affecting cooling performance and operational reliability.

At the same time, data centers are competing with other fast-growing sources of demand. Housing expansion, industrial development and agriculture all draw from the same finite resource base. In England alone, 1.5 million new homes are planned over the next five years.

These pressures sit on top of water systems already under strain. Leakage, ageing infrastructure, delayed reservoir projects and environmental compliance challenges have eroded public trust, with analysis indicating that parts of England and Wales could face severe and recurrent freshwater shortages within a decade without significant investment. Yet digital infrastructure is rarely factored into national water planning with the same seriousness as energy, increasing delivery risk and making late-stage intervention more likely.

From “shovel ready” to “shovel worthy”

A recurring issue across infrastructure projects is that water enters the conversation too late. Site selection, design, and power planning are often fixed before local water capacity or long-term climate conditions are fully assessed. By then, options are limited and costs escalate.

Data centers need confidence not only in today’s availability, but in conditions over a 20- to 30-year operating life. That requires integrated planning from the outset, with water and power considered together rather than in silos. Early engagement with water authorities and regulators reduces risk, shortens delays, and improves long-term resilience.

The next phase of AI infrastructure will be shaped by which projects are genuinely deliverable under real-world constraints. Being “shovel worthy” means recognizing water as a core design input, not an afterthought.

Power enables AI growth. Water sustains it. Neglect either, and even the most ambitious data center strategy becomes vulnerable.