As artificial intelligence (AI) continues to revolutionise various aspects of our lives, a pressing question arises: how much water is being consumed in the process? Recent discussions have highlighted the significant water usage associated with AI, leading to concerns about its environmental impact amidst an already strained global resource landscape. Let’s dive into this intriguing issue and explore the facts behind AI’s water consumption.
The Water Footprint of AI Technology
AI systems, such as ChatGPT and Google Gemini, operate from vast data centres that require a tremendous amount of water for cooling and energy generation. The energy-intensive nature of these facilities means that every interaction—be it a simple query or a complex command—demands a substantial amount of water to keep systems running smoothly. While estimates vary widely, they paint a troubling picture of AI’s growing thirst.
For instance, Sam Altman, CEO of OpenAI, asserts that a typical query to ChatGPT consumes less than 1/15 of a teaspoon of water. In contrast, a study from Google Gemini claims an average prompt uses only 0.3 ml. However, research from the University of California suggests a much higher figure, estimating that ChatGPT consumes around 500 ml of water for every 10 to 50 medium-length responses. Alarmingly, projections from the UK government’s Digital Sustainability Alliance indicate that AI’s water usage could surge from 1.1 billion to a staggering 6.6 billion cubic metres by 2027—equivalent to over half of the UK’s total water consumption.
Why Does AI Require So Much Water?
The primary reason AI systems are water-hungry is their reliance on data centres, which need cooling mechanisms to prevent overheating. These centres not only utilise water for cooling but also require it during the manufacturing of hardware and the generation of electricity. According to the Lincoln Institute of Conventional Policy, a mid-sized data centre can consume as much water as a small town, while larger facilities could match the daily water requirements of a city with a population of 50,000.
This escalating demand for water is worrisome, particularly in regions where water is already scarce. Nearly 68 per cent of data centres are situated near critical biodiversity zones, which depend on clean water supplies. As AI proliferates, the strain on local water resources intensifies, leading to an alarming potential for water scarcity and pollution. The Government Digital Sustainability Alliance warns that by the end of the decade, demand for fresh water could exceed supply by 40 per cent, exacerbating the already pressing issue of water stress in many areas.
Are These Concerns Justified?
The debate surrounding AI’s water consumption has sparked a mix of alarm and scepticism among experts. While some argue that the environmental ramifications are dire, others suggest the concerns may be exaggerated. Andy Masley, director of Effective Altruism DC, posits that the water used by AI per individual interaction is negligible compared to traditional water consumption. For example, producing a single pair of jeans requires around 7,500 litres of water—far more than the water needed for countless AI prompts.
Masley estimates that an individual would have to submit more than 1,000 prompts in a single day to see a meaningful increase in their daily water footprint. Interestingly, staying home and engaging with AI might contribute to a lower overall water footprint than the energy expenditure associated with going out.
However, Sam Gilbert, a researcher at the University of Cambridge, emphasises that the real issue lies not in individual usage but in the cumulative impact of data centres on local ecosystems and water supplies. He calls for greater transparency from companies regarding their environmental footprints.
On the other hand, sociologist Nick Couldry from the London School of Economics warns that the increasing demand for AI could lead to unsustainable pressures on water resources. He notes that regardless of current calculations, the dependence on AI will inevitably require more data processing, which could exacerbate environmental challenges.
Water Management and AI: A Double-Edged Sword
While the water usage of data centres raises valid concerns, there is also potential for AI to contribute positively to water conservation efforts. For instance, AI technology is already being leveraged to detect leaks and enhance the efficiency of water distribution systems. A water company in Surrey recently implemented AI solutions to significantly reduce leaks, demonstrating the technology’s capacity to aid in sustainable water management.
Google, known for its commitment to sustainability, claims its data centres are among the most efficient globally, employing air-cooling methods to limit water use. The tech giant also pledges to replenish 120 per cent of the water it consumes, highlighting a proactive approach to mitigating its environmental impact.
Why it Matters
The conversation around AI’s water consumption is crucial as it intersects with broader environmental issues. As AI technology continues to shape our future, understanding its resource demands becomes vital. With the potential for increased consumption and environmental strain, it’s imperative that both consumers and tech companies take responsibility. By promoting sustainable practices and investing in more efficient technologies, we can harness AI’s benefits while safeguarding our planet’s precious water resources for generations to come.