Can AI Go Green?
This blog post is abstracted from research conducted by James Landis, a DevryBV Summer Intern ‘25 .
From chatbots to predictive tools, artificial intelligence (AI) has rapidly woven itself into the fabric of our daily lives. With around 8% of Americans already using a tool like ChatGPT as their go-to search engine, the pace of AI adoption is surging. But as we embrace the convenience and productivity gains, many experts are pausing to ask a critical question: Is our pursuit of AI innovation coming at too high a price for our planet and society?
The question isn't about halting progress. It's about becoming a "sustainable steward" of AI—a concept that means maximizing the benefits while methodically anticipating the environmental, social, and governance (ESG) risks. The reality is that the AI boom has serious ESG implications we can’t afford to ignore. Acknowledging these risks is the first step toward finding solutions to make AI sustainable.
The Environmental Price Tag of AI.
The “cloud” isn’t ethereal. Technology is powered by big, physical data centers that consume large amounts of energy, water, and rare metals.
A carbon-heavy footprint: Training a single large AI model can produce around 626,000 pounds of carbon dioxide, which is roughly the same as 300 round-trip flights between San Francisco and New York. According to the U.S. Department of Energy, data centers account for 4 percent of total U.S. electricity consumption. This figure is projected to grow to 12% by 2028.
A thirst for water: Data centers rely on millions of gallons of water daily for cooling. By 2027, the global demand for AI is projected to require 1.1 to 1.7 trillion gallons of water annually, which is about half of the United Kingdom’s yearly consumption.
The e-waste problem: The computer chips at the heart of AI systems have a short lifespan of just three to five years, creating a continuous stream of electronic waste. In 2022, the world generated 62 million metric tons of e-waste, and only about 22% of it was recycled.
This isn't just about big tech; it's about us. Nearly one in five American adults uses AI daily, often unaware of how seemingly simple actions—like a quick search or asking a question—contribute to this resource-intensive cycle.
The Social and Ethical Costs: Bias and Sycophancy.
Beyond the environmental impact, AI poses a significant social concern: bias. AI systems are trained on vast datasets, and if that data is skewed, the AI's user responses become susceptible to bias and false information. Large language models (LLMs), which power tools like ChatGPT, have also come under criticism for being "skewed towards Western views" and performing best in English.
Another major issue is sycophancy, the tendency for an AI to mirror a user's opinion to appear more agreeable and "helpful." This is influenced by a process called Reinforcement Learning from Human Feedback (RLHF), which trains models to align with user preferences.
For example, researchers found that when they prompted an AI with a description of a politically conservative user and asked a question about government size, the AI’s response favored a smaller government. When the user was described as liberal, the AI’s answer switched to support a larger government.
This isn't just an academic issue. In one case, a man on the autism spectrum was convinced by ChatGPT that he could bend time, leading to a 17-day stay in a mental health facility. When AI prioritizes agreeability over factual accuracy, it can reinforce existing biases and even pose a risk to vulnerable individuals.
Moving Toward a Sustainable AI Future.
The challenges of AI’s growth may seem overwhelming, but they aren't insurmountable. It’s possible to address these issues through an ESG framework, which can be used to evaluate an industry’s effectiveness in managing its environmental, social, and governance impacts.
Here are some ways we can work toward sustainable AI:
Expand renewable energy. Currently, 56% of data center electricity comes from fossil fuels. We need to transition to cleaner sources. Interestingly, AI can assist with this by helping to optimize electricity grids and improve energy efficiency, providing a potential solution to the very problem it creates.
Use biomimetics in designing data centers. Some companies are looking to nature in order to address the data center's carbon footprint. Microsoft, for example, is focused on creating data centers considering high productivity, water stewardship, biodiversity enhancement, and prioritizing waste reduction, heat reuse, and ecosystem integration. Is a net-zero data center possible? Scientists and engineers are on the case.
Regulate for responsibility. The AI supply chain is complex, from chip manufacturing to daily data center operations. Regulations and reporting standards should be put in place in order to mandate transparency in AI. An example of this can be seen in the European Union's new AI Act, which is promoting transparency and safety with data in high-stakes industries such as healthcare and employment. While some argue regulation stifles innovation, a well-designed regulatory framework can protect against harm while allowing AI the space to grow.
Prioritize assistance, not replacement. Instead of designing AI to replace human jobs, we should focus on creating tools that assist people, making tasks easier and more efficient. Historically, every technological revolution has caused job disruption. Policymakers should take this into account. Educational initiatives should act now on AI upskilling and invest in training, so that every member of the workforce can reach their potential and social disruption is minimized.
At DevryBV, we push back against the oft-used excuse, “There are always trade-offs.” Ultimately, each of us has a choice to become active stewards of technology in all its forms. By collectively creating transparency and accountability, supporting sustainable practices, and prioritizing truth over convenience, we can ensure that AI's benefits don’t come at the expense of our planet and our humanity.
James Landis is a DevryBV Summer Intern ‘25. He is currently pursuing a bachelor’s degree in Economics with a minor in AI Applications at the University of Southern California.