Investigating the Ecological Impacts of Data Centers

In leading the MCSC’s work on Data & Computing, Noman Bashir has been exploring effective decarbonization strategies in this space, with an emphasis on generative artificial intelligence (Gen-AI) – a large-scale technology-induced shift that is rapidly expanding. While the perceived advantages of Gen-AI are vast, neglecting to consider the negative effects alongside these potential benefits can lead to uncontrolled growth with lasting consequences on the environment. When exploring some of the key drivers of this growth, why sustainability considerations for Gen-AI are lacking, and how ignoring the potential costs could lead to negative social and environmental impacts, data centers are central to the conversation.

With the rapid growth of Gen-AI, data centers’ energy demands are increasing. The quality of Gen-AI algorithms correlates with the size of required computing systems, necessitating a larger number or more powerful sets of computers for more sophisticated data sets and models. Because of this pressure, there are both rising power demands on existing data centers and the need to build new ones. Some organizations, AI giants in particular, are deploying more data centers and computing infrastructure than ever before.

“Our capacity to build sustainably cannot keep pace with the data center construction necessary to support Gen-AI, and the global electricity consumption of data centers is expected to continue to rise,” explained Noman, whose research focuses on designing, prototyping, and analyzing experimental computer systems that enable decarbonizing societal-scale systems, such as computing systems, electric grids, buildings, and transportation.

To investigate the ecological impact of these data centers, Noman has been working with Wacuka Ngata, who is studying Materials Engineering, through MIT’s Undergraduate Research Opportunities Program (UROP). Wacuka’s research interests center on the broad environmental footprints of data centers, looking beyond established research on energy use and carbon emissions to examine their localized effects.

“Factors like water consumption, noise and air pollution, and the impact on surrounding communities are underrepresented in current research initiatives involving data centers,” she explained.

To address this gap, Wacuka has been supporting Noman in developing a comprehensive database of empirical evidence – which will include these often-overlooked factors – aimed at informing better decision-making for sustainable computing. The database will highlight information gathered from stakeholder (local communities, environmental groups, and neighborhood coalitions) interviews, and highlight the specific concerns that communities have regarding ongoing expansive data center projects, being developed to meet growing demand and use of Gen-AI. The conversations will help Noman and Wacuka identify common and recurring issues and themes that are not yet well-documented in existing research – ultimately challenging current boundaries, identifying opportunities for future studies, and allowing new ideas and solutions to emerge.

“The details that we are capturing in the database will identify specific ecological concerns that will inform next steps for reducing data centers’ environmental footprints,” said Wacuka.

To begin identifying these key areas of concern and improvement, Wacuka will be reaching out to stakeholders in Northern Virginia, an area known as “Data Center Valley” for containing the world’s highest concentration of data centers.

“Growing up in Northern Virginia, the data center capital of the world, influenced my perspective on the rapid expansion of computing infrastructure,” Wacuka explained. “While most people think of ‘the cloud’ as something intangible, I had a physical reminder of it in my backyard. I noticed over the years that many forested lands were cleared to make way for these computing facilities, and as I saw protest boards pop up on the side of roads, I became increasingly curious about their broader environmental impacts – not just in terms of energy use but also their effects on surrounding communities and ecosystems.”

As the demand for computing continued to grow, Wacuka questioned how to optimize these systems to reduce their physical and environmental footprint. This curiosity led her to her current research, where she is exploring ways to minimize the ecological impact of data centers.

After months of preparations, the next step in Noman and Wacuka’s work is to conduct the stakeholder interviews – and part of this extensive preparation has been to strategically and thoroughly craft the outreach in ways that will effectively uncover the valuable information they need to move forward. To do so, Wacuka leaned on interdisciplinary perspectives within the MCSC.

She collaborated with MCSC Impact Fellows Michelle Westerlaken and Laurent Lioté, whose expertise in the social dimensions and policy spheres informed this large and important stakeholder engagement aspect of her work. They provided guidance on how to design the interviews to minimize confirmation bias and ensure that Wacuka crafted open-ended questions that gave participants the space to share insights freely. Michelle and Laurent also emphasized the importance of entering interviews with a clear idea of each interviewee’s potential role, while remaining flexible to participant-driven discussions.

“It has been incredibly valuable collaborating with Laurent and Michelle, as their expertise in qualitative research, including conducting stakeholder interviews, provided crucial guidance in both the preparation and (soon) the execution phases of our work,” said Wacuka.

Michelle and Laurent’s insights will continue to help shape how Wacuka and Noman structure their database and interpret stakeholder feedback, ensuring that the research captures a comprehensive and unbiased perspective on the ecological impacts of data centers.

In addition to interviewing key community stakeholders, discussions with data center operators, tech companies, and utility providers will allow Noman and Wacuka to better assess the effectiveness of current sustainability measures as they relate to data centers and Gen-AI broadly. Identifying areas of improvement and understanding what measurement methods are already being implemented will help them evaluate opportunities for sustainable solutions.

Conversations with local governments and economic development agencies may also pave the way for a clearer understanding of where decision-making power truly lies in the development and regulation of data centers.

“Learning more about the key decision-makers behind data centers is another pillar to our research that I am eager to explore,” Wacuka concluded. “These conversations will be crucial for later idea-generation on actionable solutions that are both impactful and feasible to implement.”