Artificial Intelligence (AI) has undeniably transformed our world, but at what cost? A recent revelation underscores an often overlooked aspect of AI development: its environmental impact, particularly on water consumption. As tech giants such as Microsoft, OpenAI, and Google race to capitalize on the craze for generative AI, they are grappling with the escalating costs of their AI tools, including the need for vast quantities of water to cool powerful supercomputers, along with the price of expensive semiconductors. The thirst for AI is not only financial; it is also literal, as these companies are drawing water from the watersheds of rivers to cool their AI systems.
Beyond the financial aspects, the secrecy surrounding these operations adds another layer of complexity. Take, for instance, the birth of OpenAI’s advanced large language model, GPT-4. Few in its birthplace of Iowa knew about its existence until a Microsoft executive announced that it was “literally made next to cornfields west of Des Moines.” These models require analysis of patterns across a massive trove of human-written text, a process that consumes significant electricity and generates ample heat. To keep the systems cool, tech companies pump water, often from nearby rivers, into cooling towers outside their warehouse-sized buildings. The environmental toll of AI development is becoming increasingly apparent, yet remains largely unaddressed by the industry.
AI Development: A Thirsty Business
As the race to capitalize on generative artificial intelligence (AI) intensifies, tech leaders, including Microsoft, OpenAI, and Google, face an escalating dilemma. Their AI tools carry a hefty price tag, not just in terms of expensive semiconductors, but also in water consumption. These AI products require an abundance of water to cool powerful supercomputers used to teach AI systems like ChatGPT to imitate human writing.
The Hidden Cost of AI
In the heart of central Iowa, an area more recognized for its cornfields than its contribution to cutting-edge technology, lies the birthplace of OpenAI’s advanced large language model, GPT-4. Despite its essential role in AI development, few locals were aware of this until a Microsoft executive revealed the fact in a speech.
Building such a large language model necessitates the analysis of patterns across an enormous trove of human-written text. This process demands a significant amount of electricity, producing a great deal of heat that needs to be cooled. Data centers, often warehouse-sized buildings, pump water into cooling towers from local watersheds to keep the system cool, particularly on hot days.
The Environmental Impact
Microsoft’s latest environmental report disclosed a 34% surge in global water consumption from 2021 to 2022. The increase is attributed by outside researchers to the tech giant’s AI research. Shaolei Ren, a researcher at the University of California, Riverside, explains that the majority of this growth is due to AI, particularly Microsoft’s partnership with OpenAI.
Google also reported a 20% increase in water use during the same period, which Ren largely attributes to its AI work. Despite the rise in consumption, Google has managed to maintain steady water usage in Oregon, where public attention to its water use is high. However, in Iowa and outside Las Vegas, Google’s water use has doubled.
The Future of AI Development
In response to these findings, Microsoft and OpenAI have committed to addressing the environmental impact of AI development. They are investing in research to measure AI’s energy and carbon footprint and exploring ways to make their systems more efficient.
Microsoft’s partnership with OpenAI began with a $1 billion investment in the San Francisco-based startup in 2019. As part of the deal, Microsoft provides the computing power needed to train the AI models. To meet this need, the companies turned to West Des Moines, Iowa, where Microsoft has been accumulating data centers for over a decade.
Microsoft’s president, Brad Smith, revealed in May that the company built an "advanced AI supercomputing data center" in Iowa. This facility, nestled among the cornfields, enabled OpenAI to train its fourth-generation model, GPT-4, which powers premium versions of ChatGPT and some of Microsoft’s own products.
Takeaways
The development of AI technology comes with a significant environmental cost, particularly in terms of water consumption. As the demand for AI tools grows, tech companies must address these environmental impacts and seek more sustainable methods of development. The future of AI should not only be measured by its technological advancements but also by its sustainability.
