The rapid growth of Artificial Intelligence (AI) and cloud computing infrastructure is driving an unprecedented increase in electricity demand from data centers, which are projected to consume up to 6-12% of U.S. electricity by 2028 (Shehabi et al., 2024). Nuclear power plants (NPPs) are increasingly being considered as reliable low-carbon energy sources capable of supplying the firm baseload power required by these facilities (U.S. White House, 2025). However, the co-location and operational coupling of NPPs with data centers introduces new challenges (Talbot et al., 2025). Data centers have strict availability and power-quality requirements that must be maintained to ensure continuous operation of latency-sensitive AI workloads (Ahmed et al., 2021). To meet these requirements, automated operational responses, such as rapid load rejection or voltage ride-through protection may be activated in response to grid or facility disturbances, resulting in rapid downstream load variations that propagate through the shared electrical interface and induce NPP operational transients, potentially affecting plant safety margins. Additional concerns arise from shared cooling and backup systems, and increased cybersecurity risk (Talbot et al., 2025). Existing Probabilistic Risk Assessment (PRA) methodologies for nuclear installations explicitly account for component and system reliability within plant boundaries, but they do not typically consider tightly integrated external infrastructures whose operational behavior may influence nuclear safety through electrical, thermal, and cyber dependencies.
To overcome these limitations, this work proposes a conceptual framework for extending nuclear PRA to account for the integration of data center infrastructures as coupled external systems. Adopting a system-of-systems perspective, the framework models data centers as dynamic semi-dispatchable loads and captures the shared electrical and thermal dependencies among the systems. The additional cybersecurity risk arising from operational coupling is quantified by extending the initiating event set to include cyber-induced plant disturbances, such as load rejections and voltage transients resulting from compromised digital control pathways. The framework provides a foundation for risk-informed deployment and operation of data centers powered by NPPs.

Keywords: Nuclear Power Plants (NPP), Data centers, Probabilistic Risk Assessment (PRA), System-of-systems, Infrastructure Interdependencies.

References
Ahmed, K. M. U., Bollen, M. H. J., & Alvarez, M. (2021). A Review of Data Centers Energy Consumption and Reliability Modeling. In IEEE Access (Vol. 9). https://doi.org/10.1109/ACCESS.2021.3125092
Shehabi, A., Smith, S., Sartor, D., & Brown, R. (2024). United States Data Center Energy Usage Report (LBNL-2001637). https://doi.org/10.71468/P1WC7Q
Talbot, P., Bryan, J., McDowell, D., Enriquez-Contreras, L., Alam, S. M. S., Tacke, J., Culler, M., Epiney, A., & McJunkin, T. (2025). Navigating Integration: Key Challenges for Data Centers, Nuclear Stakeholders, and Utility Operators (INL/RPT-25-87663).
U.S. White House. (2025). Deploying Advanced Nuclear Reactor Technologies for National Security. Executive Order 14299.