Is Chat GPT A Guzzler? Understanding AI Tools’ Water Consumption

In the pursuit of groundbreaking AI innovations, Microsoft and OpenAI nearly triggered a potable water crisis in the US State of Iowa. But there’s more.

The one that started the trend, ChatGPT apparently consumes some half a litre of water, every time a user engages it with a series of 5 to 50 prompts.

Essentially, the price that humanity and the environment must pay for generative AI platforms to succeed is getting more and more massive every day.

The tech race to tap into the generative AI frenzy is getting more prominent amid firms often remaining tight-lipped about the precise environmental toll of their creations.

Iowa’s Role In The Creation Of GPT-4

According to a new report, Microsoft and OpenAI squeezed Iowa’s Des Moines and Raccoon dry to cool their supercomputer, which was essential for training the GPT AI models.

Very few people were aware of Iowa’s role in the creation of GPT-4, until a Microsoft executive revealed it was “literally made next to cornfields west of Des Moines.”

The creation of a large language model is known to involve intensive computing, which relies on considerable electricity usage and generates a substantial amount of heat.

Data centres frequently use water, often funnelled to external cooling towers near their facilities, to maintain optimal operating temperatures.

Sizeable Resource Usage Underlying ChatGPT

Compared to Microsoft’s water-intensive data centres in Arizona, West Des Moines offers amiable conditions for AI model training.

Iowa’s climate often allows the use of outside air for cooling the supercomputer. Only when the temperature spikes beyond 29.3 degrees Celsius, the facility relies on water cooling.

Nonetheless, the quantity of water consumed remains substantial, especially during the hot summer months made more unbearable by the climate emergency.

Microsoft’s global water consumption surged by 34% from 2021 to 2022, reaching nearly 1.7 billion gallons, according to its latest environmental report.

The enormous spike is largely attributed to the tech giant’s prominent AI endeavours. It appears most people are still not aware of the sizeable resource usage underlying ChatGPT.