Microchip has introduced a 3.3kV SiC power module for solid-state transformers (SST) that reduces the power conversion steps in AI data centers. The product is designed to simplify the power supply structure between the high-voltage power grid and server racks, and to lower the burden of serial connections that occurs in existing low-voltage SiC-based designs.

Microchip Technology announced on the 27th the launch of the '3.3 kV HV-D3 mSiC Power Module,' targeting AI hyperscale data centers and high-voltage power applications. This product integrates a 3.3 kV SiC MOSFET and a Schottky diode into a 62 mm package and is primarily intended for use in SST-based power conversion systems.

SST is a method that supplies power more directly from the medium-voltage power grid to server racks by reducing power conversion steps compared to the existing large low-frequency transformer-centered structure. As the power density of AI servers increases, so do conversion losses and cooling burdens, leading to a growing need for technology that simplifies the power supply structure.

This module is designed to reduce the number of series-connected devices compared to existing low-voltage SiC alternatives in designs connected to 13.8kV or 34.5kV power grids. Since an increase in the number of series connections complicates circuit configuration and insulation design, medium-voltage SiC modules can help reduce the design burden of SST.

The product features a 6kV insulation structure, CTI Grade 600 materials, and an extended creepage design. It utilizes a silicon nitride substrate to enhance thermal conductivity and power cycling performance, and supports 100 to 300A applications.

In addition to AI data centers, the HV-D3 mSiC power module can be applied to megawatt-class charging infrastructure, railway power systems, medium-voltage motor drives, and industrial and defense power systems. As the demand for high-voltage insulation and high-efficiency power conversion increases, the scope of application for SiC-based medium-voltage power modules is expected to expand.