The massive use of artificial intelligence requires significant energy consumption, leading to increased greenhouse gas emissions, water use, and the depletion of mineral resources such as rare earth elements. A study conducted in January 2025 shows that data centers hosting A.I. consume 4,600 times more energy than those hosting search engines like Google. Other studies indicate that if usage continues at this rate, the amount of water required to operate data centers is estimated at between 4.2 and 6.6 billion cubic meters, roughly equivalent to half of the United Kingdom’s annual water consumption.

This excessive energy use is largely due to the fact that more and more users rely on A.I. several times a day. To counter this, solutions are being explored. For example, the Itrium 1 project by Totalinux in Jouy-en-Josas (Yvelines) aims to cool data centers by immersing them in a non-conductive oil that produces no noise, no CO₂ emissions, and no greenhouse gases.

However, the Secretary General of Data For Good, Lou Welgryn, states that it is essential to question the final use of these technologies. In other words, it is better to ask whether the request made to A.I. is truly necessary or whether it could be done independently.