Artificial Intelligence is currently the most talked-about topic in the world. When we talk about AI, we usually imagine chatbots, smart apps or futuristic robots. But what keeps AI running? It doesn’t live in a laptop but lives in massive data centres. These buildings are filled with thousands of processors, storage units and cooling systems. To keep these systems running, data centres use massive amounts of water every day. This has created a massive environmental dilemma.
Every time we ask AI a question, water quietly leaves a river or reservoir somewhere in the world. Large data centres can consume up to 5 million gallons of water per day. This is equivalent to the water use of a town of 10,000 to 50,000 people. These data centres require vast amounts of electricity to run AI operations. Much of this electricity comes from fossil fuel power plants, which also need water for cooling.
According to the University of California, each 100-word AI prompt uses roughly one bottle of water, approximately 519 millilitres. Every time we type a prompt into ChatGPT, it’s processed by powerful servers that perform thousands of calculations to generate a response. That may seem small, but multiplied by billions of AI users worldwide, the numbers become staggering.
Training these AI models like ChatGPT or Gemini is even more resource-intensive. Because they consume more energy and water than that directly used for cooling.
Tech giants like Google, Microsoft and Meta reportedly used roughly 2.2 trillion litres globally in 2022 for power cooling. By 2027, global water withdrawal for AI data centres could reach 4.2-4.6 billion cubic metres.
Freshwater is a limited and vital resource. Only 3 per cent of Earth’s water is freshwater. And of that, only 0.5 per cent is accessible and safe for human consumption. According to Food and Water Watch, as of 2023, nearly 80 per cent of the water consumption of Google AI data centres in the US came from drinking water sources. Using this water for AI adds extra pressure on already scarce freshwater.
Overall, this also accelerates climate change and increases carbon emissions. Scientists say Antarctica is changing faster than expected. It’s raising the risk of higher sea levels and stronger climate disruptions. The continent that once helped cool the planet is now warming it. This shows how fragile earth’s systems have become. The rising water and energy demand of AI can add even more pressure to a planet already under stress.
In Dhaka, roughly 70 per cent of the municipal water supply still comes from deep groundwater extraction, according to WASA. The situation has worsened over the decades. In some densely populated areas, groundwater levels are now 60-70 metres below the surface. It was less than one metre in the 1970s. If a large data centre were to be built here, it could worsen the city’s freshwater crisis. The Buringnga river, which is already choked by pollution, could face further strain. Without careful planning, Dhaka’s communities and ecosystems would bear the cost of global AI growth.
To address this crisis, tech companies are investing in innovative and sustainable cooling technologies. Circular water solutions such as closed-loop cooling systems, wastewater recycling and rainwater harvesting can reduce freshwater usage by a huge percentage. Microsoft has recently started implementing adiabatic cooling. This uses outside air to cool servers when temperatures allow.
Despite the efforts, this remains a complex challenge to solve. Some tech firms have pledged to become “water positive”, meaning they will restore more water than they consume. But such promises remain hard to verify.
While the rise of AI promises incredible innovation and efficiency, it also comes with a great environmental cost. Every time a machine ‘thinks’ of us, it consumes not just energy but also the most essential resource for life: water. Ensuring that progress does not come at the expense of water security is essential for creating a sustainable digital future.
