THE HIDDEN COST OF CHATGPT: HOW AI'S GROWTH IS DRIVING WATER USE IN DATA CENTERS

The rapid rise of artificial intelligence tools such as ChatGPT has sparked growing discussions about their environmental impact, particularly the amount of water required to operate the massive data centers that power these technologies.

According to reporting by the Associated Press, Microsoft-backed OpenAI relied on a powerful supercomputer located in West Des Moines, Iowa, to help train GPT-4, the advanced language model behind ChatGPT. The computing process required enormous processing power, generating large amounts of heat that had to be controlled using cooling systems, many of which depend on water.

Training large AI models involves analyzing vast amounts of text and performing trillions of calculations. These intensive workloads require thousands of high-performance computer chips operating continuously inside data centers. To prevent overheating, many facilities circulate water through cooling systems, particularly during periods of high temperatures.

Microsoft reported that its global water consumption increased by 34 percent between 2021 and 2022, reaching nearly 1.7 billion gallons. Researchers have linked much of that increase to the company's expanding investment in artificial intelligence and its partnership with OpenAI, although Microsoft has not attributed the increase exclusively to AI.

Researchers are continuing to study the environmental footprint of generative AI. Shaolei Ren, a researcher at the University of California, Riverside, has noted that AI systems have measurable water and energy requirements because of the infrastructure needed to train and operate large language models. However, experts emphasize that there is no fixed amount of water used for every ChatGPT prompt, as consumption depends on several factors, including server location, weather conditions, cooling technology, electricity sources, and the complexity of each request.

The Associated Press also reported that Microsoft's Iowa data centers withdrew approximately 11.5 million gallons of water in July 2022, the same period when OpenAI was completing the training of GPT-4. Local officials have since encouraged Microsoft to adopt technologies that reduce peak water demand while allowing continued expansion of its facilities.

In response to environmental concerns, Microsoft said it is investing in research to better understand AI's energy, carbon, and water footprint while developing more efficient computing systems. The company has also reaffirmed its long-term sustainability goals, including becoming water positive, carbon negative, and zero waste by 2030.

OpenAI has likewise acknowledged that training advanced AI models requires significant computing resources and stated that it continues to explore ways to improve efficiency as demand for AI services grows.

As artificial intelligence becomes increasingly integrated into daily life, researchers say understanding the environmental costs behind AI infrastructure will be essential. While AI offers significant benefits across education, business, healthcare, and technology, experts stress that future innovation should be accompanied by greater transparency and continued efforts to reduce the environmental impact of the systems that make these tools possible.