AI COMPUTING

Thanks to their high-voltage capability, thermal robustness, and high-frequency performance, wide-bandgap semiconductors-particularly Silicon Carbide (SiC) and Gallium Nitride (GaN)-have become essential to improving the energy efficiency of AI computing infrastructure.

Boosting Energy Efficiency to Meet Growing Power Demand

As AI workloads continue to expand, data center power consumption has increased dramatically. SiC and GaN power devices significantly reduce power conversion losses across critical systems. In server power supply units (PSUs), SiC-based modules can achieve system efficiency levels above 97%, substantially reducing energy waste compared to traditional silicon-based solutions.

HVDC Architecture and Ultra-High Power Density

The power industry is rapidly shifting toward 800V high-voltage DC (HVDC) architectures to address the inefficiencies of conventional power distribution, particularly when system power demand exceeds 200 kW. Wide-bandgap devices enable next-generation solid-state transformer (SST) systems capable of converting 13.8 kV AC directly to 800V DC. This eliminates multiple conversion stages, improves transmission efficiency by up to 5%, and reduces copper usage by approximately 45%.

At the same time, the high-frequency switching capability of GaN devices enables the use of much smaller magnetic components, reducing overall PSU size by 30% to 50%. This is especially valuable for space-constrained data center environments where power density is critical.

Reduced Cooling Requirements and Higher System Reliability

SiC’s thermal conductivity-approximately three times higher than that of silicon-combined with GaN’s low switching losses, significantly reduces heat generation. This lowers cooling requirements, improves Power Usage Effectiveness (PUE), and enhances long-term system reliability and operational stability.

In summary, wide-bandgap semiconductors improve energy efficiency, increase power density, and strengthen system reliability. They are becoming the essential “power heart” behind sustainable AI computing growth and one of the key enabling technologies for overcoming the energy barriers of high-performance computing.