In an era of artificial intelligence and big data, traditional data center facilities no longer meet market demands. Since the fourth quarter of last year, many clients have been deploying H100 GPUs, specifically designed by NVIDIA for AI computation, and have higher power density requirements. When it comes to the recent developments in artificial intelligence (AI), “Even if clients do not explicitly state their purposes, we can see the trend from their requirements of data center facilities, said Johnny Liu, President of Chief Telecom.
Johnny Liu, President of Chief Telecom
With global cloud service providers expanding their Asia-Pacific footprints to Taiwan and the increasing demand for cloud data services from domestic industries, a research report by SinoPac Securities Investment Service Corporation indicates that Chief Telecom is the only company in Taiwan that can directly connect with the world’s top three cloud service providers—AWS, Azure, and GCP. Currently, Chief Telecom’s data centers, including HD and LY, have an occupancy rate of over 90%. Chief’s latest LY2 AI data center is expected to reach an occupancy rate of 80% by the end of 2025.
Traditional server racks usually consume up to 10kW per rack. However, in the AI era, with the increase in data transmission and enhanced computing power, the power density of a single rack cabinet can probably soar up to 400kW. Predicting the future trend of high-performance computing (HPC), Chief Telecom has secured the support of Taiwan Power Company with 20MW power capacity for LY2 AI data center, enabling it to meet the rising demand for AI computing power.
To ensure smooth data center operation and mitigate any risks associated with power supply, Chief Telecom’s LY2 data center is equipped with over 10 generators with a capacity of 2.25kW each, as well as an uninterruptible power supply (UPS) system. Even if external power is interrupted, LY2 data center has enough power capacity to maintain normal operations for a month, ensuring that customer data and operations remain unaffected. Additionally, LY2’s infrastructure includes civil engineering with a load-bearing capacity of 2000 kilograms per square meter and space reserved for both air-cooling and liquid-cooling systems to provide a stable operating environment for customers’ servers.
LY2 AI data center