High computing demands and data privacy regulations from many countries in the world have resulted in the expansion of data centers. With this, energy consumption by data centers has increased to an alarming level. Data centers are highly dynamic due to time-dependent server heat generation and cold-hot aisle arrangements, making it difficult to have real-time control for efficient thermal management. Traditional cooling strategies based on conservative set points affect energy consumption. Instead of expensive field measurements, CFD analysis of the flow field inside the data center can provide insightful data to assist the heat release from the racks. The detailed rack-level flow field inside the data center is missing in the literature as most of the studies are based on approximate results using the RANS-based k–ε model. However, DES-based models have shown the ability to resolve complex flow fields. With this finding, rack-level CFD analysis of the data center is performed using DES and standard k–ε techniques. At first, average rack inlet and outlet air temperatures in steady-state were validated with experiments within the accuracy of 1.4 ℃. The distributions of turbulent kinetic energy and mean velocity inside the cold aisle were examined. The recirculation region in the cold aisle was well-captured by the DES and qualitatively validated with the experiments compared to the k–ε model. The k–ε model failed to predict SHI, RTI, and β metrics, whereas the DES model successfully captured recirculation and self-heating of the upper servers. An acceptable trade-off between computational cost and accuracy for the simulations would be a pivotal parameter for the selection of either DES or the k–ε model for the data center CFD analysis. The porous media assumption of the servers can bring uncertainties for turbulent quantities and hence further DES model for a data center can provide additional insights in this study.