The high energy and water demands of AI are causing concern among experts, who warn that the increasing use of AI could outstrip global energy resources. Arm Holdings CEO Rene Haas suggests that one solution could be the development of more energy-efficient chips for training and powering AI models. The high water usage of data centers could also have severe impacts in areas where water is already scarce, such as Arizona, where Microsoft has been accused of covering up the water usage of its data center.
Key takeaways:
- China is increasing its focus on AI and opening new data centers, which are expected to consume around 343 billion gallons of water, equivalent to the residential water use of 26 million people, and could rise to 792 billion gallons by 2030.
- Training and maintaining AI models generate a lot of heat, requiring water to cool down the hardware. China could triple the number of data centers by 2030, leading to a significant increase in water usage.
- The AI boom is leading to a significant amount of water being used globally, including in the US. For instance, Microsoft consumed 185,000 gallons of water just in training GPT-3, and Google used 5.6 billion gallons of water in 2022.
- Experts are concerned about the rising energy and water usage due to AI. Arm Holdings Plc CEO Rene Haas suggests finding new ways to train and power AI models with more energy-efficient chips as a potential solution.