The Footprint of Generative AI: Climate Change's Silent Catalyst

Ivy Gu

Mar 33 min read

It’s been roughly two years since ChatGPT was released to the public, becoming a valuable tool for learning, creativity, and innovation across many different industries. However, as artificial intelligence continues to modernize the 21st century, its silent contribution toward climate change has become increasingly concerning.

The Hidden Environmental Cost of Generative AI

Data centers supporting generative AI consume massive amounts of energy that is driven by millions of users who rely on these platforms daily. These systems consist of Graphics Processing Units (GPUs) which generate vast amounts of heat when powering AI. To prevent overheating, these centers rely on water-based cooling systems, also known as “cooling towers.”

The environmental consequences are already significant as Microsoft used 700,000 liters of water to train GPT-3 alone. When these AI models are put to use, chatbots use massive amounts of water to respond to simple questions. A study by the University of California, Riverside, in collaboration with The Washington Post, found that generating a 100-word email with ChatGPT uses approximately 519 milliliters of water—the equivalent of a standard water bottle. With millions of daily users on platforms like ChatGPT, the demand for cooling solutions escalates, further increasing consumption of energy and resources.

In North America, data center energy demands have risen. Scientists estimate that in just one year, energy consumption of data centers in the region grew from 2,688 megawatts to 5,341 megawatts. As a result, tech companies are ramping up efforts to build new data centers. “They’re building them as fast as they can,” says Steve Gaer, the mayor of West Des Moines, Iowa where data center production has surged. According to the West Des Moines Water Works, Microsoft reported using approximately 11.5 million gallons of water to cool its data centers in Iowa in July 2022, just before the completion of GPT-4’s training.

The situation is only getting worse as generative AI continues to innovate. Google’s 2023 Environmental Report, stated a 20 percent increase in water usage, largely tied to the rising demand for AI. Despite multiple efforts by companies to address sustainability, they continue to face challenges in meeting climate goals. For example, Google restored only 18 percent of the water it consumed in 2022, far below the goal of 120 percent by 2030.

Much of the problem lies in the fact that everyday users of generative AI are unaware of its environmental impact. “An everyday user doesn’t think too much about that,” says Noman Bashir, a Computing and Climate Impact Fellow at the MIT Climate and Sustainability Consortium. Despite growing concerns, engagement with AI platforms continues to rise, and tech companies are under pressure to innovate. Generative AI models have exceptionally short lifespans, due to companies releasing new versions every few weeks to meet the growing demand for AI applications. As a result, the energy used to train older models is often wasted, adding to the environmental cost of this evolving technology.

Innovation and Sustainable AI

By 2030, data center facilities could consume up to 21 percent of the electricity supply worldwide and water usage has been escalating as 1.1 billion people worldwide lack access. In response, Google recently stated, “Fresh, clean water is one of the most precious resources on Earth, we’re taking urgent action to support water security and healthy ecosystems.” Major tech companies like Microsoft have also come forward to invest in alternative energy sources like solar and wind to reduce their carbon footprint. Microsoft furthered that it has been investing in cooling technologies designed to end water consumption completely in its data centers. However, achieving these goals remains uncertain as critics question whether these commitments are realistic.