ChatGPT queries use water of one almond
The rapid evolution of artificial intelligence, while reshaping our digital world, has brought an often-overlooked environmental cost: the staggering volume of water required to power the massive data centers that fuel it. As AI development accelerates, the conversation has turned to sustainability, leading to high-profile figures like OpenAI CEO Sam Altman facing scrutiny over water consumption.
In a recent discussion, Altman addressed the concerns surrounding AI data centers’ water use, suggesting that the issue has been blown out of proportion and does not reflect the latest advancements in cooling technology. He framed the debate by making a mind-boggling comparison: producing a single almond in California requires roughly the same amount of water as 38,000 ChatGPT queries. This stark figure attempts to put the abstract concept of AI usage into a tangible, earth-based context.
However, this comparison invites further scrutiny when examining the true metrics of both almond farming and data center operations. Growing almonds, one of the most water-intensive crops in the U.S., requires about 1.1 gallons of water per tree. While Altman’s estimate is dramatic, the actual water footprint of AI infrastructure reveals a complex picture.
Historically, training older AI models utilized systems that were significantly less efficient. For example, training GPT-3, for instance, reportedly consumed nearly 185,000 gallons of water using older systems. This underscores the necessity of technological evolution in how we handle digital infrastructure.
The true water consumption of data centers, however, is the critical focus. Large-scale facilities can use immense quantities of water. Reports have surfaced detailing consumption levels, such as one site in Fayette County, Georgia, which used 29 million gallons of water over just 15 months. This raises serious questions about the environmental accountability of this burgeoning digital infrastructure.
Fortunately, the industry is responding to this pressure with rapid innovation in cooling solutions. Companies are deploying advanced technologies that dramatically cut water consumption. For instance, Nvidia has introduced liquid cooling systems that operate “hotter than a hot tub,” which have demonstrated the potential to reduce water use by up to 100%. Similarly, Microsoft is implementing closed-loop cooling systems that mimic the efficiency of a restaurant, aiming to slash consumption from millions of gallons.
Even Amazon has made notable strides, claiming that its data centers consume only 0.075% of the water Americans use for watering their lawns and gardens, highlighting the pursuit of extreme water efficiency in massive operations. These advancements demonstrate that while the scale of AI infrastructure is vast, the commitment to sustainable engineering is steadily catching up to the technological demands.
Credit
