Now that the data center boom is well underway, it’s beginning to dawn on the average citizen that these facilities consume an enormous amount of water. We’re talking millions of gallons daily for large-scale data centers. But this isn’t news to the companies that power data centers, such as LinkedIn, for whom becoming “water positive,” or returning more clean water to the environment than consumed, is a goal they aim to achieve by 2030. A team of graduate students had the opportunity to contribute to the company's goal by helping them understand their indirect water usage. 

The four students from the University of Michigan School for Environment and Sustainability (SEAS), Stella Li (MS ’26), Treye Meadows (MS ’26), Roxanne Wang (MS ’26), and Bonny Wysocki (MS ’26), started the work by spending 18 months assessing the scope 2 water use from purchased electricity at LinkedIn. Scope 2 focuses on the indirect water consumption needed to power the organization, which, for LinkedIn, primarily relates to the in-person offices and other points of presence. 

Meadows explains that LinkedIn’s U.S. work makes up approximately 97% of its total energy usage. With the intensity of the use of data centers, LinkedIn wanted to better understand how it can prepare for the future.

The team went through a variety of processes in an attempt to determine the best way to collect and analyze the needed data. They explain how they tried to develop their own methodology, but as they continued to research, they found that the existing tools were the best option. 

“We ended up choosing two methodologies and gathering data from the EIA [U.S. Energy Information Administration],” says Wysocki. 

They also used tools such as the Water Impact Tool to obtain water consumption and withdrawal data, Aqueduct 4.0 for the baseline water stress scores, and AWARE 2.0 to identify remaining available water to humans and ecosystems post-consumption. 

“We utilized both AWARE and Aqueduct because we thought they would mesh well together. No model is perfect, so utilizing two complementary methodologies can help with the gaps in each,” explained Meadows. 

The team highlights their advisor, SEAS Assistant Professor Benjamin Goldstein, and the client as incredibly helpful resources, who pointed them in the right direction and assisted in learning how to use new tools. 

After months of reviewing the literature and analyzing data, the team is now wrapping up, refining details and creating data visualizations. They will be providing a folder with their methodologies, a slide deck and a final report to LinkedIn and its advisors. The team is hoping that their research is going to help LinkedIn with their decision-making on water usage, but also hopes that it is a stepping stone in the broader discussion of data centers. 

The project members say that some of their favorite parts of the work include learning how to clean data, creating their own numerical models and having the opportunity to work on something so relevant to the current period. 

Meadows adds, “The demand for data centers is growing. If we are going to see more of them, then it should be our goal to minimize their impacts.”