**
As the pace of artificial intelligence development accelerates, a pressing concern arises: the considerable amount of water required to power and cool the data centres that support these technologies. Recent analyses indicate that while the water consumption attributed to AI may seem minimal on an individual basis, its cumulative effect could significantly strain global water resources. This article delves into the nuances of AI’s water usage and the broader implications for environmental sustainability.
The Water Usage Debate
Artificial intelligence operates primarily through vast networks of data centres that require substantial amounts of water for cooling and power generation. Estimates regarding the water consumption associated with AI operations vary widely, sparking a debate among experts and stakeholders. For example, Sam Altman, CEO of OpenAI, has suggested that a single query to ChatGPT consumes less than one-fifteenth of a teaspoon of water. Meanwhile, a study by Google Gemini posits that the average prompt uses less than 0.3 millilitres.
However, contrasting assessments have emerged, with research from the University of California indicating that ChatGPT could use approximately 500 millilitres of water for every 10 to 50 medium-length responses. Alarmingly, a report by the UK Government Digital Sustainability Alliance forecasts that AI could escalate global water usage from 1.1 billion to 6.6 billion cubic metres by 2027, a figure that surpasses half of the UK’s total annual consumption.
Why Water Matters in Data Centres
The reliance of data centres on water is multifaceted. Not only do they utilise water to cool systems and prevent overheating, but they also depend on it for electricity generation and hardware manufacturing. The Lincoln Institute of Conventional Policy has highlighted that a mid-sized data centre consumes as much water as a small town, while larger facilities can draw up to five million gallons daily—comparable to the usage of a city housing 50,000 residents.
The implications of this water consumption extend beyond mere numbers. Local communities, particularly in regions already grappling with water scarcity, face increasing pressure as data centres tap into local water supplies. A working group from the Government Digital Sustainability Alliance has noted that nearly 68 per cent of data centres are located near protected biodiversity areas, which rely heavily on clean water sources. As demand for fresh water is projected to outstrip supply by 40 per cent by the decade’s end, the sustainability of such practices comes into question.
Perspectives on AI’s Water Footprint
Despite the growing concerns, opinions among experts vary regarding the severity of AI’s water consumption. Andy Masley, director of Effective Altruism DC, argues that the water footprint from individual AI interactions is negligible compared to conventional uses. He posits that hundreds of thousands of ChatGPT prompts would require less water than producing a single pair of jeans, which the UN estimates consumes around 7,500 litres. Masley further asserts that an individual would need to submit over 1,000 prompts daily to increase their water footprint by just 1 per cent.
Conversely, Sam Gilbert, a researcher at the University of Cambridge, stresses that the primary concern lies not with individual usage but with the cumulative impact of data centres on local ecosystems. Gilbert calls for greater transparency from companies regarding the environmental implications of their operations. He acknowledges that while estimates of ChatGPT’s water usage might be exaggerated, the overall context of water consumption remains critical.
The Future of AI and Water Sustainability
In light of these discussions, the question arises: can AI contribute to water conservation in other sectors? Shaolei Ren, an engineering professor at the University of California, points out that AI technologies can enhance water management by detecting leaks and optimising distribution systems. For instance, a water company in Surrey recently implemented AI to mitigate leaks, and the World Economic Forum anticipates that AI-enabled solutions could reduce water use by up to 15 per cent in the United States.
Notably, companies like Google are actively pursuing measures to improve their water efficiency. According to a spokesperson, Google’s data centre in Waltham Cross employs air-cooling to minimise water usage. The company claims to replenish 120 per cent of the water it consumes on average, reflecting a commitment to local watershed health.
Why it Matters
The intersection of artificial intelligence and water consumption poses significant challenges for sustainability. As AI technologies continue to proliferate, the demand for water will increase, potentially exacerbating existing environmental concerns. While some argue that AI’s water footprint is relatively small on an individual level, the collective impact of data centres warrants serious consideration. It is imperative that the tech industry takes proactive steps to mitigate its environmental impact, ensuring that the benefits of AI do not come at the expense of our planet’s vital resources.
