Youāll be amazed to learn this chilling fact: your favorite AI tool, ChatGPT, consumes a large amount of the world’s water supply. You may be hungry for knowledge, but your ChatGPT is thirsty for water that is four times more than previously thought.
Let’s take a moment and educate yourself about using technology and its consequences on the environment. According to a new study by a researcher, Shaolei Ren at the University of California, the water consumption required for 5-50 ChatGPT prompts is approximately 500 milliliters per interaction, which makes two liters of water. This hidden cost of AI can just be one of the scariest facts of the tech-driven era.
Truth Unleashed…
The OpenAI fetched water from the watershed of the Raccoon and Des Moines rivers in central Iowa. This water is used to cool powerful supercomputers as they trained their AI systems to mimic human writing.
Supercomputers use this water to cool down as they generate heat while consuming energy. The analysis and computation of human-written text provided by large language models such as ChatGPT consumes a significant amount of power. On hot days, data centers must pump water to a cooling tower outside their warehouse-sized structures.
An example of a data centreās operational water usage. Source: oecd.ai
In response to mounting concerns, Microsoft stated in a statement:
“We are strengthening how we manage water within Microsoft while also working to improve how the world assesses and utilizes water today and in the future.ā
The OpenAI acknowledged the water usage problem associated with this innovation and has committed to making LLMs more efficient.
In addition to Microsoft, Google is also contributing to water consumption for growing AI technology. As per Google’s environmental report of 2023, it used over 5.6 billion gallons of water in one year. This shows that the company has used 20% more water than previously as the company shifts towards more AI efforts. After analyzing this study, Google has planned to form a data center in Arizona that will work on “air-cooled technology” to mitigate the shortage of water.

Source: The Times
Shaolei Ren, a professor and researcher at Riverside, wrote that OpenAI provided the figure used for the original water consumption calculation. However, as per a new report by Microsoft, ChatGPT-3 and 4 will consume four times more water than they were using previously.
A recent OECD policy analysis found that AI’s water impact varies dramatically depending on where it is trained and housed. For example, AI consumes between 1.8 and 12 liters of water for each kWh of energy usage across all of Microsoft’s global data centers. Ireland proves to be the most water-efficient location, while Washington is the least.
Nvidia is far from absent in this problemā¦.
According to a Reuters report, Nvidiaās latest AI server racks generate 120 kW of energy and therefore produce a great amount of heat. David Craig, CEO of Iceotope, a British company that cools data centers, remarked that, āit is human-meltingly hot.ā
The study went on to show that an old three-bar heater is the equivalent of just 1 kW. āImagine a one-square-meter area crammed with 100 to 120 of such heaters. They are that hot, two meters high and one meter wide. Now consider thousands of these racks running within a single data center.”
Mitigation Moves by Tech Giants
Businesses are investigating data center cooling systems to combat the environmental effect of technology. Google has been trying to recycle wastewater, whereas Microsoft has run experiments using underwater data centers cooled by seawater. These initiatives seek to lower water consumption and help sustainability.
Both OpenAI and Microsoft are trying to develop more environmentally friendly systems. By 2030, they have ambitious objectives to reduce their carbon footprint and water use.
Ren encourages AI businesses to optimize their models and algorithms, as well as distribute workloads to more water-efficient places. Google DeepMind created an artificial intelligence system that assisted the corporation in reducing its cooling center energy usage by 40%.
On the way to sustainable artificial intelligence development, nevertheless, several fundamental issues arise: How do we weigh artificial intelligence’s great advantages against its environmental costs? What measures can be done to lower the water consumption of data centers? And how can we guarantee that AI evolves in a manner that supports long-term sustainability?
Discovering answers to these issues would be vital for building a more sustainable future as we push this technology.
For the latest technology news and insights, read our blogs on Inside Tech World.