Infineon Technologies has introduced a cutting-edge 12 kW power supply unit (PSU) reference design tailored for AI data centers and high-performance server platforms. This design showcases exceptional power density and conversion efficiency, leveraging a mix of silicon (Si), silicon carbide (SiC), and gallium nitride (GaN) semiconductors.
As AI workloads push the boundaries of compute and energy demand, Infineon’s new PSU reference design addresses the need for scalable, efficient power delivery. The 12 kW PSU integrates a three-level flying capacitor interleaved power factor correction (PFC) topology on the AC/DC front end, followed by a full-bridge LLC resonant converter with planar magnetics for the DC/DC stage.
Key performance highlights:
- Peak efficiency exceeding 99.0%
- High power density enabled by wide bandgap semiconductors
- Modular architecture suitable for rack-level deployment
This design is aimed at hardware designers, R&D engineers, and power system developers building next-generation AI infrastructure.
Infineon’s use of SiC and GaN devices reflects the industry’s shift toward wide bandgap materials for high-frequency, high-efficiency power conversion. These materials allow for:
- Reduced switching losses
- Smaller passive components
- Improved thermal performance
By combining these with traditional silicon, Infineon achieves a balance of cost, performance, and scalability.
Rather than a commercial PSU, this 12 kW reference design serves as a development platform. It provides engineers with a validated blueprint to accelerate product development, optimize thermal and electrical performance, and reduce time-to-market.
Infineon also emphasizes the design’s compatibility with AI server rack architectures, making it suitable for hyperscale data centers and edge AI deployments.
According to Richard Kunčič, Infineon’s Senior VP and GM of Power Systems, the company’s goal is to “preserve every single possible Watt” in the face of AI’s growing energy appetite. This aligns with broader industry efforts to reduce data center carbon footprints while scaling compute capabilities.
Share to your social below!
