The Hidden Cost of AI: How Much Water Does ChatGPT Really Use?
In recent years, artificial intelligence has made huge strides, revolutionising the way we live and work. Tools like ChatGPT have become household names, offering instant answers, creative support, and business insights at our fingertips. But behind the magic lies a lesser-known environmental cost — water.
Yes, water.
Every time someone prompts ChatGPT, a small but significant amount of water is used to keep the infrastructure running smoothly. AI models such as ChatGPT are hosted on powerful data centres, many of which require advanced cooling systems to prevent servers from overheating. These systems often rely on evaporative cooling — a process that consumes vast quantities of water to maintain optimal temperatures.
So how much water does one interaction use?
According to a 2023 report from researchers at the University of California, it’s estimated that ChatGPT consumes around 500ml of water for every 20 to 50 questions asked. That’s roughly equivalent to a standard bottle of water for a single extended chat. Multiply that by millions of users worldwide, and the scale becomes apparent.
At a personal level, the impact may seem negligible. But when you consider the training phase of large AI models — which can require thousands of high-performance GPUs running for weeks or even months — the water usage becomes staggering. Training GPT-3, for example, was estimated to have consumed over 700,000 litres of fresh water, primarily to cool the servers during computation-heavy sessions.
Why water?
Cooling is essential to maintain the performance and longevity of data centre hardware. While some centres use air-based cooling systems, water-based methods are generally more energy-efficient and cost-effective — albeit with a significant environmental trade-off. In regions where water is scarce or where clean water is drawn from freshwater reserves, this can place undue stress on local ecosystems.
What does this mean for sustainability?
As businesses and individuals grow more dependent on AI tools, it’s important to be aware of the environmental footprint — not just in terms of electricity and carbon emissions, but also water use. While AI companies are taking steps to improve efficiency and shift to greener data centre models, users should also consider how they engage with these tools. Conscious usage, especially at scale, plays a role in managing demand and supporting sustainable innovation.
A global perspective
With climate change intensifying water scarcity in many parts of the world, the hidden water consumption behind digital tools becomes more than just a technical concern — it becomes a global environmental issue. As we continue to embrace AI, we must do so with eyes wide open, balancing innovation with responsibility, and ensuring that the invisible cost of progress doesn’t outweigh its promise.