Single-phase immersion cooling for data centers and HPC: Recent advances and system-level challenges

Single-phase immersion cooling (SPIC) has emerged as a promising solution to address the escalating thermal demands of data centers and high-performance computing (HPC). This review consolidates recent advances in single-phase immersion cooling system, including discussions on dielectric fluids, cooling architectures, and intelligent control frameworks, with emphasis on experimental validation and system-level deployment. Progress in dielectric fluid is also addressed in details. The common dielectric fluid includes synthetic hydrocarbons and nanofluid-enhanced oils that were examined in association with thermal performance and compatibility with electronic components. At the control frontier, models subject to predictive control, digital twins, and reinforcement learning have enabled real-time optimization, predictive fault detection, and adaptive energy management. In parallel, sustainability metrics such as Carbon Usage Effectiveness (CUE), Water Usage Effectiveness (WUE), Renewable Energy Factor (REF), and Life Cycle Assessment (LCA) extend evaluation beyond traditional power usage effectiveness (PUE), incorporating water footprint, carbon impact, and waste heat recovery. Comparative assessments confirm SPIC's operational stability and scalability, but also highlight persistent challenges in coolant reliability, system retrofitting, and certification standards. By bridging the fluid/thermal science, intelligent control, and sustainability metrics, this review provides a comprehensive framework for advancing SPIC as a cornerstone of sustainable, scalable data center cooling.