The AI Drought

Planet or profit? The devastating consequences of the AI boom.

It has become painfully apparent, AI has destroyed the world as we knew it. Now, as the dust settles we are watching the new world emerge; record job losses, climate damage, emerging studies showing the already detrimental effects on our brains- and even AI marriages (yes, people are really marrying AI bots).

But there's one issue being lost and obscured amongst the ruble: the AI drought.

Whenever you prompt AI, whether that is to write your shopping list or draft the email you have been putting off, something enormous happens behind the scenes. Your single line of text is sent across continents into a warehouse filled with thousands of humming servers, each one burning through electricity and generating intense heat as it attempts to predict your next word. The heat builds and begins to rise through the racks, and the cooling systems scramble to keep the machines alive. Water is pumped in from local supplies and pullThis so-called digital revolution is in reality a physical industry with a vast and hungry, near insatiable footprint. The data centres built to power AI are using staggering volumes of fresh water at an obscene and terrifying scale. The Environmental and Energy Study Institute rang the alarm bells: a single large data centre can consume up to five million gallons of water in a single day, a volume comparable to the annual needs of a small town. Yes, a small town. And that's in a single day, at a single data centre. Globally, data centres now consume an estimated five hundred and sixty billion litres every year. To put that insane number into perspective, that amounts to more than two hundred thousand Olympic sized pools. 

Even more alarming, the location of these data centres.

Nearly two thirds of the data centres built since 2022 are located in high water stress regions. The industry is placing its thirstiest facilities in the driest places because land is cheap, regulations are weaker and communities have less power to fight back. If that's not cause for concern, I don't know what is.

We are witnessing a post code lottery of man made water shortages, of AI droughts. When a tech giant arrives in town and begins extracting millions of gallons of local water every week the burden falls first on the families who already live with fragile infrastructure and rising bills. Low income neighbourhoods are the ones that feel the first drop in water pressure. They are the ones told to limit shower time and reduce garden water use while a concrete fortress on the edge of town runs twenty four hours a day to generate endless prompts.

It is undisputable, this is injustice. Would you like a data center built near you? Behind the glamorous public narrative of AI there is a quieter insidious one. It is the story of low income communities absorbing the shock of an industry that treats water as an infinite resource simply because they are not the one paying the price when it runs out.ed through industrial chillers, then expelled as vapour into the air, each request evaporating a little more of the community’s drinking water. What feels like a tiny and harmless digital action on your screen becomes a real world physical event that drains water from rivers and aquifers, strains municipal systems and ties local residents to the silent thirst of the global AI machine.

In Newton County, Georgia a single data centre owned by a major technology company uses five hundred thousand gallons of water every day. That is ten percent of the entire county’s consumption diverted to one industrial site. Residents describe sudden drops in water pressure, unexplained outages, strange fluctuations in reliability. Local officials insist that everything is under control, but the numbers do not lie. One building is using the same amount of water that entire neighbourhoods depend on.

Communities in Arizona, Nevada and parts of California report that groundwater levels are falling faster than before and many point directly to the arrival of high demand facilities that draw from the same aquifers as households and farms. When a data centre extracts water around the clock it does not merely stress the system, it competes with the basic needs of the people who live there. Farmers face reduced irrigation allocations. Families find that taps sputter in the evening. Environmental groups warn that fish and wildlife in already fragile ecosystems are being pushed closer to collapse.

And yet, despite the evidence right in front of us, this pattern is repeating across the world. In the Aragon region of northern Spain a proposed facility owned by Amazon has been licensed to use more than seven hundred thousand cubic metres of water every year. That amount could irrigate two hundred and thirty three hectares of corn, yet it is being channelled into a complex designed to help train AI models and process global cloud traffic. Local residents and farmers fear that the region’s water shortages will intensify and that the economic value claimed by the project will not match the ecological price they will ultimately pay.

The UK Environment Agency has openly admitted that the rapid growth of AI infrastructure makes it increasingly difficult to predict water shortages. Regional authorities warn that the digital sector has outpaced their capacity to model future demand. Every new data centre introduces an unpredictable strain and planning systems are struggling to catch up.

These are the real life consequences of AI, felt by people like you and me. They are unfolding in kitchens, where families notice the tap pressure fall. They are unfolding on farmland where crops fail due to lack of irrigation. They are unfolding in our rivers and streams. And the people carrying the burden are not the ones set to profit, they are not the shareholders or the executives, the billionaire founders. They are ordinary people, possibly soon to be you and I. After all, how can we compete against a competitor as powerful as AI.

It is easy to say that water is abundant on a planetary scale, yet this is a misleading comfort. Only a small fraction of global water is freshwater and an even smaller share is accessible or safe for human use. The real danger lies not in total volume but in timing and geography. When vast quantities are withdrawn in regions already experiencing drought or stress the impact is immediate and severe. The statistics we are seeing should be a global warning siren, yet they remain largely overlooked because regulators and water agencies do not even have accurate data on how much water data centres use, how much they reuse or how much vanishes through evaporation. This opacity hides the true scale of the threat. As AI development shifts into hotter climates the demand for cooling will intensify and water use will rise just as climate change reduces freshwater availability and extends drought across entire regions. Municipal systems may simply fail under the strain. Aquifers could fall to levels that take decades to recover. The burden will fall hardest on low income, rural and drought prone communities. Water may not be finite in the strict scientific sense, but in the places where people live and in the moments when they need it most it becomes painfully scarce and AI is accelerating that scarcity.

So, how do we even begin to solve this, or at least reduce the harm it is causing? If the AI industry is to exist without draining the world around it, transparency must come first. Companies that build and operate data centres should be required to disclose exactly how much water they use on site and indirectly through the power grid. At present many do not and the absence of this information allows the scale of extraction to remain almost invisible. Facilities should not be placed in water stressed regions simply because the land is cheap or the planning laws are weak. Cold climates and regions with abundant supply or recycled water should be prioritised and technologies that reduce dependence on evaporative cooling must be adopted as standard. No facility should be approved unless governments and water agencies have fully accounted for the burden it will place on local infrastructure. Importantly, communities must have a genuine voice in the process and guarantees that their own water security will not be sacrificed for the benefit of an industry that delivers few local rewards.

We need to understand that every prompt and every click has a physical cost and that the convenience of AI is being subsidised by rivers, reservoirs and neighbourhoods. The AI drought is not a metaphor but a real, growing threat and its consequences will fall hardest on the communities already stretched by drought, inequality and climate stress. If we continue to prioritise profit over the basic conditions of life the taps will run dry long before the industry slows.

The choice is stark and it is already upon us. Planet or profit. The drought is the reckoning.