As AI servers incorporate greater computing capacity, managing the heat they generate has become increasingly challenging for conventional air-cooling methods. To address this, Taiwan’s Industrial Technology Research Institute (ITRI) collaborated with Flotrend Corporation and used Siemens’ Simcenter STAR-CCM+ software to simulate liquid cooling approaches, including two-phase immersion cooling. Instead of relying solely on physical prototypes, the team applied simulation and digital twin methods to evaluate the thermal performance of different coolants and cooling configurations for densely packed AI chips. The initiative aims to support informed cooling system design for AI data centers while addressing energy efficiency and sustainability considerations in high-performance computing.
Image courtesy: Siemens