ROHM's 750V withstand voltage SiC MOSFET has been adopted in a battery backup unit (BBU) for AI server power. This is an instance where SiC power semiconductors have been applied as core components in the backup power supply in the AI server market, which is shifting toward high-voltage DC-based power structures.

ROHM recently announced that its SiC MOSFET 'SCT4013DLL' has been incorporated into the power supply of a ±400V power supply architecture for AI servers. As competition for GPU performance continues following the spread of generative AI, server power demand has increased, and the data center industry is considering the introduction of HVDC-based power structures to reduce power loss.

A BBU is a device that compensates for power at the server rack level in the event of a power outage or momentary voltage drop. For systems with large data processing volumes and high power consumption, such as AI servers, the ability to rapidly maintain and control power in the event of an anomaly is critical. Accordingly, there is a growing need for power semiconductors that reduce losses and operate stably even in high-voltage and high-current environments.

The SCT4013DLL applied this time is a SiC MOSFET with a voltage rating of 750V. According to ROHM, this product supports a maximum junction temperature of 175℃ and is designed to ensure stable operation even in BBU environments where heat generation is a significant burden due to high power density and high voltage.

In the next-generation 800VDC power supply structure, the voltage supplied to the battery pack inside the BBU is suggested to be around 560V. ROHM explained that 750V-class SiC MOSFETs can be applied under these voltage conditions.

SiC power semiconductors are more advantageous than silicon-based devices in high-voltage and high-temperature environments, leading to an expanding scope of application in data center power supplies, following electric vehicles, industrial equipment, and energy storage systems. As the power structures of AI servers become higher voltage, power conversion efficiency and thermal management performance are expected to emerge as key variables in power system design.

ROHM plans to expand the development and supply of SiC, GaN, and silicon-based power devices targeting the future AI server and data center markets. In addition, the company intends to meet the demand for high-efficiency power systems by proposing power solutions combined with analog ICs.