BERKALA SAINSTEK
Vol. 14 No. 3 (2026)

Comparative Study of Air and Immersion Cooling in Data Centers: Energy Efficiency, PUE, and Carbon Emissions in Indonesia

Agung, Desy (Unknown)
Pratiwi, SA (Unknown)
Dewi, Rezayanti Novia Putrika (Unknown)
Shalihah, Adinda Rahmah (Unknown)



Article Info

Publish Date
13 Oct 2026

Abstract

The rapid expansion of digital infrastructure has substantially increased data center energy consumption, with cooling systems representing a major contributor to overall facility energy demand. Conventional air-cooling systems face increasing thermal management challenges as computing densities rise, particularly in Artificial Intelligence (AI) and High-Performance Computing (HPC) applications. This study evaluates the potential of immersion cooling as an energy-efficient technology for data centers by comparing its energy performance and carbon emission implications with conventional air cooling. A comparative analytical approach was employed using thermodynamic principles, energy performance indicators, standard operational benchmarks, and data derived from relevant literature. The assessment was conducted using a modeled data center with a 100 kW IT load; therefore, the reported energy and carbon reduction values represent model-based estimates rather than measurements from an operational or experimental data center. The model estimates that immersion cooling can reduce Power Usage Effectiveness (PUE) from 1.80 to 1.20, corresponding to a 33.3% reduction in total facility power consumption. Based on the modeled operating conditions, this reduction is estimated to yield annual energy savings of approximately 525,600 kWh. Using an Indonesian national grid emission factor of 0.85 kg CO₂/kWh, the corresponding avoided Scope 2 emissions are estimated at approximately 446.76 metric tons of CO₂ annually. Furthermore, literature-based evidence indicates that direct contact between dielectric cooling fluid and electronic components can enhance heat removal and support higher rack power densities of approximately 40–100 kW per rack, which is relevant to AI and HPC applications. Overall, the model-based results indicate that immersion cooling has the potential to improve data center energy efficiency and reduce electricity-related carbon emissions, while supporting the transition toward more sustainable and energy-efficient digital infrastructure in Indonesia.

Copyrights © 2026






Journal Info

Abbrev

BST

Publisher

Subject

Agriculture, Biological Sciences & Forestry Biochemistry, Genetics & Molecular Biology Chemistry Mathematics Physics

Description

Berkala SAINSTEK (BST) merupakan jurnal ilmiah yang memuat artikel hasil penelitian di bidang sains dan teknologi. Secara khusus BST diperuntukan bagi penulis internal mahasiswa Universitas Jember bidang fisika, matematika, biologi, kimia, teknik sipil, teknik mesin, teknik elektro dan sistem ...