**
As artificial intelligence continues its relentless march forward, a troubling concern has emerged: the substantial water consumption associated with these technologies. A growing body of evidence suggests that the operation of AI systems, particularly those housed in data centres, may place an undue burden on already strained water resources. This article delves into the complexities surrounding AI’s water usage, contrasting estimates and perspectives from industry experts.
The Thirst of AI: How Data Centres Consume Water
Data centres are the backbone of modern AI applications, providing the necessary infrastructure to process vast amounts of data. This energy-intensive operation requires significant cooling, which, in turn, depends heavily on water. Estimates of water consumption vary widely across the industry. For instance, Sam Altman, CEO of OpenAI, claims that a typical interaction with ChatGPT consumes less than 1/15 of a teaspoon of water. Meanwhile, a study from Google Gemini suggests that each prompt requires less than 0.3ml of water.
However, these figures stand in stark contrast to research conducted by the University of California, which posits that ChatGPT’s activity could actually consume around 500ml of water for every 10 to 50 medium-length responses. Alarmingly, a report by the UK Government Digital Sustainability Alliance anticipates that AI-driven demand could escalate global water usage from 1.1 billion to 6.6 billion cubic metres by 2027—an amount that represents more than half of the UK’s total water consumption.
Growing Concerns Over Water Scarcity
The concerns surrounding AI’s water footprint are not merely theoretical. Critics argue that data centres, by tapping into local water supplies, exacerbate existing challenges for communities already grappling with water scarcity. Notably, nearly 68% of data centres are located near protected biodiversity areas, which rely on clean water. As demand for water continues to grow, experts warn that the anticipated shortfall could reach 40% by the decade’s end.
The Government Digital Sustainability Alliance highlights that over half of global data centres are situated in river basins at high risk of pollution, raising questions about the safety of local water supplies. As communities struggle to secure access to clean water, the expansion of AI technologies could lead to increasingly severe water stress.
The Debate: Are AI’s Water Needs Overstated?
The discourse around AI’s water consumption is rife with conflicting viewpoints. Some experts, like Andy Masley of Effective Altruism DC, suggest that the actual impact of individual AI usage is relatively minor. He argues that the water footprint associated with a single person’s use of AI pales in comparison to more traditional consumption patterns, such as the 7,500 litres needed to produce a single pair of jeans. Masley estimates that one would need to engage with ChatGPT for over a thousand prompts daily just to increase their water footprint by a mere 1%.
This perspective is echoed by Sam Gilbert from the University of Cambridge, who emphasises that the focus should not be on individual usage but rather on the cumulative effect of data centres on local ecosystems. Gilbert calls for greater transparency from tech companies regarding the environmental footprint of their operations.
Conversely, sociologist Nick Couldry of the London School of Economics warns that regardless of the calculations, the reliance on AI will inevitably lead to unsustainable demands on natural resources. He argues that as society increasingly integrates AI into daily life, the pressure on water supplies will only intensify.
Addressing the Challenge: Can AI Help Mitigate Water Issues?
While concerns about AI’s water consumption are valid, there are also opportunities for AI technologies to play a role in water conservation. Researchers like Shaolei Ren from the University of California point out that AI can enhance the efficiency of water distribution systems and identify leaks, potentially leading to significant water savings. For instance, a water company in Surrey has already begun using AI to reduce leaks across its network, showcasing the technology’s potential to address its own environmental challenges.
Google has also made strides in this area, with its Waltham Cross data centre employing air-cooling methods to minimise water usage. A company spokesperson highlighted their commitment to replenishing 120% of the water consumed, underscoring an effort to balance operational needs with environmental stewardship.
Why it Matters
The conversation around AI’s water consumption is an urgent reminder of the complex interplay between technological advancement and environmental sustainability. As AI continues to evolve, it is imperative for industry leaders and policymakers to address the inherent resource demands of these systems. A proactive approach could not only mitigate the potential negative impacts on water supplies but also harness AI’s capabilities to foster sustainable practices across sectors. With the stakes this high, the intersection of technology and ecology must remain at the forefront of discussions as we navigate the future of AI.