Infineon Targets AI Servers With Newly Added CoolSiC MOSFETs

Infineon has expanded its CoolSiC Gen 2 lineup to include MOSFETs developed for lower voltage levels. With the proliferation of data centers and large AI servers, the lower-voltage class of the CoolSiC portfolio can help designers select a MOSFET that provides maximum efficiency and thermal performance.

 

Leveraging silicon carbide and a lower breakdown voltage, Infineon’s newest CoolSiC Gen 2 MOSFETs enhance the efficiency of AI servers. 

 

These new devices join Infineon’s existing CoolSiC Gen 2 portfolio, which offers low-resistance MOSFETs for high-power applications in the 650-V and 1,200-V classes. With the release of Infineon’s CoolSiC Gen 2 400-V class of MOSFETs, designers now have more options in a broader swath of power electronics designs.

 

Adding to the MOSFET Ecosystem

Optimizing a MOSFET's parameters is no easy task. Ideally, a switching MOSFET would have an infinite breakdown voltage when turned OFF, act like a perfect short-circuit when turned ON, and transition from ON to OFF infinitely fast. In the real world, however, manufacturers must strike a balance between all these parameters while also considering thermal resistance and package size.

In the case of power MOSFETs, the breakdown and on-resistance have a proportional relationship: as breakdown increases, so too does the on-resistance. An optimal MOSFET would have a high enough breakdown voltage to prevent failures but not so high that on-resistance suffers.

 

AI, servers, and data centers demand more power in smaller volumes. MOSFETs must target a narrower market to meet these efficiency standards.

 

Infineon addresses this tradeoff with its latest additions to the CoolSiC Gen 2 portfolio. For designers who don’t need the robustness of 650 V or 1,200 V MOSFETs, the new additions improve on-resistance for applications demanding the best efficiency and volumetric power density. 

 

AI-Targeted Power Supply Efficiency

The 400 V class of Infineon’s CoolSiC Gen 2 MOSFETs are available in 10 varieties (IMBG40R011M2H datasheet linked), with an on-resistance ranging from 11–45 mΩ in a TOLL or D2PAK-7 package. In addition to the 400 V drain-source breakdown voltage, the FETs can support up to 144 A of current draw.

The devices also exhibit good thermal performance, with a 0.35 °C/W and 40 °C/W junction-case and junction-ambient thermal resistance. As a result, the 400 V class of CoolSiC MOSFETs can be deployed in much smaller devices without creating thermal performance issues. The device’s on-resistance also exhibits little change across the entire temperature range, ensuring adequate performance even in harsh environments.

 

The MOSFET is less dependent on RDS(on) and temperature, enabling it to work across the entire operating range without sacrificing efficiency. 

 

Infineon has already used the 400-V MOSFETs in its 12 kW AI power supply unit reference design. The PSU's output power ranges from 3 to 12 kW while maintaining a minimum efficiency of 97.5%. In fact, Infineon has reported up to 99.5% efficiency, a 0.3% improvement over the previous solutions using the 650-V MOSFET. In other words, waste power has dropped by 37.5% using the 400 V class.

 

Powering Power Electronics

While no MOSFET can be considered the perfect switch for every application, the 400 V class of CoolSiC Gen 2 MOSFETs offers designers a broader portfolio of FETs to choose from when optimizing server and data center power supplies.

 

A combination of MOSFETs with differing strengths must be used to continue improving efficiency for power-hungry and space-constrained applications.

 

While the 0.3% jump in efficiency may not immediately seem impactful, the reduction of waste power can have second-order impacts that enable AI servers to continue shrinking without sacrificing performance.

 

---

 

All images used courtesy of Infineon.