The development of artificial intelligence (AI) technologies, such as ChatGPT, comes at a cost that extends beyond monetary expenses. One major factor that is often overlooked is the significant amount of water required to sustain these cutting-edge systems. Companies like Microsoft-backed OpenAI have been utilizing the waters of the Raccoon and Des Moines river basins in Iowa, USA to cool their powerful supercomputers, which facilitate the machine learning process for their AI systems.
The demand for AI tools and technologies from major tech developers like Microsoft, OpenAI, and Google is steadily rising, resulting in substantial costs ranging from expensive semiconductors to increased water consumption. However, these companies rarely disclose specific details about the environmental implications of their operations. For instance, few people were aware that the birthplace of OpenAI’s most advanced language model, GPT-4, was in the cornfields of western Des Moines, Iowa until a Microsoft executive revealed it in a speech.
Constructing large language models requires extensive computational power, which in turn consumes a significant amount of electricity and generates substantial heat. To keep data centers cool on hot days, water is pumped to external cooling towers. Consequently, Microsoft’s latest environmental report indicates that their global water consumption has increased by 34% between 2021 and 2022, totaling nearly 1.7 billion gallons. This significant rise can be attributed to their investments in AI, particularly in generative AI like ChatGPT.
According to researcher Shaolei Ren from the University of California, Riverside, ChatGPT consumes approximately 500ml of water for every series of 5 to 50 questions posed to it, considering both direct and indirect water consumption. Ren’s upcoming article estimates the environmental impact of generative AI systems like ChatGPT. These findings raise concerns about resource usage and the need for better awareness to promote resource preservation.
In addition to OpenAI, Google has also reported a 20% increase in water consumption during the same period, largely associated with its AI initiatives. Google’s water consumption remained stable in Oregon, where it gained public attention, but doubled outside of Las Vegas. Furthermore, Google’s data centers in Council Bluffs, Iowa, consume more potable water than any other location.
Both Microsoft and OpenAI have acknowledged the need for improvement in their energy and water usage. Microsoft stated that they are investing in research to measure the environmental impact of AI and working on means to enhance the efficiency of large-scale systems. OpenAI echoed these sentiments, emphasizing their commitment to improving processing power utilization.
As these technology giants continue to explore the potential of AI, it is essential for them to address the environmental implications. Efforts to reduce carbon footprint, increase renewable energy usage, and enhance water efficiency should be prioritized to achieve sustainable development goals.
Q: How much water does ChatGPT consume?
A: ChatGPT consumes approximately 500ml of water for every series of 5 to 50 questions, depending on server location and season.
Q: Why is water consumption a concern for AI technologies?
A: Water consumption is a concern for AI technologies due to the significant amount of water required to cool the powerful computing systems used in machine learning processes.
Q: How are major tech companies addressing the environmental impact of AI?
A: Tech companies like Microsoft and OpenAI are investing in research to measure the environmental impact of AI while working on making their systems more efficient and sustainable. They are also exploring renewable energy sources to power their data centers and reduce their carbon footprint.