Behind Every AI Answer… A Drop of Water

Behind Every AI Answer… A Drop of Water The rising water footprint of artificial intelligence and the urgent need for green intelligence.

“Green intelligence is not an ethical luxury; it is a condition for survival”.

With this statement, Tariq Al Hosani, Founder and Chairman of Zero Gravity Group, highlights one of the most pressing challenges emerging alongside the rapid rise of artificial intelligence: its hidden environmental cost.

Every time we ask artificial intelligence to write, summarize, or analyze information, a vast digital infrastructure immediately goes to work. Behind each interaction lies an immense network of data centers processing billions of requests every day. These machines do not operate without consequence. Each computational process generates heat that must be cooled instantly to prevent system failure.

Today, most data centers rely on water-based cooling systems. Water circulates through massive cooling towers where it absorbs heat and then evaporates into the atmosphere. The problem is that this water does not return to rivers, aquifers, or any water cycle that benefits human use—it simply disappears.

According to available estimates, 78 percent of the water withdrawn by major data centers is potable water. In other words, the same water that could supply homes or irrigate dry land is instead used to cool digital infrastructure.

Al Hosani explains: “Technology itself is not the problem. It is our only tool for survival, development, and prosperity. But intelligence that is not designed from the beginning with the logic of sustainability-what we may call green intelligence-may transform from a solution into a postponement of the crisis”.

Research by Professor Shaolei Ren of the University of California, Riverside estimates that a typical AI conversation involving 20 to 50 prompts consumes around 500 milliliters of water. Training a single artificial intelligence model from scratch can require about 700,000 liters of water.

The scale becomes clearer when considering global usage. ChatGPT alone processes at least one billion queries every day, while Google reported withdrawing 37 billion liters of water in 2024, of which 29 billion liters evaporated permanently.

The environmental footprint extends even further. Manufacturing a single smartphone requires 12,670 liters of water, while a Bitcoin transaction consumes roughly 16,000 liters. For Al Hosani, the real challenge is no longer technological performance alone. “The question today is not how to make intelligence faster or smarter,” he says. “The real question is how to make it consume less of what cannot be replaced”.

Solutions already exist, including closed-loop cooling systems, immersion cooling technologies, and more efficient chips. In 2024, Microsoft introduced a data-center design that requires zero water for cooling. Still, Al Hosani believes transparency must come first:“What cannot be measured cannot be managed. We need a global index for the water footprint of every AI model”.

Every time we press the question button, somewhere a drop of water pays the price.