Data Centers Doubling as Water Guzzlers Cooling Off AI

    Published on:

    As generative AI grows in popularity, the water consumption of this technology is also skyrocketing.

    This technology is energy intensive and also consumes large amounts of water to cool the generation AI servers. This is despite the fact that AI is being touted as a potential solution to reducing greenhouse gas emissions and the negative effects of climate change.


    According to SiliconANGLE reportMicrosoft reported that water consumption in data centers jumped 35% from 2021 to 2022.

    In other words, the company used 1.7 billion gallons of water, or more than 6.4 billion liters, in 2022 alone. This is enough to fill more than 2,500 Olympic-sized swimming pools.

    Google, another major generative AI company, also reported a significant increase in water consumption in its data centers during the same period.

    Ann environmental report According to the graph, the amount of water consumed by Google increased by 20% in 2022 compared to the previous year, but the rate of increase was uneven across data centers.

    According to SiliconANGLE, the company’s Oregon facility has previously attracted negative headline, We have had steady water usage for the past year.

    But its Council Bluffs data center used more potable water than any other data center.

    Also read: German military tests AI weapons system in Metaverse

    Water consumption will increase due to the spread of AI

    Experts attribute this increase to increased usage of generative AI popularized by OpenAI after the launch of ChatGPT in November 2022.

    “Most people are not aware of the underlying resource usage of ChatGPT,” California scientist Xiaolei Ren told The Associated Press. Ren has conducted research to assess the impact of generative AI on the environment.

    “If we don’t know how resources are being used, we can’t contribute to resource conservation,” the scientist said.

    In a paper to be published later this year, Ren estimates that OpenAI’s GPT-3 gulped down about 85,000 gallons of water during training. In the same paper, Ren states that ChatGPT can consume about 500 milliliters of water for each discussion with about 25 to 50 questions.

    Various factors are at work

    Water usage for data center cooling depends on many factors, including location, season, and the type of technology used in a particular data center.

    While the huge water consumption may alarm environmentalists, Holger Müller, an analyst at Constellation Research, told Silicon Angles that the more worrying issue is energy consumption.

    “While much of the water used in data centers is recycled, even by nature itself, the same cannot be said for energy,” he said.

    “Also, note that while there is a lot of attention on cloud data centers, cloud data centers typically have higher utilization rates and outperform on-premises facilities in terms of efficiency.”Additionally, Microsoft and Google both have commercial incentives to reduce energy and water usage.

    The data center doubles as a watering hole to cool the generative AI servers

    Companies work to solve problems

    When asked about their huge water consumption, Microsoft and OpenAI said they are aware of the challenge and are looking into reducing it.

    Microsoft outlined its plans for the next seven years. The company has pledged to become even more “water positive,” carbon negative and zero waste by 2030.

    OpenAI also promised to improve water usage efficiency. Part of the solution involves reducing energy requirements, combined with efficient algorithms and hardware.

    Another solution involves locating data centers at sea. This is an initiative that Microsoft is working on. Implemented in 2018, it has been reported that energy efficiency has improved. The tech giant also suggested that underwater data centers do not mean the need for cooling systems.


    Leave a Reply

    Please enter your comment!
    Please enter your name here