Data centers across the globe are quickly becoming significant contributors to environmental concerns, with their energy usage rivaling that of some of the world’s largest nations. A report from the United Nations University reveals that the rise of artificial intelligence is expected to double their consumption of water and energy, as well as pollution levels, over the next four years.
In 2022, the energy consumption of global data centers reached 448 trillion watt-hours, a figure surpassed by only ten countries. This usage resulted in the emission of approximately 208 million tons of carbon dioxide, equivalent to the annual emissions of Argentina. Moreover, the energy production required to power these centers consumed about 1.2 trillion gallons of water, according to the study on AI’s environmental impact.
READ MORE: States are struggling to meet their clean energy goals. Blame data centers
Projections for 2030 indicate that data centers will account for nearly 3% of global electricity consumption, totaling 935 trillion watt-hours. If data centers were considered a country, they would rank sixth in terms of power usage. This would result in nearly 440 million tons of carbon dioxide emissions, as highlighted in the report which primarily focused on energy usage without delving into water usage for cooling purposes.
“If you look at these numbers, we’re seeing scales comparable to nations,” remarked Kaveh Madani, a water scientist and co-author of the study. “The demand is enormous.”
The rapid expansion of AI significantly contributes to the growth of data centers. Currently, AI accounts for approximately 20% of data centers’ energy usage, a figure expected to rise to 40% by 2030.
Global Perspective on Ecological Impact
The report carries weight due to the credibility of the United Nations, according to Fengqi You, a Cornell University professor specializing in AI sustainability. The study provides a comprehensive view of the carbon, water, land, and life-cycle impacts associated with AI, offering transparency on an issue often obscured by limited disclosures.
WATCH: Electric grid faces political roadblocks as it struggles with data center demand
“The general public should be concerned, but not panicked,” You advised. Meanwhile, Jean Su, an energy justice advocate, emphasized the significance of the report as the first global analysis to illuminate the environmental repercussions of AI.
Caleb Max, President of the National Artificial Intelligence Association, pointed out the growing efficiency of AI and its societal benefits: “AI is rapidly becoming part of our everyday lives and adding benefits that improve safety, live longer, work more efficiently, enhance food production, and reduce poverty.”
Josh Levi, president of the Data Center Coalition, affirmed the industry’s commitment to responsible growth: “We remain committed to working with policymakers, local communities, and industry partners to ensure that as data centers grow, they do so responsibly, transparently, and in ways that reflect the best available practices.”
Minimizing Energy Consumption
Madani, who also received the Stockholm Water Prize, highlighted the tangible environmental costs of AI. He pointed out that while AI might seem cleaner than devices like cars, its infrastructure involves significant energy usage.
WATCH: The growing environmental impact of AI data centers’ energy demands
Madani suggested that users could help reduce AI’s energy consumption by being more concise in their queries. The report noted that a 30% reduction in word count could lead to a 25% decrease in energy usage, saving an amount of electricity comparable to the annual consumption of 700,000 people in Africa.
“If you’re too polite, then that extra ‘please’ you put there can make a huge difference,” Madani explained. “You’ve got to be very precise and be short.”
AI-powered queries, such as those generated by ChatGPT, are much more energy-intensive than basic text classification tasks, like spam filtering. The complexity of AI systems correlates with increased energy demands for training and operation, as seen with GPT-3’s energy-intensive training process.
Miriam Aczel, another study co-author, noted that operational requests account for about 90% of AI’s power usage, with GPT alone handling 2.5 billion prompts daily.
Efficiency and Energy Use
Despite advancements in efficiency, there’s a paradox where increased efficiency leads to higher usage, resulting in overall higher energy consumption. Madani noted that when data centers use renewable energy, it can deplete clean electricity supplies, leading to increased reliance on fossil fuels elsewhere.
The study faced challenges due to a lack of transparency from companies about their data center operations and energy consumption, according to Aczel and Madani. “We cannot manage what companies do not disclose,” Cornell’s You stated.
Original Story at www.pbs.org