This study systematically examines direct liquid cooling (DLC) solutions for high-density data centers, leveraging CFD-based simulations for a 2U GPU server and facility-level design analysis for a 30 MW data center. The results highlight how coolant supply temperature (12–50 ℃) and flow rate (0.25–10 LPM) affect cold plate inlet–outlet temperature differences, pressure drop, and GPU stability. Low flow rates reduce pump energy but risk GPU overheating. By integrating CFD findings into ASHRAE S-Class configurations, we show that low supply temperatures (12–18 ℃) necessitate extensive chiller usage, while higher supply temperatures (42–50 ℃) can eliminate chillers but require adequate flow rates for thermal safety. The optimized approach yields a cooling PUE as low as 1.04, underscoring the energy-saving potential of DLC compared to traditional air-cooled systems. These findings provide actionable guidance on balancing coolant parameters to minimize operational costs while safeguarding IT equipment performance. Hence, the approach underscores how carefully selecting TCS supply temperature and flow rate is crucial for ensuring both GPU reliability and cost-effective data center operation, especially in emerging high-density compute environments.