Just this week, the Trump administration has introduced its new ‘AI Action
Plan’ with the direct aim to speed up the adoption of AI and the build of more data centres, all the while cutting regulation to enable this. As the world continues its AI race, how much is too much?
Concerns have long been raised about the amount of energy AI uses. The
International Energy Agency has stated that, by 2030, electricity demand for AI-optimised data centres is forecast to more than quadruple. Similarly, increasingly large amounts of water are required to produce the energy needed to power AI.
On top of that, data centres themselves have started using water in order to cope with the boom in AI usage. Historically, data centres have used air cooling systems, which was sufficient to cool systems up until AI came along. Now, air cooling is no longer
enough to prevent systems from overheating. Generative AI uses up to
33 times more energy than machines running task-specific software, which is why liquid cooling mechanisms have been implemented in many data centres.
The problem with liquid cooling, however, is that it relies on clean water in order to prevent corrosion to the systems or the growth of bacteria. The solution: Drinking water. And while the water used in the cooling mechanism is designed to be re-introduced
into the natural water source, up to 80% evaporates in the process. “What this means is that this type of water is gone, and we are extracting water from a water circuit that is necessary for irrigation, for human consumption and hygiene,” Lorena Jaume-Palasí,
founder of the Ethical Tech Society, commented
in the BBC.
Dr Venkatesh Uddameri, an expert in water resources management, is
additionally quoted stating that a “typical data centre can use between 11 million and 19 million litres of water per day, roughly the same as a town of 30,000 to 50,000 people.”
Data centres put pressure on the water sources of local communities
While many tech giants like Microsoft, Facebook and Google
have pledged to be water neutral by 2030, AI has changed the equation. One of the latest concerns has been Elon Musk’s self-proclaimed ‘largest data centre in the world’ for xAI, Colossus, build in Memphis in 2024. Apart from
sidestepping environmental laws, the “expected water demand is 5+ million gallons per day in an area where arsenic pollution threatens the drinking water supply,” according to a
local non-profit concerned about the data centre’s environmental impact.
Memphis is not alone. Research by
Bloomberg News found that in the US, “two thirds of new data centres built or in development since 2022 are in places already gripped by high levels of water stress.”
In the UK, Prime Minister Starmer aims to become a world leader in AI, having pledged a
£2 billion investment to data centre growth and digital skills development. However, some of these centres are proposed to be built in areas already under threat of
water shortages. On top of that, the government has
designated data centres as ‘critical national infrastructure’, which means they will face fewer planning restrictions than traditional developments.
How do we fix the AI water problem?
There is good news as well, which is that companies are already working on technology to be more water efficient. Closed-loop and immersion cooling are being adopted by companies like Microsoft in order to implement zero-water cooling systems.
Closed-loop cooling eliminates the need for evaporating water, allowing systems to re-circulate coolants continuously. Immersion cooling is another technique applied, which sees servers submerged in non-conductive fluids. According to
DataCentre, “these fluids either circulate passively or follow a two-phase process, evaporating to absorb heat and condensing for reuse. There are no pumps, fans or water involved.”
Communities around the world are urging their governments to ensure water and energy are part of the equastion. In the UK, the
Royal Academy of Engineering has asked the government to ensure tech companies accurately report how much water and energy their data centres are using as the country gears up to become a ‘world leader in AI.’ The report also asks to introduce sustainability
requirements for all data centres, including the move to zero-water cooling in order to preserve precious fresh water reserves.
The crucial step is ensuring the water problem is being considered industry-wide, and that companies are being held accountable for their water and energy usage, as well as their dedication toward reducing their environmental impact. Unchecked AI expansion
will only move us closer to
Day Zero.