Generative AI and Machine Learning are increasingly being used to better manage water resources, with current applications in detecting leaks and reducing river pollution. However, Microsoft’s AI push saw the company’s global water usage increase 34% year on year in 2022, with Google seeing an increase of 20% in the same period.
With water scarcity becoming a significant environmental and social concern, and experts proposing that the water crisis be treated with the same urgency as climate change, there is frequent opposition to the development of new data centers on environmental grounds. As such, tech companies are currently facing challenges in addressing their water usage, and must be asked difficult questions about whether the ends justify the means.
The escalating water scarcity crisis
Climate change is exacerbating water scarcity around the world, leading to droughts, crop failures and wildfires. Within the EU, both the frequency and severity of droughts have risen, impacting nearly 20% more areas and people between 1976 and 2006, with the total costs of such droughts estimated at €100 billion. A recent report by the Global Commission on the Economics of Water predicts that by the end of the decade, demand for fresh water will outstrip supply by 40%.
Efficient water resource management and conservation is becoming an environmental imperative, with experts claiming that circularity of water systems is as important as the road to Net Zero. Efficient management not only helps secure a consistent water supply for human consumption, industry and agriculture, but also enhances the ability to withstand water shortages caused by climate fluctuations.
Additionally, conservation techniques contribute to the reduction of water wastage and ecological harm, safeguarding ecosystems and biodiversity. With the rising concern of water scarcity, responsible management becomes crucial in protecting communities, ecosystems, and economies as they confront the challenges posed by an unpredictable climate.
The role of AI in water management
There is great potential for the effective use of AI in water quality monitoring. In South-West England, a pilot project in Devon is using AI to predict and prevent water pollution. Connected sensors in rivers and fields can gather data on local environmental conditions, which are then combined with satellite imagery. This enables the real time identification of pollution incidents, from sources such as agricultural runoff and sewage discharges, prompting timely preventative measures. If successful, the project could have a significant impact in how water pollution is addressed, and ensuring cleaner water in marine environments.
Water management companies are beginning to adopt AI technology to detect leaks and efficiently manage waste, which could reduce costs and ensure the optimal utilisation of resources. Thames Water in the UK has begun to adopt AI technology for managing both its clean water and wastewater systems. Considering the widespread use of dowsing rods by water companies in the detection of leaks and pipes (an ancient pseudoscientific practise widely derided as witchcraft) the adoption of high tech AI detection solutions by UK utility providers will hopefully prove more effective in conservation and management of water resources in the long term.
The soaring water footprint of Big Data
Concurrently, tech giants like Microsoft, Google and OpenAI are under heightened scrutiny regarding their environmental impact as they strive to meet the surging demand for AI tools, particularly in the generative AI sector. Data centers, vital for AI processing, consume substantial amounts of water primarily for cooling purposes.
The training of AI models has led to a noticeable upswing in water consumption by tech giants, with both Microsoft and Google witnessing large year-on-year increases. Preliminary research suggests that interacting with AI models like ChatGPT or Google Bard through a series of questions consumes roughly 500 millilitres of water. This estimate doesn't encompass indirect water use, such as the water required for cooling the power plants that supply electricity to data centres.
While energy consumption in data centres is well documented, their water usage remains underexplored. Data centers directly consume water for cooling, in cases drawing more than half from potable sources, and indirectly through non-renewable electricity generation. Although data center water use is becoming an issue of greater significance for tech companies, transparency and measurement issues persist.
Considerable steps are being taken to try to limit the water used by data centres. Water use efficiency is a relatively new metric for measuring the impact of data centres on water resources, with a US Government report estimating that US based data centres use 1.8 litres per kilowatt hour. The Climate Neutral Data Center Pact (CNDCP) goes some way to addressing this challenge, with signatories pledging to reduce this number to 0.4 for their data centers on the European continent. According to the CNDCP:
The proposed limit of zero-point-four litres of water per Kilowatt-hour of computer power (0.4l/kWh) takes into account the diverse range of technologies, climates and types of data centre building to ensure that the metric is technology and location neutral.
The proliferation of data centers, while integral to technological advancement and the digital age, has encountered mounting opposition on environmental grounds, highlighting the complex conflict between technological progress and sustainability objectives. Environmental activist groups worldwide are mobilising to oppose the construction of the demonstrably water-intensive and polluting data centres, demanding transparency, sustainability, and democratic involvement in data center development. The pressure from these protests is challenging the tech industry to be accountable, and drawing attention to the physical infrastructure necessary to accommodate big data. With an increase in enterprise investment in data predicted in the near future, balancing the need for data-driven innovations with ecological responsibility, specifically water usage, is a pressing challenge.
As the world grapples with water scarcity and the impacts of relentless advances in technology, the debate over the impact of Big Data in water management is one of critical importance. Data center water usage is still under-measured and under-reported, and while big data and machine learning hold immense potential to address our water woes, the surging water footprints of tech giants demand scrutiny and accountability, as a matter of urgency.