AI Models and Data Centers Are Consuming Large Amounts of Water Alongside Energy
Many are aware that AI models consume vast amounts of energy during their operation, but what’s less known is that they also require a significant amount of clean water. According to Silicon Angle, data centers (DCs) powering AI systems are heavy water consumers, and tech giants like Google and Microsoft are seeing their water usage continue to rise.
A report by Associated Press reveals that Microsoft’s DC water usage increased by 34% from 2021 to 2022, totaling 6.4 billion liters in the past year. During the same period, Google’s water consumption rose by 20%. Experts attribute much of this increased consumption to the growing demands associated with AI workloads.
According to some estimates, OpenAI’s GPT-3 model, which powers chatbots like ChatGPT, is responsible for using over 320,000 liters of water during training. In fact, ChatGPT alone “consumes” approximately 0.5 liters of water per chat session consisting of 25-50 queries. These figures also account for indirect water usage, which includes the water used by power plants supplying DCs.
For Google, the statistics vary significantly from one DC to another, depending on location, season, cooling technology, and power station water consumption. However, many industry specialists emphasize that while the rise in water consumption is concerning, the surge in energy consumption is an even greater concern.
A significant portion of the water used is recycled and reused, unlike energy, which can’t be reclaimed. Furthermore, Microsoft, Google, and other DC operators are keen on reducing water consumption and will take necessary steps to achieve this goal.
Previously, Microsoft announced its intention to become “water-positive,” “carbon-negative,” and zero-waste by 2030, with a goal to reduce DC water consumption by 95% by 2024 compared to 2021. However, there are still no precise definitions for these terms, particularly “water-positive.” It is assumed that the company aims to return more water to nature than it consumes.
Pepsi serves as an example, redirecting water through its snack manufacturing equipment in the Mexico area. The water is purified to drinking water quality and sent to another facility for washing potatoes for chips. In this way, water indeed cycles more within the company than is initially drawn from natural sources.
OpenAI, a company with Microsoft as its largest investor, acknowledges the high levels of AI energy and water consumption and is actively working on improving efficiency in this area. This may involve developing more efficient algorithms and equipment for AI, although this will take some time to implement.
- I'm Vasyl Kolomiiets, a seasoned tech journalist regularly contributing to global publications. Having a profound background in information technologies, I seamlessly blended my technical expertise with my passion for writing, venturing into technology journalism. I've covered a wide range of topics including cutting-edge developments and their impacts on society, contributing to leading tech platforms.
- Troubleshooting15/11/2023Intel Fixes Critical Vulnerability Affecting All Processors – CVE-2023-23583
- Business15/11/2023Google Pays $8 Billion to Samsung for Default ‘Play Market’ and Search in Galaxy Devices
- Technology04/11/2023North Korea Upgrades Mobile Networks with Huawei Equipment Imports
- Technology03/11/2023Chinese Scientists Invent Passive Saltwater Cooler, Boosts CPU Speed by a Third