“Bubble” is probably the word most associated with “AI” right now, though we are slowly understanding that it is not just an economic time bomb; it also carries significant public health risks. Beyond the release of pollutants, the massive need for clean water by AI data centres can reduce sanitation and exacerbate gastrointestinal illness in nearby communities, placing additional strain on local health infrastructure.

AI’s energy consumption is massive and increasingly water-dependent

Generative AI is artificial intelligence that is able to generate new text, photos, code and more, and it has already infiltrated the lives of most people around the globe. ChatGPT alone is reported to receive around one billion queries in a single day, pointing to huge demand at the individual level.

This, however, is only the tip of the iceberg. Companies such as Google, Apple and Microsoft are now embedding AI into their key products. Applications that utilise search results are quickly moving to have AI as a new standard in their algorithms. Whether it is shopping on Amazon or booking a flight or a hotel, AI is now being used in searches, and that demands more energy. As an example, a single AI-powered Google search is estimated to use up to 30 times more energy than its standard version.

Graphics Processing Units (GPUs) are the current industry answer to this. They are chips that demand energy and produce heat. Though the thousands of small cores on GPUs enable parallel processing of massive, repetitive maths carried out by AI algorithms, a single chip can use up to 700 watts. This means that three chips alone can use roughly the same amount of energy as a home electric oven.

The large amount of heat produced by data centres is cooled by up to hundreds of thousands of gallons of fresh water each day. With thousands of heat-generating chips stacked next to and on top of one another, a simple fan does not do the trick. Instead, water is pumped or immersed between and around chips in order to avoid system overheating. A recent report from the United Kingdom’s Government Digital Sustainability Alliance predicts that AI will increase global water usage from 1.1 billion cubic metres to 6.6 billion cubic metres by 2027.

Some companies are attempting to use seawater in cooling. However, fresh water continues to be widely used for cooling in many facilities. Water recycling is another option, but not a simple one. Several companies use a “closed-loop system” to reduce the total amount of water needed. Nevertheless, dust and minerals collected during cooling can degrade water quality over time, requiring treatment or replacement.

How AI-driven water scarcity threatens public health

Data centres being placed where water is already scarce can quickly translate into a healthcare burden, even before pollution becomes an issue. In 2023, Microsoft reported that 41 percent of its water withdrawals were from areas with water stress. Google, on the other hand, said that 15 percent of its water consumption occurred in areas with high water scarcity. Amazon did not disclose comparable figures.

It is well established that water scarcity correlates with infections, malnourishment and declining hygiene. While most such studies focus on areas that are already impoverished, in many cases, these are exactly the places where data centres are planned to be built. In addition, the underlying cause remains the same. Less fresh water for local populations pushes households to prioritise drinking and cooking over washing hands, food or bathing. Naturally, this also leaves less water available for cleaning living spaces.

The World Health Organization recognises that unsafe drinking water and poor sanitation and hygiene are conducive to the spread of diseases such as cholera and other diarrhoeal illnesses, along with a range of other pathogens. To make matters worse, several diseases associated with water scarcity can pass from person to person, raising the risk of sustained local transmission.

The health burden on children is particularly alarming, as infections and deaths are more common than among adults. In fact, 84 percent of the global burden of diarrhoeal disease is borne by children under five, and infections with diarrhoeal pathogens have been linked to cognitive impacts later in childhood.

Although it is too early to draw direct causal links between AI data centres and water-related diseases, the known facts make this a significant concern. It is established that AI data centres can significantly deplete local water supplies. It is also established that communities with poor water access face heightened risks of gastrointestinal disease and other illnesses.

To claim that AI data centres are directly causing gastrointestinal disease would be poorly supported. However, the warning signs are increasingly difficult to ignore. When risks are foreseeable and severe, governments should not wait for people to start dying before putting preventative policies in place.

Marginalised communities are already reporting polluted water

In Newton County, Georgia, in the United States, Meta has built an AI data centre, and residents have reported discoloured, sediment-filled water coming out of their taps, which they attribute to the facility. Similarly, in Fayette County, residents have reported sediment in their water, which they believe coincided with nearby data centre construction. Another report from California suggests that a data centre planned along the San Francisco Bay in Bayview-Hunters Point has raised concerns about compounding environmental burdens in an already polluted community. In all these cases, the local population includes a significant Black and African American presence, a pattern that has raised environmental justice concerns.

Accumulated residues can result in effects ranging from acute gastrointestinal illness to chronic conditions such as cancer. Microbial contamination can cause poisoning and acute disease, while chemical residues are associated with long-term harm, often acting as a slow, invisible threat.

With plans for data centres in African countries such as Nigeria, Egypt and South Africa, further questions arise about who will bear the brunt of their environmental impacts and whether affected communities will receive sufficient protection or support. Weak regulatory oversight in some of these countries makes this uncertain. In many cases, serious community effects may go unreported altogether.

Why corporate water promises and regulation matter

Only 0.5 percent of the planet’s water is fresh water, and water is not only needed for data centres. It is also required for the power plants that generate electricity for them. The manufacture of chips and wiring similarly demands water, making water use an AI supply-chain issue rather than merely a data-centre problem.

Many companies are promising sustainability, with some even claiming they aim to be “net water producers” or “water positive”. Even if such targets are achieved, which remains questionable, they must deliver benefits to the communities from which water is extracted. Providing more water for affluent areas while depleting supplies in places such as Newton County may satisfy corporate accounting standards, but local residents will still suffer the consequences.

To meet their ethical obligations to the public, governments must rapidly catch up with the pace of AI expansion and data-centre construction. A healthy population is a productive one, and a lower public-health burden can reduce government spending while supporting development. More fundamentally, there is a collective moral obligation to build a sustainable future for coming generations by safeguarding water security and averting environmental catastrophe. This begins with legislation mandating transparent corporate reporting on water use and enforcing meaningful standards for sustainable management. Regulation must prioritise human wellbeing over short-term, extractive technological growth. As with climate change, unrestrained innovation risks further harm to both people and the planet.

The views expressed in this article are the author’s own and do not necessarily reflect Al Jazeera’s editorial stance.