As colleges and universities increase their adoption of artificial intelligence, and particularly generative AI, there is a parallel, rising need for AI literacy instruction. Since librarians are experts in information literacy and technology, they are often on the frontlines of providing training in AI literacy. Indeed, this is reflected in Ithaka S+R’s recently published 2025 US Library Survey, where academic library leaders were asked what they anticipate the most significant impacts of AI will be on their libraries. The most common answer, selected by 83 percent of respondents, was increased demand for AI literacy instruction.

Dominant frameworks for AI literacy usually focus on the foundational knowledge learners need to employ AI and understand how it works. The ethical and social implications of AI—including AI’s environmental impacts—are also often included among learner competencies. Librarians, faculty, students, and others across higher education are increasingly and understandably concerned about how widespread adoption of generative AI tools is affecting the climate and the planet broadly speaking. In order for the higher education community to make informed choices about AI tools, AI literacy curricula must address these pressing environmental dimensions of the technology.

Today, we are publishing a LibGuide focused on the Environmental Impacts of AI, as a part of our Incorporating Environmental Perspectives into AI Literacy project, funded by the Mellon Foundation. The LibGuide’s objective is to help users attain a baseline understanding of the varied environmental consequences behind AI technology. The LibGuide’s articles, reports, podcasts, videos, data trackers, and other types of resources address environmental impacts throughout the AI lifecycle. These include energy and water use, emissions, mining, hardware construction, and e-waste, as well as resources about data centers and their effects. The guide also links to research on how AI could be developed more sustainably, tools that track AI’s environmental footprint, policy recommendations, and more.

The LibGuide intends to offer librarians, instructors, and others within the higher education community a gateway for learning more about how AI development and use are impacting the environment. By providing a set of resources that can be inserted within broader AI literacy training, the LibGuide makes it easier for those teaching or learning about AI to include content about AI and the environment. The LibGuide can serve as a launch pad for further research into specific aspects of AI’s environmental impacts, building from the baseline of resources it provides.

The materials in the LibGuide reveal that researchers still do not have a full grasp of AI’s environmental impacts, and opinions vary on how to think about and act on information available to date. Research in this area is evolving rapidly, and estimates from different sources—whether for measurements of emissions, water or energy use, or other impacts—do not always align with each other. Major technology companies are also not always transparent about resource use and other environmental impacts. Despite these challenges, the resources included in the LibGuide document what we do know about the impacts of widespread AI adoption on the planet.

We are deeply thankful to the four members of our project’s advisory committee who are offering us guidance and feedback on our work. Below, each of these advisors recommends a resource or two from the LibGuide and explains why the resource is important.

Beth Filar Williams, user experience research librarian, Oregon State University Library

If you’re looking for a resource to learn the basics of AI and its environmental impacts, this primer is a great place to start. It covers, in simple language, the energy and water usage, mineral extraction, and GHG emissions related to AI and data centers. It also shares basic information on the lack of transparency disclosures from large AI companies, regulation and policy of various countries, and has a useful glossary of terms along with many cited resources. Dr. Luccioni, one of the main authors of this primer, is a leading researcher on this topic, making this resource well vetted. If you prefer to listen than read, also consider her 10 minute TED talk that covers the basics but also how we could do it right by considering smaller focused language models than these few huge LLMs which use an incredible amount of resources and are operated by just a handful of companies.

Chris Rabe, program lead, Universal Climate, Massachusetts Institute of Technology (MIT) Open Learning; Education Program Director, MIT Climate Project

Most new resources on AI rightly focus on the extreme energy and water demands of the data centers that run it. What often gets lost is the labyrinth of material infrastructure at global scales that powers all of computing. In the MIT Case Study The Cloud Is Material: On the Environmental Impacts of Computation and Data Storage, Steven Gonzalez Monserrate uses ethnographic research, personal photography, and a figurative style to “materialize the immaterial” and show that the “cloud” is nothing like a cloud. It is a global infrastructure dependent on rare mineral extraction, undersea fiber optic cables, millions of data servers, vast cooling systems, and endless piles of e-waste.

Reading it is essential for understanding the material consequences of AI, and a powerful reminder that every prompt and query has a physical footprint shaping ecosystems, communities, and economies worldwide.

Eira Tansey, founder and manager of Memory Rising

When engaging with resources to understand the environmental impacts of artificial intelligence, grasping the math and scale around the resource demands can be difficult. This MIT Technology Review, We did the math on AI’s energy footprint. Here’s the story you haven’t heard, is one of my favorites I’ve encountered so far for understanding the interactions between data centers and the electrical grid. For visual learners, this essay is both informative and an accessible resource to add to instruction materials.

It’s important to understand the underlying grid concerns that AI is exacerbating. This is where the International Energy Agency (IEA)’s 2025 report Energy and AI is particularly helpful. Different countries around the world have varying forms of clean energy adoption. Unfortunately the United States (where many of the major data centers are currently located with more being built) is behind much of the rest of the world when it comes to renewable energy transition. According to the IEA, “In the United States, data centres account for nearly half of electricity demand growth between now and 2030” (p. 14). This report, and others from the IEA, are important to contextualize the development and adoption of AI within the global energy landscape.

Sarah Tribelhorn, sciences and sustainability librarian, San Diego State University

One standout tool featured in the guide’s “Trackers & Tools” section is EcoLogits. This open-source tool, developed by CodeCarbon, is specifically designed to estimate the energy consumption and environmental footprint of generative AI models. Unlike general trackers, it provides precise estimates for the resource-intensive process of using large-scale models, including electricity, carbon footprint, water, metals and minerals, and fossil fuels. By utilizing EcoLogits, developers and researchers can gain a data-driven understanding of their digital footprint across the AI life cycle. As we move forward, adopting tools like these is a vital step for any organization aiming to balance technological innovation with long-term planetary sustainability.

For more information about our Incorporating Environmental Perspectives into AI Literacy project, please reach out to Claire Baytas (claire.baytas@ithaka.org).