Real-time water consumption of artificial intelligence worldwide
Every 100 words generated by AI consumes about 500ml of water.
This estimation is based on the average processing of 500 billion tokens per day by AI systems
like
ChatGPT, Google Gemini, Meta Llama, and others.
THE INVISIBLE COST OF PROGRESS
This page does not aim to criticize the water consumption caused by artificial intelligence.
The purpose of this project is to bring to the table, in a curious, objective, and conscious way, a
reflection on the impacts—visible and invisible—that accompany technological progress.
AI is reshaping markets, redefining roles, transforming education, healthcare, science, and society as a
whole.
No advancement happens without impact. Whether in the adaptation of social and economic systems,
cultural transformations, or the consumption of resources—often unnoticed.
Discussing the impacts of artificial intelligence is not about slowing progress. It's about paving its
path in a conscious and informed.
After all, we cannot stop progress. And progress isn't free.
HOW IT WORKS
This counter represents a global estimate of the water consumption associated with the use of generative
artificial intelligence.
The estimate is based on recent academic studies, including projections for 2027 suggesting that global
AI-related water use could reach up to 6.6 trillion liters per year.
For visualization purposes, we adopted a rate of 158,000 liters per second, based
on the average of these academic projections. Since January 1, 2025, the counter has been accumulating
in real-time the estimated total amount of water consumed by global AI systems.
These figures are consistent with the recent growth recorded by major companies like Google, Microsoft,
OpenAI, and Meta, whose data centers already consume billions of liters of water annually.
FAQ
Are these numbers accurate?
→ These figures are rigorous estimates, based on recent academic studies
and sustainability reports from the world's leading technology companies. While methodological
variations exist among sources, the order of magnitude is widely validated by experts and
consistent with the rapid growth of generative AI.
Where do these numbers come from?
→ The data is based on a structured research effort conducted using deep
AI-powered search mechanisms, with results available in this report. The material compiles water
usage reports from companies such as Google, Microsoft, and Meta, as well as academic
projections on both direct consumption (data center cooling) and indirect usage (power
generation required to operate large-scale AI models).
Is this project against the use of AI?
→ No. This project has no critical or ideological stance. Our goal is to
spark intelligent, objective, and conscious reflection on the real environmental impacts of
accelerated technological progress.
Does AI really use that much water?
→ Yes. Large AI models require significant volumes of water — both for
cooling the data centers where they run, and for generating the electricity needed to train and
operate them. The global expansion of generative AI has led to a substantial increase in
resource consumption, already measured in billions of liters per year, with projections reaching
into the trillions.