AI boom could strain global water and power supplies by 2030, UN report warns

Artificial intelligence may appear to exist only in the digital world, but a new report warns that the data centres powering the technology could place severe pressure on the world’s electricity, water and land resources by 2030.

The report by the United Nations University Institute for Water, Environment and Health (UNU-INWEH) says AI is not merely software. Behind every model and online service is a vast physical network of data centres, energy systems and hardware supply chains that require large amounts of natural resources.

Without more effective management, the rapid expansion of AI could intensify pressure on local communities, widen global inequality and shift environmental burdens to regions least able to absorb them, the report says.

Data centres’ impact goes beyond carbon

The report warns that the environmental footprint of AI data centres should not be measured by greenhouse

gas emissions alone.

Electricity use remains a major concern, but the report says water consumption and land use are also critical factors that are often overlooked in sustainability assessments.

It cautions that relying on a single measure, such as carbon emissions, may hide other environmental costs, especially when energy policies are changed in ways that reduce emissions but increase pressure on water or land.

Professor Kaveh Madani, director of UNU-INWEH, said the report was not an argument against artificial intelligence, but a call for its responsible use.

He said stronger action was needed to address unintended consequences and ensure that the AI revolution develops within the planet’s limits.

Electricity demand could reach 945 terawatt-hours

By 2030, data centres powering AI worldwide could consume up to 945 terawatt-hours of electricity per year, according to the report.

That would be about three times the annual electricity consumption of Pakistan, Bangladesh and Nigeria combined — countries with a total population of more than 650 million people.

If AI data centres were counted as a country, the industry could become the world’s sixth-largest electricity consumer by the end of the decade, accounting for about 3% of global electricity use.

AI’s water demand could surge

The report also estimates that AI-related water consumption could rise to 9.3 trillion litres by 2030.

This water would be used mainly for cooling servers and supporting electricity generation needed to keep AI systems running continuously.

The volume is equivalent to the basic annual household water needs of around 1.3 billion people in sub-Saharan Africa, highlighting the scale of the potential burden on water resources.

Land use also under pressure

AI infrastructure could require more than 14,500 square kilometres of land by 2030 — an area nearly 10 times the size of Bangkok.

The report says this land would not be used only for data centre buildings. It would also include space for power plants, reservoirs, fuel extraction and mineral supply chains needed to produce and operate AI hardware.

Professor Shaolei Ren, an expert in AI sustainability, said AI’s impact extends far beyond models and algorithms.

He said data centres, energy systems and hardware supply chains all create substantial physical and environmental consequences.

Everyday AI use drives most energy demand

The report found that everyday AI use, including inference and routine user queries, may account for 80-90% of total AI energy demand.

This challenges the common belief that training AI models is always the most energy-intensive stage.

Popular AI services now process billions of user prompts each day, meaning ordinary use at scale can become a major driver of resource consumption.

The type of task also matters. Generating one image can use up to 1,000 times more energy than a standard text classification task, while producing a short video may require up to 200,000 times more energy than a basic AI operation.

Alex Hernandez, a researcher at MILA, noted that these figures may still underestimate the true impact because some of the data was based on older models such as GPT-4.

Green energy may not solve every problem

The report says efforts to make data centres greener, including the shift to renewable energy, are important but may create new challenges.

For example, replacing coal with bioenergy could reduce carbon emissions by 70%, but may increase the water footprint 30-fold and the land footprint 100-fold.

The report says this shows how solving one environmental problem can create another if decisions are based only on carbon emissions.

Miriam Aksel, the report’s lead author, warned that judging AI sustainability only by carbon could create a misleading picture.

She said renewable energy may make AI appear cleaner, while in reality it could shift environmental pressure to regions that do not benefit equally from the technology.

Digital divide and e-waste risks grow

The report also raises concerns about inequality in the AI economy.

Only 16% of countries currently have dedicated AI data centres, while 90% of this capacity is concentrated in the United States and China.

More than 150 countries lack their own AI infrastructure, a gap that risks deepening the global digital divide.

At the same time, electronic waste from AI hardware is expected to become a growing crisis. By 2030, AI-related hardware waste could reach 2.5 million tonnes a year, equivalent to discarding 250 Eiffel Towers annually.

Much of this waste is likely to end up in lower-income countries, where safe disposal systems may be limited.

Local impacts already visible

The report says pressure from data centres is already being felt in some countries.

In Ireland, data centres accounted for 21% of national electricity use in 2023.

In Uruguay and Mexico, plans to build large data centres during periods of severe drought have triggered public concern and protests over whether clean water for households could be diverted to industry.

Call for responsible AI ecosystem

Dr Tshilidzi Marwala, rector of the United Nations University, said AI has the potential to create wealth and improve well-being, but fairness will not happen automatically.

He said governments and technology developers must urgently build an AI ecosystem that is responsible, transparent and guided by planetary limits.

The report calls for greater transparency from AI companies, stronger government planning for resource use, and more selective use of AI by consumers and organisations.

It says AI should be used where it is necessary and where its benefits clearly outweigh its environmental cost, especially for high-impact tasks that require large amounts of energy, water and computing power.