When we ask ChatGpt or GoogleGemini we don’t think about it, but every request to AI has its own environmental cost. An increasing cost as the computing capabilities of these technologies and user demands increase. The environmental cost of AI does not concern the future, but it is an issue that we should keep in mind for the present, given that the use of AI is becoming increasingly pervasive in Italy too.
Artificial intelligence consumes more and more water and energy. One fact above all should make us reflect: by 2030, the data centers that power artificial intelligence will consume an amount of water equal to the needs of 1.3 billion people, the equivalent of the entire population of sub-Saharan Africa. Not only that: electricity demand will rise to 945 terawatt hours, triple the energy used by Pakistan, Bangladesh and Nigeria combined.
The new United Nations report raises the alarm Environmental cost of AI energy consumption: carbon, water and soil footprints. The study highlights how the entire supply chain – from the construction of infrastructure to the disposal of electronic waste, up to the daily use of algorithms – urgently needs global regulation. «This report is not an indictment against AI, a technological transformation that is improving the lives of billions of people, but a call for its responsible use to make it sustainable and fair.” said Kaveh Madani, director of the UN Institute for Water, Environment and Health.
SOME CONCRETE CASE
The UN report is not limited to future projections, but photographs dynamics already underway by analyzing real cases. We report two of them.
Ireland: having become the European hub of data centers, in 2023 the island’s “digital factories” absorbed 21% of all national electricity, exceeding the total consumption of the entire urban population of the country.
Uruguay: Plans to build a huge, water-guzzling data center coincided with the historic drought of 2023, which dried up freshwater reserves in the capital Montevideo, making tap water temporarily undrinkable for citizens.
FOR A RESPONSIBLE ECOSYSTEM
But how is the ecological footprint of AI calculated? A common mistake is to evaluate the impact by calculating only carbon dioxide emissions. UN scientists point out that the ecological footprint of the technology is three-dimensional and affects carbon emissions (energy required to run servers) as well as land use (land occupied by physical infrastructure) and water footprint (millions of liters of fresh water used every day to cool superprocessors). Isolating just one of these factors prevents us from understanding the true scope of the problem. The report concludes that the development of the sector needs global and political guidanceas it is necessary to define a “responsible AI ecosystem” based on transparency, efficiency, equity and environmental justice.
