Artificial intelligence in the water industry: friend or foe?
The artificial intelligence (AI) revolution is transforming many industries, and the water industry is no exception. AI-powered systems help automate water management processes, increase efficiency and reduce water safety risks. But there is also a downside to the rapid rise of this new technology: training and maintaining AI models requires huge amounts of potable water.
ChatGPT consumes a half-litre bottle of water for every standard conversation comprising between 5 and 30 questions.
Jonas Theyssens, Equity Analyst at KBC Asset Management
Growing trend towards going digital
Water has been a prominent item on the economic and political agenda for some time. Potable water is a scarce commodity. Water pollution is consistently underestimated. And climate change means that we are facing more frequent extreme weather events, which can lead to both flooding and water shortages. ‘AI can help us take control of the future of water. Market research is predicting investments to the tune of 6.3 billion USD in AI by the year 2030. These investments are part of a growing trend in the water industry towards going digital using smart solutions,’ said Anthony Sandra, Portfolio Manager at KBC Asset Management.
Three major focus areas for AI
Researchers worldwide use internet-connected sensors to gather data. The water industry then uses this data to address three key focus areas:
1. Leak detection
More than 15% of drinking water in Flanders is lost during the transportation in the network from production sites to end consumers, mainly through hidden leaks. ‘In total, we’re talking about more than 150 million litres per day, equivalent to the contents of more than 60 Olympic-sized swimming pools - and at a time when every drop of water counts,’ Sandra explains. Leak detection is therefore high on the priority list in the fight for water.
‘Closer to home,’ says Sandra, ‘IMEC has developed a leak-localisation system; the SmartWaterGrid project uses AI technology to detect the slightest deviation in water pressure using real-time data. This reduces the time taken to detect leaks from several weeks to a few hours, leading to a decrease in water loss and lower maintenance costs.’
AI can help us take control of the future of water. Market research is predicting investments to the tune of 6.3 billion USD in AI by the year 2030.
Anthony Sandra, Portfolio Manager at KBC Asset Management
2. Efficient water management
Having some buffer capacity is an essential element of sustainable water management, but is not always easy to achieve, and certainly not in an urban context.
‘Aquafin recently developed a project in Edegem, near Antwerp, to use private rainwater wells more dynamically through the deployment of AI,’ Sandra explains. ‘When heavy rain is forecast, AI proactively sends a signal to a specific pump to allow the water to seep into the garden in a controlled manner, thus freeing up space within the well. Rainwater can then be recaptured during heavy downpours and so help prevent flooding.’
3. Water quality
Manual checks are not only time-consuming, they are also really just snapshots of a moment in time. AI allows real-time monitoring of water quality and enables potential problems to be detected more quickly.
The other side of the coin
Clearly, then, AI can play an important role in the water industry; but there is also a downside to the rapid rise of this technology. ‘Training and maintaining AI models requires huge amounts of potable water,’ explains Jonas Theyssens, Equity Analyst at KBC Asset Management.
ChatGPT consumes a lot of water
Recent research by the Universities of California and Texas shows that language models like ChatGPT consume about half a litre of water for every five to 30 questions asked. ‘That environmental impact may not seem too bad on an individual basis, but because we use it massively and intensively - ChatGPT's website currently has more than 100 million active users - the overall impact is huge,’ says Theyssens.
The latest sustainability report from Microsoft, which is engaged in a key partnership with OpenAI, the company behind ChatGPT, showed that Microsoft's water consumption increased by more than 30% in 2022. Its major competitor Google, which has its own language model on the market, LaMDA, also saw its water consumption rise by 20%. Together, these two tech giants accounted for an additional consumption of 3.2 billion litres of water in one year. ‘That is the equivalent of about 1 300 Olympic pools,’ clarifies Theyssens.
The environmental impact of language models like ChatGPT may not seem too bad on an individual basis, but because we use them massively and intensively, the overall impact is huge.
Jonas Theyssens, Equity Analyst at KBC Asset Management
What is all that water needed for?
Computing processes that involve a lot of computing power, such as mining cryptocurrencies or training AI models, require huge data centres filled with servers which are processing data non-stop. To ensure that these servers do not overheat, they need to be cooled either using traditional air-conditioning systems, which is very expensive, or using direct cooling with water. Fresh water is pumped from a cooling tower and routed through the data centres via pipelines. To prevent clogging, rust and algae formation, only potable water is used.
Big tech companies' promise to be 'water-positive'
The environmental footprint of AI models, and especially the CO2 emissions, has led to advances in energy efficiency. Unfortunately, the huge water footprint of AI models has so far received too little attention.
‘With increasing pressure on CO2 emissions and rising energy costs, data centres are mostly being built in places where cheap solar and wind energy is available. 30% of the world's data centres are located in the Southwestern United States, where the climate is mild but where water has also become a scarce commodity,’ Theyssens explains. Many data centres struggle with the trade-off between using renewable energy and saving on water consumption.
If not properly addressed, this footprint will pose a major obstacle to the socially responsible and environmentally-friendly evolution of future AI models.
Jonas Theyssens, Equity Analyst at KBC Asset Management
‘It’s not as if the big tech companies aren’t doing anything to address the problem,’ adds Theyssens. ‘Many of them, such as Amazon, Meta, Google and Microsoft, have promised to become 'water-positive' by 2030, meaning they will replenish more water than they consume.’ Data centres can be made more energy-efficient centres so that they need less water for cooling. Using treated wastewater for cooling instead of pumping up potable water can also significantly reduce the environmental impact. On top of that, billions will be invested in restoring wetlands, building water basins and rainwater harvesting systems, as well as in infrastructure and wastewater treatment.
‘There’s a clear need for more transparency,’ Theyssens adds. ‘For example, more monitoring and proper analysis of water consumption in data centres. Developing guidelines for data centre sustainability is also becoming a must for
policymakers. Because if this footprint is not properly addressed, it poses a major obstacle to socially responsible and environmentally-friendly evolution of future AI models.’
Thematic investing gives investors an opportunity to invest in the key trends that shape our society.
Anthony Sandra, Portfolio Manager at KBC Asset Management
Smart water management offers opportunities for investors
In today's world, water resources are stretched to the limit. So technologies to enable smart water management are becoming increasingly important. AI can be invaluable here, provided the technology is used wisely and provided it gets its own environmental footprint under control. The big tech companies are already showing the necessary good intentions and also have the financial capacity to make the necessary investments.
‘Thematic investing gives investors an opportunity to invest in the key trends that shape our society. An active investment approach is key to selecting businesses that can create sustainable value in the long run,’ Sandra concludes.
Want to read more about responsible investing?
This article is informational only and should not be considered investment advice.
