Artificial intelligence is presented as the technology of the future, but its infrastructure runs on electricity, water and land of the present. An analysis by Rebelión Ecologista focuses on the footprint already left by data centers that train models and serve AI queries on a global scale.
The question is uncomfortable and concrete: what environmental cost does the infrastructure that sustains AI tools pay while their growth is announced as inevitable? Because every query to a chatbot, every image generation and every model training runs on servers that drink megawatts.
You may also likeONU reclama a grandes tecnológicas asuman costes ambientales de la IAA report by the International Energy Agency (IEA, 2024) estimates that data centers consumed about 1.5% of global electricity in 2023 — about 460 TWh — and projects that figure could double by 2026, driven mainly by AI and crypto mining workloads.
Water is the other major vector. A study published in *Environmental Research Letters* (Li et al., 2023) calculated that training GPT-3 consumed around 700,000 liters of water for cooling in Microsoft centers. A query to ChatGPT shows between 0.5 and 2 liters depending on the region and type of server.
There is also a less visible effect: the territorial footprint. Hyper-data centers require huge extensions close to cheap energy sources. In the United States, Virginia – "Data Center Alley" – already concentrates 70% of the world's Internet traffic and faces conflicts over electricity demand, noise and residual thermal.
On the Latin American side, the debate is just beginning. Chile, Brazil, and Mexico are receiving hyperscaler projects leveraging renewable energy, but water and electricity impact measurement frameworks remain opaque. The expansion of AI requires environmental auditing now, not when it is too late.
You may also likeAI will double data center power consumption by 2030Sources: [Rebelión Ecologista — original article](https://rebelionecologista.com/). Global electricity consumption data: [IEA — Data Centres & Data Transmission Networks, 2024](https://www.iea.org/reports/data-centres-and-data-transmission-networks). AI water footprint: Li et al., "Making AI Less Thirsty", *Environmental Research Letters*, 2023. Image: archive.




