Integrated process design strategies: Nuclear-powered hyperscale datacenter & cooling

The rapid proliferation of artificial intelligence (AI) services and workloads is reshaping datacenter infrastructure, creating unprecedented demands for computational throughput and associated thermal management. Datacenters therefore require not only large amounts of electricity but also comparable cooling capacity. Although advances in rack-scale cooling improve heat capture from IT equipment, facility-level challenges remain in how cooling is supplied and how the resulting heat is ultimately rejected to the ambient environment. At present, most facilities rely on grid-connected electricity or fossil-fuel-powered islanded configurations. In this context, this study investigates light-water-reactor-based nuclear energy in combined cooling and power configurations as a potential framework for not only providing electricity but also supporting system-scale cooling in AI datacenters. Two representative facility-side cooling pathways—electricity-driven vapor-compression chiller (VC) and thermally driven absorption chiller (AC)— together with their nuclear integration pathways are evaluated using Aspen Plus thermodynamic process simulations. Four nuclear energy integration strategies based on either a single-stage VC system or a double-effect LiBr AC system are proposed and compared for a representative 77 MWe light-water small modular reactor. This work contributes to a nuclear-specific thermodynamic comparison framework that places electricity-driven and heat-driven cooling on a common basis by explicitly accounting for the electric-generation penalty associated with steam extraction. The results show that VC offers the highest peak cooling efficiency under favorable heat-rejection conditions, whereas absorption-based pathways are more robust to condenser-temperature increase and therefore become attractive under hot-weather, water-constrained, or heat-recovery-oriented operating conditions.