By: Dwight Links
The ‘Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints’ report from the UN University’s Institute for Water, Environment and Health (UNU-INWEH) points to the impact of the rapid global artificial intelligence (AI) developments.
“One of the most consequential dimensions of AI that remains comparatively under-examined is its environmental footprint and the justice implications that follow,” the report highlights.
The key focus areas have been the size of land required for data centres housing AI infrastructure, the amount of power to keep it running, and the water required to cool the processing.
“Expansion involves physical infrastructure and supply chains, including data centres, chips, electricity generation, cooling systems, water withdrawals, land occupation, critical minerals, and eventual e-waste,” notes the report.
According to Professor Kaveh Madani, co-author of the report, there is a misperception that the technology and hardware running the AI processing is somehow cleaner and cheaper to operate, but that is not the case.
“A lot of people think that the environmental footprint of AI reduces, as technology improves and processes become more efficient. But that is only a partial picture of the overall problem,” stated Madani.
Environmental Costs of AI
The report notes that the scale and ability invested in training the AI models is a subject of appetite for water, electricity and carbon emissions.
“The training of frontier models demands immense energy. GPT-4 likely consumed 50 to 70 GWh of electricity over 100 days, roughly 40–55 times more than GPT-3, which is 1.287 GWh over a 34-day period,” illustrates the report.
GPT-4’s training energy is said to be equivalent to the annual residential electricity consumption of over 460 000 people in Sub-Saharan Africa.
“GPT-4’s training carbon footprint of 25 000 tonnes of carbon-dioxide-equivalent (CO₂e) would require the sequestration capacity of 420 000 tree seedlings grown for 10 years, or about equal to the number of trees in 105 Hyde Parks in London,” further describes the report.
The most obvious remedy that policymakers look at is how the footprint is mitigated.
Another worrying depiction is that the water footprint associated with training GPT-4 was about 600 million litres, enough to meet the minimum annual domestic water needs of 81 000 people in Sub-Saharan Africa, or to fill 237 Olympic-sized pools, indicates the report.
“Projections for models like GPT-5 suggest training electricity requirements of 100 GWh, equivalent to the annual residential electricity usage of 770 000 people in Sub-Saharan Africa,” the university describes.
According to the UNU-INWEH, training models such as the GPT-5 is estimated to have a carbon footprint of 42 000 tonnes of CO₂e, requiring 700 000 tree seedlings – about equal to the number of trees in 40 Central Parks in New York or 155 times the trees in Toronto’s High Park over 10 years to offset.
“The water footprint of GPT-5 training is estimated at 1 billion litres, enough to meet the annual domestic water needs of more than 135 000 people in Sub-Saharan Africa,” shared Madani.
True Costs of AI
“If you map where data centres are built against where water stress is worst, you tend to see the same regions in some instances,” Dr. Mir Matin pointed out.
Matin is a manager of the UNU-INWEH’s Geospatial, Climate and Infrastructure Analytics Programme and a co-author of the report.
“And the communities living near these sites are not necessarily the ones using the AI being run there. That asymmetry is the issue. Without fixing it, we’ll just be repeating older patterns, where some places carry the costs and other places capture the benefits,” added Matin.
“While the AI infrastructure comes with environmental costs, they also have major economic, security and sovereignty advantages that encourage the wealthier countries to build more data centres,” the report states.
This is as only 32 countries in the world host AI-specialised data centres, and 90% of that capacity is concentrated in 2 countries, while more than 150 countries currently have little to no access to sovereign AI computing.
Market Investments
The global AI market is expanding rapidly, which is projected to grow from US$189 billion in 2023 to nearly US$5 trillion by 2033, the researchers note in their report.
Current trends they observed would mean global AI expenditure is projected to exceed US$2.5 trillion in 2026.
Generative AI accounted for over 20% of the global AI market in 2026, and is projected to reach 40% by 2030.
Corporate AI investment exceeded US$580 billion in 2025, while generative AI alone attracted nearly US$34 billion in private investment.
