As artificial intelligence systems continue expanding across the global economy, environmental researchers and analysts are warning that the AI boom may be creating an increasingly serious – and often overlooked – problem: water consumption.
Much of the public discussion surrounding AI infrastructure has focused on electricity demand and carbon emissions. But the giant data centers powering AI systems also require enormous quantities of water to cool servers and generate electricity.
An investigation published by Proof News argues that water usage linked to AI development is rising rapidly as major technology companies race to build larger and more powerful AI systems.
According to the report, Microsoft’s water consumption increased by 34 percent in 2022 compared with the previous year, while Google’s rose by 22 percent.
The article says Microsoft’s operations consumed enough water in 2023 to fill approximately 15.6 billion water bottles.
Researchers say the true scale of AI-related water use remains difficult to calculate because technology companies often disclose limited information about their data-center operations.
One estimate frequently cited in discussions about AI’s environmental impact suggests that interacting with AI chatbots such as ChatGPT may indirectly consume the equivalent of a bottle of water after dozens of user queries.
The report references research conducted by a team at the University of California, Riverside, which has studied data-center water use for years.
The researchers estimated that a bottle of water’s worth of water may be consumed for roughly every 10 to 70 AI queries, depending on how the data center is powered and cooled.
That figure includes:
- water used directly to cool servers; and
- water used indirectly in electricity generation.
The article notes that many estimates focus only on “water consumption” – water that is permanently lost or not returned to the system – while excluding much larger quantities temporarily used during cooling processes.
In regions already facing water shortages, even temporary diversion of water resources can become a significant issue, especially during heatwaves and droughts linked to climate change.
The report also argues that the AI industry faces a difficult tradeoff between renewable energy and water use.
Hot, dry regions often provide strong solar-energy potential but may face water scarcity, while cooler regions with greater water availability can require more energy generation.
Some countries and regions are already struggling to balance those competing pressures.
The article cites Ireland as an example of a location where data-center expansion has become so intense that authorities introduced restrictions on building additional facilities due to concerns over electricity and infrastructure demand.
At the same time, the global AI race is accelerating rapidly.
Technology companies are investing billions of dollars in new data centers as demand grows for AI systems capable of handling:
- large language models;
- image generation;
- cloud computing;
- automation; and
- real-time AI services.
The report argues that while many technology firms present AI as a tool that could help solve environmental problems, the infrastructure supporting those systems is itself becoming a major consumer of energy and natural resources.
“Big Tech’s energy and resource use is trending in the wrong direction,” the article states. “They’re using more, emitting more, and hoping AI magically solves all these problems. Instead, it’s making those problems worse.”
As governments and companies continue expanding AI infrastructure worldwide, analysts say debates over data-center electricity use may increasingly be accompanied by concerns about water availability, environmental sustainability and resource competition.
Leave a Reply