Massive data centers powering the boom in artificial intelligence and other digital services are increasing demand on the U.S. electrical grid and straining natural resources. A report released today by Duke University researchers surveys the sustainability challenges these data centers present, examines potential solutions and outlines pragmatic recommendations for companies, utilities, regulators and policymakers.
The analysis conducted by Ian Hitchcock and Merritt Cahoon—researchers at the Sanford School of Public Policy’s Deep Tech Initiative—focuses on large-scale cloud computing providers known as hyperscalers. These companies, such as Amazon, Google, Microsoft and Meta, are rapidly expanding their data centers to meet rising demand for AI, cloud computing, big data analytics and digital transformation across industry and government.
“This report reframes hyperscale data centers not simply as passive consumers of energy and water, but as active agents in the energy transition and water management,” said Jackson Ewing, director of energy and climate policy at the Nicholas Institute for Energy, Environment & Sustainability. “In collaboration with policymakers and stakeholders, hyperscalers could emerge as accelerants of clean technology deployment, grid modernization and operational innovation rather than impediments to sustainable futures.”
Energy use from hyperscalers is forecast to grow substantially, but estimates vary from anywhere to 6.7% to 12% of total U.S. load by 2030. The report notes that uncertainty, coupled with pressure to connect new data centers to the grid quickly, is causing concern over shielding everyday residential and commercial ratepayers from the costs of building out new energy infrastructure.
Another consideration is the energy sources for new data centers. Many hyperscalers have made sustainability commitments, leading them to seek renewable energy or invest in accelerating next-generation technologies. However, hyperscalers can build data centers more quickly than the energy infrastructure needed to service them can be permitted and constructed.
While the energy demands of data centers have been widely publicized, they also consume large amounts of water for cooling. This can put stress on local resources, particularly in water-scarce regions that attract operators with low-cost land and electricity. Yet the report notes that water availability is often not a priority during site selection processes.
The authors also found a lack of reliable data and transparency from hyperscalers on both energy and water use. That makes it harder for policymakers, regulators, utilities and communities to make sound sustainability decisions.
Media Inquiries
Duke experts are available to speak with reporters interested in learning more about the findings in the report.
Contact: ni-comm@duke.edu
Hitchcock and Cahoon offer several recommendations for addressing these challenges:
- Hyperscalers should continue their efforts to accelerate deployment of renewable energy and commercialization of next-generation clean firm power, such as advanced geothermal, small modular nuclear reactors and carbon capture utilization and storage.
- Electric utilities could explore employing load flexibility strategies to meet load growth to avoid costly infrastructure projects to service load that may not materialize. An analysis published in February by Duke scholars estimates nearly 100 GW of large loads could be added to the grid without costly infrastructure upgrades, provided those loads occasionally cut their power use during times of peak demand.
- Creating opportunities for the very different cultures of technology developers and utility operators to interact will be critical to thoughtfully managing the implications of digital infrastructure build-out.
- Policymakers could require data centers to produce data about the anticipated energy and water use of their operations before approving new construction.
- The availability of renewable energy and water should take greater priority in data center siting decisions.
The sustainability challenges posed by data centers are often local in nature, so solutions can vary. With that in mind, the report includes case studies that highlight some of the different approaches cities and states are taking.
“Solutions to the sustainability challenges that data centers pose will have to be interdisciplinary and involve the private sector, technology and policy experts, and communities,” said David Hoffman, managing director of the Deep Tech Initiative, interim director of the Duke Initiative for Science & Society and professor of the practice at the Sanford School of Public Policy. “This makes Duke University, with its commitment to interdisciplinary problem solving, an effective home for this research. The recommendations in this report offer a roadmap for further research, collaboration and engagement.”
In May, Provost Alec D. Gallimore launched a new AI initiative to facilitate collaboration among Duke experts who are advancing AI research and development, addressing related ethical issues and shaping the future of AI in the classroom.
“AI presents humanity with so many opportunities—but also some daunting challenges, including those related to sustainability. This report is an example of how Duke experts are contributing to the global conversation, helping clarify what is at stake and highlighting potential ways forward.”
Duke University Provost Alec D. Gallimore
Duke University Provost Alec D. Gallimore
###
CITATION: Hitchcock, I., and M. Cahoon. 2025. Hyperscaler Data Center Buildout: A Sustainability Bane, Boon, or Both? NI R 25-04. Durham, NC: Nicholas Institute for Energy, Environment & Sustainability, Duke University. https://nicholasinstitute.duke.edu/publications/hyperscaler-data-center-buildout-sustainability-bane-boon-or-both.
For media inquiries, contact the Nicholas Institute communications team at ni-comm@duke.edu.
