Data centre

Data centres: water consumption equivalent to that of 1.3 billion people by 2030

A UN report highlights the enormous water footprint of artificial intelligence – an aspect that is often overlooked compared to emissions – and the associated global environmental and social risks.

by Rome Editorial Staff

5 June 2026

2' min read

Translated by AI

Versione italiana

Key points

2' min read

Translated by AI

Versione italiana

By 2030, the data centres powering artificial intelligence will consume as much water as the entire population of sub-Saharan Africa – that is, over 1.3 billion people.

This is the picture painted by the new report from the Institute for Water, Environment and Health at the United Nations University (UNU-INWEH), published at a historic moment when the race for water is becoming increasingly frenzied.

Water and the land: often overlooked

The report highlights that the real problem is that the environmental impact of AI is measured almost exclusively in terms of CO₂ emissions, whilst water and land use are overlooked.

The most surprising paradox concerns renewable energy itself: switching from coal to bioenergy reduces the carbon footprint by 70%, but increases the water footprint by more than 30 times and the land footprint by 100 times.

Another common misconception, debunked by the report, concerns energy consumption: it is not the training of the models that is the main culprit, but their day-to-day use.

Producing a short video using a generative system can require up to 200,000 times more energy than a simple text query.

Real-life cases: from Ireland to Uruguay

The study analyses situations that have already occurred. In Ireland, one of Europe’s main hubs for data centres, these facilities consumed 21% of the country’s total electricity in 2023, exceeding the consumption of the country’s entire urban population.

In Uruguay, plans for a new data centre with high water consumption coincided with the drought of 2023, which had already rendered Montevideo’s tap water undrinkable.

Electronic waste and inequality

By 2030, AI infrastructure could generate up to 2.5 million tonnes of electronic waste each year, and the critical raw materials required for the hardware are concentrated in regions with weaker environmental oversight, often in countries in the Global South.

The benefits of AI are concentrated in the North; the environmental costs tend to be borne elsewhere.

What UN experts are calling for

The report proposes a “responsible AI ecosystem” based on transparency, design efficiency, environmental justice and global cooperation.

Investors and data centre operators should treat energy, water and land as material risks, not as externalities; local communities should be involved in decision-making from the outset.

“This report is not an indictment of the AI,” said Kaveh Madani, director of UNU-INWEH.

“It is a call for responsible use, which proactively addresses unintended consequences in order to make this technology sustainable and equitable.”