Generative AI is being blamed for a massive increase in the use of water for cooling in data centres.
The Financial Times this week reports that the world’s biggest tech companies have substantially increased their use of water to cool down data centres “triggering concern about the environmental impact of the generative artificial intelligence boom”.
The report says academic have warned that AI demand would drive up the “water withdrawal” to about half the amount consumed in the UK each year by 2027.
It says concern has grown in the past year as large language models consume vast amounts of energy to generate texts and other data types.
A paper by Interplex has also linked the concern about Generative AI to the operation of data centres.
“As these new applications are moving AI into the mainstream,the need for real-time, low-latency processing of large language model databases to provide on-the-fly responsiveness is having major impacts on data centres and networks. This means that existing data and infrastructures must undergo radical transformation and expansion to see the new AI demands,” Interplay warns.
In November last year, Nature magazine reported that ‘thirsty’ computing hubs could put pressure on already stretched water resources in sub-Saharan Africa.
“Yet water scarcity is rarely considered when deciding where to build data centres … typically companies care more about performance and cost.”
The big tech companies for their part have acknowledged they need to improve their game.
Microsoft’s global water consumption spiked 34 percent from 2021 to 2022 to more than 2,500 Olympic-sized swimming pools, says a report by AP. Microsoft in a blog says it will be water positive by 2030, replenishing more water than it uses.
Google has made a similar commitment. It reported it had consumed 5.6 billion gallons of water in 2022, the equivalent of 37 golf courses. It too has vowed to be “water positive” by 2030, as has Meta.
Scientists will watch to see whether the net water consumption figures increase or decrease as water replenishment projects come online yet generative AI use and water use intensifies.
There is some encouraging news at Mobile World Congress in Barcelona this week. Amazon Web Services (AWS) announced a two-year project with Greater Eastern Ranges (GER) to enhance the health and functionality of important parts of the major water catchment that supplies Greater Sydney.
“Once the project is completed, it is expected to deliver an additional 32 million litres of water each year to the Sydney catchment,” says AWS in a statement.
“This is AWS’s first water replenishment project in Australia, and a part of how AWS is working towards returning more water to communities than is used in its operations by 2030.”
AWS says the project will engage landholders, community groups and local government to improve, expand and reconnect habitats and manage key threats that suppress natural regeneration and reduce water quality, such as weeds, erosion and excess sediment.
“Weeds that spring up in the aftermath of fires, for example, prevent growth of biodiverse habitats that support a broader range of native species, and can consume large quantities of water, reducing the amount of water available to meet community needs.
“These areas were severely impacted by the Black Summer bushfires that swept through New South Wales in late 2019 and early 2020.”
The project will cover on-the-ground activity in Buxton, Picton and Thirlmere in the Wollondilly Shire area.
“As part of our water positive commitment, AWS is investing in water replenishment projects to expand community water access, availability, and quality by restoring watersheds and by bringing clean water, sanitation, and hygiene services to water-stressed communities,” says Jenna Leiner, Asia Pacific (APAC) sustainability lead for Amazon Web Services.
AWS has also announced water replenishment projects in India, Indonesia, Spain and the US.
Those concerned will be watching closely for project announcements by other multinational cloud providers and will be gauging their success.
It remains to be seen whether these replenishment projects will be effective enough to counter the likely exponential drain of water resources including potable water with data centres still heavily reliant on water for cooling.
Then there’s the ever increasing impact of climate change and resulting water scarcities that will make the situation worse. It’s the price we have to think about carefully as we embrace a Generative AI future.
On the microscopic level, you cause a data centre to gulp down a 500 ml bottle of water when you use ChatGPT to answer about 20 questions, say researchers at the Universities of Colorado Riverside and Texas at Arlington.
Published by ChannelNews Australia, February 28, 2024.