A Sustainable Tech Challenge: Microsoft’s AI advancements cause a surprising rise in global water usage

Microsoft’s latest environmental report reveals a surprising 34% increase in global water consumption from 2021 to 2022, totaling almost 1.7 billion gallons. The increase is significantly higher than in previous years, and outside researchers attribute it to the company’s advancements in artificial intelligence (AI).

Researchers estimate that Microsoft’s AI model, ChatGPT, consumes about 500 milliliters of water, equivalent to a 16-ounce water bottle, every time a user poses 5 to 50 prompts or questions. This water consumption varies depending on the location of Microsoft’s servers and the season. Google has also reported a 20% increase in water use during the same period, with a similar link to its AI work.

Microsoft’s collaboration with San Francisco-based OpenAI began in 2019 when the tech giant invested $1 billion in the startup. This partnership aimed to provide the computing power required to train AI models. Microsoft’s data centers in West Des Moines, Iowa, have played a pivotal role in this endeavor. Microsoft’s “advanced AI supercomputing data center” in Iowa was built to facilitate the training of GPT-4, OpenAI’s fourth-generation AI model.

West Des Moines, Iowa, has proven to be an efficient location for AI model training, particularly when it comes to water consumption. Compared to Microsoft’s data centers in Arizona, Iowa consumes far less water for the same computing demand. However, the water required can still be substantial, especially during the summer. In July 2022, Microsoft used approximately 11.5 million gallons of water in its Iowa data centers, accounting for about 6% of the district’s total water usage.

The intersection of AI research and environmental impact is becoming increasingly evident, with water consumption emerging as a significant concern. Microsoft’s collaboration with OpenAI, while driving AI advancements, has also raised questions about the sustainability of such endeavors. Efforts are underway to find more water-efficient ways to power AI models, reflecting the broader challenge of reconciling technological progress with environmental responsibility.