Poster #149 - Toward Greener AI Infrastructure: Government Officials’ Preferences for Governing Data Centers’ Environmental Impacts

• In Event: Friday Poster Reception

Abstract

Background and Research Question
The expansion of cloud computing and data-intensive applications, including artificial intelligence (AI), has rapidly increased growth in the number and scale of data centers, creating growing pressures on electricity systems, water resources, and local environments (Gohr et al., 2025; Mytton, 2021). These pressures are making data centers’ environmental impacts an urgent policy issue, particularly given the sector’s technical complexity, fragmented accountability, and lack of transparency (Fabrizio & Kim, 2019). While citizens may express concern regarding such impacts, governance in this domain ultimately depends less on individual restraint than on institutional design, regulation, and the allocation of responsibilities, making government officials critical since the measures they adopt shape both business practices and end-user experiences. Yet evidence remains limited on how officials evaluate institutional designs in such emerging sectors. This study thus examines how government officials evaluate alternative governance designs for managing the environmental impacts of data centers. Rather than treating related governance as a binary choice between state control and private responsibility, this study examines how officials evaluate different institutional design features and evaluations are shaped by input, throughput, and output legitimacy, a set of considerations which is highly consequential in officials’ practical governance judgments (Mosley & Wong, 2021). 

Research Design and Methods
The study uses an online conjoint survey targeting U.S. government officials to quantify their preferences over alternative institutional designs for governing the environmental impacts of data centers. Respondents evaluate three rounds of hypothetical governance designs, reporting preferred design and level of support for adoption, with profiles varying across institutional design features including regulatory authority, reporting requirements, audit and monitoring arrangements, government intervention triggers, stakeholder participation, and accountability assignment. To clarify why some features are more attractive, the analysis examines input, throughput, and output legitimacy as mechanisms. The analysis further assesses whether these relationships vary with officials’ representative orientation (Sjöblom, 2026), shedding light on the role of bureaucratic role conceptions. 

Expected Findings and Contribution
The study is expected to show officials’ support depends not simply on whether the state plays a stronger role, but on how different features are institutionally structured. In particular, governance designs combining actors’ collaboration, credible monitoring and verification, explicit intervention and accountability provisions are expected to receive higher levels of support and recommendation for policy adoption. Meanwhile, effects are likely to operate through input legitimacy (voice and participation), throughput legitimacy (fair and transparent procedures), and output legitimacy (accountability and trigger-based intervention). Variation by representative orientation may further reveal officials emphasizing the representative role of public administration are more attentive to procedural credibility, accountability, and the public anchoring of arrangements. By doing so, the study contributes in three ways: substantively, by extending research on environmental governance to the overlooked domain of AI-related infrastructure; conceptually, by treating data center governance as a bundle of institutional design choices evaluated through legitimacy dimensions; and methodologically, by showing how conjoint experiments can identify which governance designs are most likely to gain institutional support.

Author

• Tianhao Zhao, National University of Singapore
  Presenting Author