**
As artificial intelligence continues its rapid ascent in both consumer and enterprise applications, an overlooked consequence of this technological boom is its significant water consumption. Recent analyses reveal that the data centres powering AI applications, such as ChatGPT and Google Gemini, are drawing on immense quantities of water, raising concerns about their impact on local ecosystems and water supplies.
The Water Footprint of AI Technologies
The crux of the issue lies in the operation of data centres, which require substantial cooling to maintain optimal performance. Each interaction with an AI model necessitates energy and water to prevent overheating. Estimates regarding the water used per AI query vary dramatically, with OpenAI’s CEO Sam Altman asserting that a single prompt consumes less than a fifteenth of a teaspoon, while a study from Google Gemini suggests it requires roughly 0.3 millilitres. However, research from the University of California presents a far more alarming figure, indicating that ChatGPT could consume around 500 millilitres for every 10 to 50 responses.
This discrepancy in figures fuels a larger debate about the sustainability of AI technologies. A report from the UK Government Digital Sustainability Alliance anticipates that by 2027, global water usage attributed to AI could soar from 1.1 billion to 6.6 billion cubic metres—equivalent to over half of the UK’s total water consumption.
The Environmental Impact of Data Centres
Data centres are not just power-hungry; they’re also water-intensive. A mid-sized data centre can consume as much water as a small town, while larger facilities may require up to 5 million gallons daily—comparable to the needs of a city with 50,000 residents. This raises critical questions about the sustainability of such operations, particularly in regions already grappling with water scarcity.
The Government Digital Sustainability Alliance has highlighted that nearly 68 per cent of data centres are located near protected ecosystems, which rely on clean water sources. As the demand for AI continues to escalate, concerns about water stress and scarcity grow, especially in areas where local communities are already vulnerable. The Alliance warns that demand for fresh water could exceed supply by 40 per cent by the decade’s end, with over half of global data centres situated in river basins at high risk of pollution.
Assessing the Crisis: Is AI’s Water Use Overstated?
While some experts argue that the water consumption of AI is a pressing global issue, others contend that the fears may be exaggerated. Andy Masley, director of Effective Altruism DC, posits that the water used for numerous AI prompts pales in comparison to everyday water consumption. He suggests that an individual would need to generate hundreds of thousands of prompts to equal the water footprint of a single pair of jeans, which typically requires about 7,500 litres to produce.
Masley’s perspective highlights the invisible nature of water footprints, which extend beyond direct usage to include the water consumed in the production and operation of technologies. Even so, he acknowledges the need for transparency regarding data centres’ environmental impacts.
Conversely, sociologist Nick Couldry from the London School of Economics cautions that regardless of individual water usage calculations, the broader implications of AI’s increasing reliance on data processing must be scrutinised. He argues that the business models driving AI encourage constant usage, which could lead to unsustainable environmental demands.
The Seasonal Pressure on Water Resources
Compounding these concerns, engineering professor Shaolei Ren of the University of California points out that many data centres experience peak water usage during summer months, exacerbating pressure on local water supplies. Thames Water has warned that restrictions could be imposed on data centres during heatwaves when water availability is particularly strained. Ren emphasises that while total water resources may appear sufficient, the distribution and timing of water availability are critical for effective management.
Can AI Provide Solutions?
Interestingly, the dual nature of AI may offer potential solutions to the very problems it creates. For instance, AI technologies are being deployed to improve water conservation through leak detection and enhancing water distribution efficiency. In 2024, a water company in Surrey implemented AI to mitigate leaks across its network. The World Economic Forum has reported that AI-enabled solutions in the United States could ultimately reduce water usage by up to 15 per cent.
Moreover, companies like Google claim to operate some of the most efficient data centres globally, using air-cooling systems to minimise water consumption. They assert a commitment to replenishing 120 per cent of the water they consume, striving to enhance local watershed health.
Why it Matters
The intersection of technology and environmental sustainability is increasingly critical as digital services continue to expand. As AI becomes more integral to various aspects of life and business, understanding its environmental footprint—especially in terms of water consumption—becomes essential. The potential for AI to both strain and conserve water resources underscores the urgent need for transparency and accountability from tech companies, alongside innovative solutions that address the growing demand for both data and water. Balancing technological advancement with environmental stewardship will be pivotal in ensuring a sustainable future for all.