- The UN warns that the environmental footprint of AI is much, much more than just energy
- AI data centers could consume the water equivalent of 1.3 billion people by 2030
- Report calls for more diverse reporting and strong governance to protect people
A new UN report argues that the impacts of artificial intelligence are far from equal; Instead, its environmental impact is being underestimated because most discussions only focus on carbon emissions.
Instead, the United Nations is urging companies, investors and governments to also include water consumption and land use in their assessments.
This comes as AI data centers alone are expected to consume 945 TWh of electricity by 2030, the equivalent of 1.95 billion households, or three times the population of Pakistan, Bangladesh and Nigeria.
The UN is concerned about the environmental impacts of AI
Leaving aside electricity, the UN also warns that its water consumption by the end of the decade would be equivalent to 1.3 billion people in sub-Saharan Africa (9.3 trillion liters), and land use could be equivalent to 14,500 square kilometers (twice Jakarta, home to 32 million people).
But it’s much more than the environment that the AI industry is putting pressure on: unlike conventional software, artificial intelligence relies heavily on the physical campuses of data centers, network connections, cooling systems and semiconductors, greatly expanding its impacts in both Scope 2 and Scope 3.
Professor Kaveh Madani, director of the Institute for Water, Environment and Health at the United Nations University, stressed that the report should serve as a roadblock to AI. Instead, Madani calls for responsibility and sustainability.
“We have a narrow window to ensure that the backbone of our era’s technological revolution develops within planetary boundaries, and that the communities that provide the minerals critical to the advancement of AI and those that host its infrastructure and e-waste are also among those that benefit from it.”
Interestingly, while much of the debate often focuses on model training, researchers now believe that inference (everyday use after deployment) accounts for around 80-90% of AI’s power demand. ChatGPT alone is said to process around 2.5 billion messages per day, and power demand only increases as response quality improves.
Looking ahead, the United Nations is called on to make mandatory reporting on carbon, land and water footprints, as well as “efficiency by design” approaches. The document also encourages stronger governance to prevent environmental costs from being passed on to the most vulnerable communities.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds.
