UN report warns AI data centers could consume more water than global drinking needs by 2030

Artificial intelligence does not only rely on massive amounts of computing power and electricity. According to a new United Nations report, the rapid expansion of AI data centers could also drive an enormous rise in water consumption, with serious environmental consequences.

The authors warn that, if current trends continue, facilities powering AI systems may be using more water for cooling by 2030 than the total annual amount of drinking water consumed by the world’s population.

UN report: AI’s hidden thirst for water

Public debate about the environmental impact of AI usually focuses on rising energy demand, the carbon footprint of training large models and the challenges of building ever more data centers. The latest UN analysis adds another critical piece to this picture: water use.

According to the report, data centers that support AI workloads could require around 9.3 trillion liters of water each year by 2030 to cool servers and related infrastructure. The comparison used by the authors is stark: this volume would exceed the annual drinking water needs of the entire global population.

The report also estimates that by the end of this decade, AI could account for roughly 3 percent of worldwide electricity consumption. The greenhouse gas emissions associated with this activity are projected to be similar in scale to the current annual emissions of the United Kingdom.

Even today, the scale of resource use is significant. The document notes that in the previous year, data centers consumed an amount of electricity comparable to the entire energy demand of Saudi Arabia, one of the world’s largest energy consumers.

When efficiency leads to more use: the Jevons paradox

One common argument in technology discussions is that future generations of AI models will be more efficient, and that this will reduce their overall environmental impact. The UN report challenges the assumption that efficiency gains will automatically translate into lower total consumption.

The authors point to the Jevons paradox, a phenomenon first described in the 19th century by British economist William Stanley Jevons. He observed that improvements in the efficiency of coal use did not reduce total coal consumption. On the contrary, cheaper and more efficient use encouraged broader adoption, leading to higher overall demand.

The report suggests that a similar dynamic may play out with AI. As models become cheaper to run and more accessible, more companies, public institutions and individuals are likely to integrate AI into their operations and daily lives. Any savings from efficiency improvements could then be offset, or even overwhelmed, by rapid growth in the number and intensity of AI applications.

Unequal distribution of AI’s benefits and costs

The UN analysis also highlights how unevenly AI infrastructure and its environmental impact are distributed across the globe. Large-scale cloud and data center capacity dedicated to AI currently exists in only 32 countries.

Moreover, about 90 percent of AI cloud computing power is concentrated in just two nations: the United States and China. This means that while AI services may be used worldwide, the majority of energy and water consumption tied to these systems is clustered in a small number of regions.

This imbalance raises questions about who gains the economic and technological benefits of AI, and who bears the environmental and social costs, especially in areas that may already struggle with water scarcity or strained power grids.

Environmental price of AI’s rapid growth

The UN report does not argue for halting AI development, but it underlines that its environmental footprint—especially water use—needs to be treated as a central policy and engineering challenge, not an afterthought.

As demand for AI services continues to grow, the choice of data center locations, cooling technologies, water management strategies and energy sources will all play a crucial role in determining how heavy the environmental price of this technological wave will be.