AI is consuming water at a rate we can't sustain 

Dr. Muhammad Wakil Shahzad, Associate Professor and Head of Mechanical and Construction Engineering at Northumbria University, explores a potential solution to the environmental toll of AI-powered data centres.

Data centres are the hidden backbone of our digital ecosystem. Every Google search, Netflix stream, cloud-stored photo, and ChatGPT response relies on vast networks of powerful computers housed in enormous facilities scattered worldwide.

These data centres consume immense amounts of electricity—but increasingly, they are also devouring significant quantities of water. Unlike the water used in our homes, much of this water doesn’t return to the ecosystem. Environmental scientists are becoming increasingly alarmed by this trend. A recent study (yet to be peer-reviewed) predicted that by 2027, AI’s water consumption could surpass the total water use of the UK in one year.

Inside a data centre, thousands of servers run 24/7, generating massive amounts of heat. If not properly cooled, these machines risk overheating and failing. This was evident in 2022 when a record heatwave in the UK, with temperatures soaring to 40°C, caused outages in Google and Oracle data centres in London. Cooling these systems is a critical concern, and this is where water comes in.

Water for Cooling

One of the most common cooling methods is mechanical refrigeration. These systems, like large industrial fridges, use refrigerants to absorb heat from the servers and expel it through a condenser. However, a significant portion of the water used in this process evaporates and cannot be reused.

A 1-megawatt (MW) data centre—equivalent to the energy consumption of 1,000 homes—can use up to 25.5 million litres of water per year. With the UK’s total data centre capacity estimated at about 1.6 gigawatts (GW), and global data centre capacity at approximately 59 GW, the scale of water consumption is staggering.

Unlike household water, which usually cycles back through treatment systems, water in cooling systems vanishes into the air as vapour. While it’s still technically part of the Earth's water cycle (returning as rainfall), not all of it can be recovered. For this reason, the water used in data centres essentially becomes "lost" to the local water balance. This is especially concerning in water-scarce or drought-prone regions—where two-thirds of data centres have been constructed since 2022.

AI’s Growing Impact

With the rise of AI technologies like ChatGPT, image generators, and voice assistants, data centres are under increasing pressure. These systems require far more computing power, which in turn produces more heat. To keep up, data centres are using even more water for cooling.

The International Energy Agency (IEA) reported in April 2025 that data centres now consume more than 560 billion litres of water annually, with projections suggesting this could increase to 1,200 billion litres per year by 2030.

Searching for Alternatives

A possible alternative is direct evaporative cooling. This method pulls hot air through water-soaked pads, cooling it through evaporation before circulating it back into server rooms. While this approach is energy-efficient—particularly in warmer climates—it introduces higher humidity levels, which can damage sensitive equipment. This added complexity necessitates additional systems to control the moisture in the air, making the design of data centres more intricate.

At Northumbria University, my research team and I have developed a different solution: separating moist and dry air streams with a thin sheet of aluminium foil, similar to kitchen foil. In this system, the hot, dry air flows near the moist air stream, and heat is transferred through the foil without mixing the moisture. This method cools server rooms without increasing humidity levels, reducing the risk to equipment.

Initial trials of this technology at our university's data centre have shown that it is not only more energy-efficient than traditional cooling methods but also uses less water. Plus, the system operates entirely on solar power, eliminating the need for compressors or harmful chemical refrigerants.

Looking Ahead

As AI continues to expand, the demand on data centres will likely skyrocket—and so will their water consumption. If we’re going to avoid a serious environmental crisis, a global shift in how we design, regulate, and power digital infrastructure is essential. The need for innovative, sustainable cooling solutions is more urgent than ever.