There is a silent detail behind the magic ofartificial intelligence that we all have in our pockets: while we ask ChatGPT to write a text or Midjourney to create a sublime image, somewhere in the world a data center roars. Thousands of servers heat up like turbines, sucking electricity and water to stay cold, tireless, with the same energy hunger as a small city. It is the new invisible industrialization, where steel and coal have been replaced by transistors and algorithms.
Artificial intelligence illuminates our everyday world, but behind every magical prompt lies a roar of servers that devour energy like a whole little nation. As we type a question, thousands of data centers pump electricity and water to keep their overheated hearts cool, revealing the paradox of “smart” technology draining the planet’s resources.
The International Energy Agency predicts that data center electricity consumption, driven by AI, will double by 2026, exceeding 1,000 TWh per year – the equivalent of Japan’s electricity needs. A single GPT-4 training session emitted between 12,000 and 15,000 tons of CO₂, thousands of times more than previous models. And it doesn’t end there: each query to the AI, like those to ChatGPT, absorbs around 0.3 Wh, which multiplied by billions of interactions becomes an enormous burden on the global network.
Data centers not only eat electricity, but they drink water for cooling. The study of University of California Riverside he calculates that training GPT-3 required approximately 700,000 gallons of water directly in Microsoft data centers, with total estimates rising into the millions when indirect effects are included. In a world with growing water crises, this “digital thirst” amplifies the problem, especially with expected AI expansion. Big Tech’s Greenwashing
Giants like Google and Microsoft proclaim carbon neutrality, but their emissions rise due to AI: Google has seen a +48% CO₂ since 2019, despite offsets. They invest in renewables, but demand grows faster than efficiency, creating an unsustainable race.
The physical impact of AI – mines for chips, plants for power, rivers for cooling – challenges us to rethink it. Computational power is not enough: we need an ethical approach that prioritizes efficiency and conscious consumption, making artificial intelligence truly a friend of our planet.
At Un Giorno Speciale we asked for information Mario Tozzigeologist and popularizer: the intervention in the video.
Related News :