China’s data centers vital for the country’s AI ambitions are guzzling water at a huge rate. According to a study by Hong Kong-based non-profit China Water Risk, these data could consume a 343 billion gallons of water annually, equivalent to the residential water use of 26 million people.
Also Read: Huawei Watch Fit 3: Apple Inspired Smartwatch
By 2030, this figure is projected to balloon to 792 billion gallons surpassing the water needs of the entire population of South Korea.
These data centers gulp down a 1300 billion liters of water annually, a figure equivalent to the residential water usage of approximately 26 million people.
The surge is even more alarming by 2030, this consumption is expected to skyrocket to a 3000 billion liters per year.
This is a disaster for the country’s water security with projections suggesting that China’s data centers alone could soon outpace the entire water consumption of South Korea.
The thirst of these data centers is primarily from the necessity to cool down the hardware essential for training and maintaining AI models.
As the AI boom intensifies, so does the demand for computational power, exacerbating the water consumption issue.
China’s plans to triple the number of data centers by 2030, reaching approximately 11 million data center racks.
While China leads the charge, it’s crucial to recognize that the AI boom is a global phenomenon with major players like the United States also contributing to the water consumption conundrum.
Companies like Microsoft and Google have disclosed figures regarding their water usage in AI-related activities.
Microsoft and OpenAI’s training of GPT-3 alone consumed approximately 700,000 liters of water. Google’s 2023 Environmental Report revealed a water consumption of 21.1 billion liters in 2022.
AI chatbots touted as the future of human-computer interaction come with their own set of challenges. These chatbots powered by specialized chips, exhibit a level of energy intensity.
For instance engaging with OpenAI’s ChatGPT on a massive scale could lead to water consumption equivalent to 20 Olympic-sized swimming pools.
Experts warn that the energy and water demands of AI are spiraling out of control with data centers projected to consume more electricity than entire nations by the end of the decade.
Rene Haas, CEO of Arm Holdings Plc, underlines the need for breakthroughs in energy-efficient chip technology to tackle these challenges.
The time for action is now, as the explosion of AI looms large on the horizon. Chinese ICT giants, urged to become “water neutral” or “water positive,” must adopt strategies such as watershed restoration, water reuse, and rainwater collection to minimize and offset their water footprint.
More than three-quarters of China’s data center racks are located in water-stressed regions, stringent regulations and water usage standards are imperative.
Also Read: Meta Launches Llama 3 AI Model
Holistic river management approaches, from source to sea, with tighter regulations can help mitigate the risks posed by excessive water consumption in the ICT sector.
China plans to triple its number of data centers by 2030, reaching approximately 11 million data center racks.
This expansion exacerbates the water consumption dilemma especially considering that nearly half of China’s data center racks are located in regions already struggling with water scarcity.
The AI boom has triggered a global surge in water usage, with tech giants in the United States also contributing to the problem.
Microsoft, in partnership with OpenAI, reportedly consumed a 185,000 gallons of water in training GPT-3 alone. Google, in its 2023 Environmental Report, confessed to using a 5.6 billion gallons of water in 2022.
The energy-intensive nature of AI chatbots further compounds the issue, with specialized chips devouring vast amounts of electricity.
China Water Risk’s report shows that engaging with AI chatbots, such as ChatGPT, consumes more water compared to traditional online searches.
Conversing with 100 million users via ChatGPT would require water equivalent to 20 Olympic swimming pools, while the same interaction via Google searches would consume only one swimming pool’s worth of water.
Experts warn that if left unchecked the energy and water demands of AI could soon surpass those of entire nations.
Arm Holdings Plc CEO Rene Haas cautioned that data centers worldwide could consume more electricity than India.
Also Read: Nothing is Integrating ChatGPT in Nothing Ear, Ear(a) Buds
)
