A UN report warned that global data centres could consume 945 TWh of electricity by 2030, nearly triple the combined annual electricity use of Pakistan, Bangladesh and Nigeria
The rapid advancement of artificial intelligence (AI) and the data centres powering it could consume as much electricity by 2030 as the annual residential needs of 1.3 billion people in Sub-Saharan Africa, according to a report published on Tuesday by the United Nations University Institute for Water, Environment and Health (UNU-INWEH).
The report, titled Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints, warned that global data centres could consume 945 terawatt-hours (TWh) of electricity by 2030, which is nearly triple the combined annual electricity use of Pakistan, Bangladesh and Nigeria.
The report noted that these three countries are collectively home to more than 650 million people. It added that the associated water footprint of data centres would equal the minimum annual domestic water needs of all 1.3 billion residents of Sub-Saharan Africa, while the land footprint linked to their electricity use would exceed 14,500 square kilometres, roughly twice the size of the Jakarta metropolitan area, which is home to more than 32 million people.
AI data centres consume millions of litres of water for cooling systems that help prevent servers and high-performance processors from overheating during intensive computing operations. This has raised growing concerns over water use and resource stress, particularly in drought-prone and water-scarce regions where large data centres can place additional pressure on already strained supplies.
In 2025, global data centres consumed an estimated 448 terawatt-hours of electricity, the UN report stated, adding that if treated as a nation, they would have been the world’s 11th-largest electricity consumer, behind France and ahead of Saudi Arabia.
Why does every AI prompt matter?
“ChatGPT alone is estimated to process around 2.5 billion prompts per day, translating to roughly 383 GWh of electricity per year for a single product,” the report added.
The report further noted that AI’s energy consumption varies sharply depending on the type of task being performed. A typical conversational AI query is around 200 times more energy-intensive than basic text classification, while generating a single AI image can require nearly 1,450 times more energy. Meanwhile, a short AI-generated video can consume as much electricity as 200,000 spam classifications. The researchers said factors such as model choice, prompt length, output format and video resolution can significantly alter the environmental footprint of AI systems, even though many of these decisions are determined by default settings that users rarely see.
Why does AI’s environmental cost go beyond carbon emissions?
However, the researchers in the report argued that the environmental costs of AI and data centres cannot be understood through carbon emissions alone. Every unit of electricity used to train or run AI systems also carries a water footprint from cooling and power generation, and a land footprint linked to energy infrastructure. These impacts often move in different directions. For instance, shifting from coal to bioenergy may reduce carbon emissions while sharply increasing water and land use.
The researchers warned that “low-carbon” does not automatically mean “low-water” or “low-land”, and said relying on a single sustainability metric can obscure trade-offs and shift environmental stress onto already water- or land-scarce regions.
How uneven is the global burden of AI infrastructure?
The environmental burdens of AI infrastructure are being unevenly distributed across the world, the UN report stated. It pointed to Ireland, where data centres accounted for 21 per cent of total metered electricity consumption in 2023, prompting the national grid operator to pause new approvals around Dublin until 2028 due to mounting pressure on the power system. In Mexico and Uruguay, expanding data centre infrastructure has intensified concerns over water use during prolonged droughts.
The researchers further warned that AI infrastructure could generate up to 2.5 million tonnes of electronic waste annually by 2030, much of it likely to be processed in low-income countries with limited environmental safeguards, while the extraction of critical minerals needed for AI hardware continues to place pressure on regions with weak oversight.
