Ask a chatbot to plan your holiday. Stream a film on your phone. Upload a few holiday photographs to the cloud. Each action feels almost effortless,a few taps, a few seconds of waiting, and the job is done. But behind every digital interaction lies an industrial-scale network of data centres, packed with thousands of servers working day and night. And behind those servers lies an unexpected ingredient that is becoming increasingly important in the age of artificial intelligence: water.

The internet is often described as “the cloud,” a term that makes it sound almost weightless. In reality, it is anything but. Every search, video stream, AI-generated response and online transaction passes through machines that consume enormous amounts of electricity and generate equally enormous amounts of heat. Left unchecked, that heat would damage delicate electronics, slow computing speeds and eventually bring systems to a halt.

Cooling these machines is therefore not a luxury but a necessity, and water remains one of the most effective ways to do it. Unlike air, water absorbs and carries away heat quickly and efficiently. In many data centres, warm water is circulated away from the servers and into cooling towers, where part of it evaporates, taking the excess heat with it. The process is remarkably energy efficient, but there is a trade-off. The evaporated water is lost forever and must be replaced with fresh supplies, day after day.

Artificial intelligence has dramatically changed the equation. Traditional cloud servers experience peaks and lulls in activity, but AI processors operate under relentless workloads. Training large language models or generating images requires specialised chips running close to their maximum capacity for long periods, producing far more heat than conventional computing. More heat demands more cooling, and more cooling often means more water. This is why AI has quietly become one of the fastest-growing drivers of water consumption in the technology sector.

Yet the water footprint of AI extends well beyond the walls of a data centre. Keeping these facilities running requires a continuous supply of electricity, much of which still comes from coal and natural gas-fired power stations. These plants themselves consume vast quantities of water to produce steam and cool turbines. Even before a server is switched on, water has already been used extensively to manufacture its semiconductor chips, where ultra-pure water is essential for cleaning silicon wafers during production. In effect, every AI query carries with it a hidden trail of water stretching from chip factories to power plants and finally to the data centre itself.

This growing dependence on water is colliding with another global challenge: climate change. AI is fuelling an unprecedented expansion of data centres, making them among the fastest-growing consumers of electricity in the world. If that electricity is generated from fossil fuels, it adds more greenhouse gases to the atmosphere, accelerating global warming. Rising temperatures then increase the need for cooling, forcing data centres to consume even more water and electricity. At the same time, climate change is making droughts more frequent and more severe, reducing freshwater supplies in many parts of the world. The result is a troubling cycle: the infrastructure driving the AI revolution is becoming increasingly dependent on a resource that climate change is making harder to secure.

The scale of the issue is no longer theoretical. Data centres in the United States alone are estimated to consume well over 100 billion gallons of water every year, and similar facilities are rapidly being built across Europe, Asia and the Middle East. In regions already grappling with water scarcity, a single hyperscale data centre can consume as much water as a small town, forcing governments and communities to confront uncomfortable questions. Should scarce freshwater be reserved for households and agriculture, or diverted to support the rapidly growing digital economy?

Technology companies are not blind to these concerns. Many are investing in closed-loop cooling systems that recycle water instead of allowing it to evaporate. Others are experimenting with immersion cooling, where servers are submerged in specially engineered fluids that transfer heat even more efficiently. Some companies are building facilities in naturally cooler climates to reduce cooling needs, while others have pledged to become “water positive” by replenishing more water than they consume.

These initiatives are encouraging, but they are unlikely to keep pace with AI’s explosive growth on their own. Artificial intelligence is rapidly becoming embedded in healthcare, education, banking, manufacturing and countless everyday applications. Every breakthrough brings new computational demands, which in turn require more servers, more electricity and, in many cases, more water.

The remarkable capabilities of AI often make it seem detached from the physical world, but the opposite is true. Every prompt typed into a chatbot, every recommendation generated by an algorithm and every file stored in the cloud depends on an infrastructure that consumes energy and freshwater on an industrial scale. As the race to build smarter AI gathers speed, the real challenge may not be creating more powerful algorithms, but ensuring that the technology shaping our future does not deepen the climate crisis or place even greater strain on one of the planet’s most precious and increasingly fragile resources.