The Princeton Review publishes a “Top 50 Green Colleges” list annually. The list aims to highlight colleges and universities that have “superb sustainability practices, a strong foundation in sustainability education, and a healthy quality of life for students on campus.” These colleges tout LEED-certified buildings, renewable energy initiatives, and zero-waste goals. Yet, as generative AI products like ChatGPT, Claude, and Gemini infiltrate higher education, a new tension emerges. Can a college truly claim to be “green” while embracing technologies with adverse environmental impacts? Colleges that aspire to be on the “Top 50 Green Colleges” list must confront this question.
GenAI models are computationally intensive. Training GenAI models, such as Large Language Models (LLMs), requires massive computational power, thereby resulting in substantial electricity use and CO₂ emissions. For example, training GPT-4 demands “a staggering amount of electricity, which leads to increased carbon dioxide emissions and pressures on the electric grid.” AI data centers also rely on liquid-cooling systems, thereby consuming millions of gallons of freshwater per day. The water and electricity consumption of data centers is only expected to rise with the growing demand for GenAI products. According to the International Energy Agency, global data-center electricity demand reached around 460 terawatt-hours in 2022 and may double by 2030.
The development and deployment of GenAI products require specialized hardware, such as Graphics Processing Units (GPUs), which contain rare earth and critical minerals.Manufacturing a single high-end GPU can emit approximately 200 kg of CO₂ and require thousands of gallons of water per wafer. The mining of rare earths, lithium, and cobalt, which are needed to make the hardware on which GenAI products run, could lead to environmental degradation, mine runoff, and biodiversity loss in extraction regions. GenAI hardware churns turnover quickly, which is a process that results in electronic waste (e-waste). When used in data centers, GPUs often become obsolete within two to five years, therebyresulting in a projected 2.5 million tons of e-waste per year by 2030.
By endorsing GenAI products for their students, faculty, and staff, colleges may be tacitly acquiescing to the high environmental costs of GenAI. But can they still honestly claim the status of “green college” if they do?
Consider, for example, the University of Vermont (UVM), which is ranked #11 on the “Top 50 Green Colleges” list. UVM boasts aggressive waste-reduction policies and a respected environmental studies program. But UVM also partnered with the consulting firm EAB to deploy AI-driven analytics to improve student retention. According to Jay Jacobs, Vice Provost for Enrollment Management at the University of Vermont, UVM is “excited to pilot EAB’s new AI features to help staff save time, create deeper connections with our students, and have a measurable impact on their lives.” While these AI products may offer some efficiency, their development and deployment come with the aforementioned environmental costs of high energy consumption and carbon emissions, intensive water usage, raw minerals extraction, and e-waste.
Another example is Arizona State University (ASU), which is ranked #48 on the “Top 50 Green Colleges” list. ASU is celebrated for its solar power initiatives and sustainability-focused programs. But its partnership with OpenAI underscores the growing tension between aspiring to be a “green college” and a leader in GenAI adoption. Can ASU simultaneously be a sustainability champion and an AI innovator?
This tension is not only operational but also ethical. Colleges often market themselves as sustainability leaders to attract students who care deeply about climate change and its adverse effects. At the same time, they also adopt GenAI products to remain competitive and innovative. This dual pursuit risks sending a mixed message to students. Discerning students are likely paying attention. They might question why their college is investing heavily in GenAI products that exact a severe environmental toll while at the same time aspiring to be carbon-neutral.
The tension between being a “green college” and embracing GenAI reflects an even broader societal and ethical problem, namely, how to balance technological progress with environmental stewardship. Colleges, especially those that tout sustainability as a core value, must confront this problem. If colleges genuinely aim to model sustainability, they must take the carbon cost of digital innovation seriously.
As the Princeton Review continues to exert influence on colleges’ reputations around sustainability, there is an opportunity for its Green College rankings to incorporate the environmental impacts of GenAI. It can do so by including measures of digital footprints as well. Until such measures are adopted, colleges face a dilemma, namely, either redefine what it means to be a “green college” racing to adopt GenAI products or run the risk of bluewashing, which is the greenwashing of the GenAI era.