Report: Thanks to AI, China’s Data Centers Will Drink More Water Than All of South Korea by 2030

🌈 Abstract

The article discusses the growing water consumption of data centers and AI models, particularly in China, and the potential environmental impact of this trend.

🙋 Q&A

[01] The Water Consumption of Data Centers and AI

1. What are the key facts about the water consumption of data centers and AI in China?

• China's data centers could soon be consuming around 343 billion gallons of water, equivalent to the residential water use of 26 million people.
• By 2030, this number could rise to 792 billion gallons, enough to cover the needs of the entire population of South Korea.
• China could triple the number of data centers by 2030, reaching roughly 11 million data center racks.

2. How does the water consumption of AI compare to traditional online searches?

• If 100 million users were to chat with OpenAI's ChatGPT, it would consume the equivalent of 20 Olympic swimming pools.
• Doing the same via simple Google searches would only "consume one swimming pool."

3. What are the potential environmental impacts of this water consumption?

• This water usage could have devastating effects on parts of the world where water resources are already extremely scarce, such as the Arizona desert where Microsoft has a data center.

[02] The Energy Consumption of Data Centers and AI

1. What are the concerns about the energy consumption of data centers and AI?

• Arm Holdings Plc CEO Rene Haas expects the world's data centers to use more electricity than India, the world's most populous nation, by the end of the decade.
• Training and maintaining AI models is an infamously energy-demanding task that generates a huge amount of heat, requiring water to cool down the hardware.

2. What are the potential solutions to address the energy and water consumption issues?

• Experts suggest finding new ways to train and power these AI models with more energy-efficient chips as a meaningful step forward.
• Any piece of efficiency, such as improving the energy efficiency of the hardware, could make a difference.