UN: AI data centres to consume more power than Nigeria, two other countries 

Global data centres will consume nearly three times the combined annual electricity of Nigeria, Pakistan, and Bangladesh, by 2030.

This is according to a United Nations report published on Wednesday, which noted that the power consumption would be driven largely by the explosion in everyday AI usage rather than model training.

The report, published by the UN University Institute for Water, Environment and Health, projects that data centre electricity consumption will rise from 448 terawatt-hours in 2025 to 945 TWh by 2030, with AI accounting for 40% of that total, up from a fifth today.

For context, Pakistan, Bangladesh, and Nigeria together are home to more than 650 million people.

Water consumption from data centres is projected to more than double from 4.5 trillion litres to 9.3 trillion litres by 2030, enough to meet the basic water needs of 1.3 billion people in Sub-Saharan Africa.

Carbon emissions will rise from 189 million tons to 399 million tons over the same period, while the land footprint of data centres will expand from 6,900 square kilometres to more than 14,500 square kilometres.

What the report is saying 

The report identifies inference, the continuous running of deployed AI models to answer everyday user queries, as the dominant cost, accounting for 80 to 90% of total AI energy use.

• ChatGPT alone processes an estimated 2.5 billion prompts daily, consuming roughly 383 gigawatt-hours of electricity annually. Offsetting those emissions would require 2.6 million tree seedlings grown for a decade.
• AI image generation requires approximately 1,450 times the energy of basic text classification.
• A single short AI-generated video consumes as much electricity as 200,000 spam classifications. A complex AI video carries a water footprint of 4.1 litres, nearly a two-day drinking supply for one person.

The report notes that if current growth trends continue, global data center electricity consumption could nearly double to 945 terawatt-hours (TWh) by 2030.

• “At that level, data centers would consume almost three times as much electricity annually as Pakistan, Bangladesh, and Nigeria combined, countries that together are home to more than 650 million people”.

 More insights 

The report documents cases where data centre expansion has already collided with local resource limits. In Ireland, data centres accounted for 21% of total metered electricity in 2023, exceeding all urban households combined, prompting the national grid operator to pause new approvals around Dublin until 2028.

Only 32 countries currently host AI-specialised data centres, with more than 90% of global AI compute capacity concentrated in just two, the United States and China. More than 150 countries have little or no access to sovereign AI computing infrastructure.

• AI infrastructure could generate up to 2.5 million tonnes of electronic waste annually by 2030, equivalent to discarding nearly 250 Eiffel Towers each year, much of it processed in low-income economies with limited environmental safeguards.
• Critical minerals required for AI hardware are concentrated in regions with weaker environmental oversight, predominantly across the Global South, including parts of Africa.

What you should know   

Nigeria’s active push to attract data centre investment, anchored in the National Cloud Policy 2025 and accelerated by facilities like Digital infrastructure provider, Kasi Cloud’s LOS1 commissioned in Lagos in May 2026, sets the country up to capture economic benefits from AI infrastructure.

• Kasi Cloud has commissioned the first phase of its planned 100-megawatt data centre facility, a development expected to help reduce Nigeria’s dependence on foreign cloud services and retain an estimated $850 million spent annually on offshore cloud infrastructure.