Wolfspeed Launches 3.3 kV SiC Power Modules Claiming 42% Lower
- Power Electronics Product SiC & GaN Power Semiconductors
Marin Holt
- 0
- 7 minutes read
- Voltage class: 3.3 kV
- Switching loss cut vs other SiC: up to 42%
- Switching loss cut vs silicon IGBTs: more than 90%
- Solid-state transformer footprint reduction: more than 50%
Wolfspeed, the Durham, North Carolina silicon carbide maker, on May 20, 2026 introduced two 3.3 kV SiC power module families aimed at AI data centers, grid-scale battery storage, and solid-state transformers. The company says a new high-power half-bridge baseplate module, rated above 800 amps, cuts switching losses up to 42% against other silicon-carbide modules on the market and by more than 90% against silicon IGBTs, both measured at 125°C on a 1.8 kV bus.
The two parts target different jobs. The HAB900C33LM4 baseplate module handles high-current conversion. The WolfPACK IBB020A33GM4 and IBB020A33GM4T are baseplate-less full-bridge modules built for modularity: engineers stack them in multi-level, series, or parallel arrangements for solid-state transformers and modular renewable systems. Wolfspeed reports the WolfPACK design shrinks solid-state transformer footprints by more than 50%.
What’s Actually New
Both families use Wolfspeed’s Gen 4 SiC and a packaging change rather than a relabeled prior part. Sintered die attach and a copper die-top construction replace the silicon-gel encapsulation of earlier modules, which the company ties to longer power-cycling life, the failure mode that retires modules in hard-switching converters. The 2 kV-plus DC-link rating and improved cosmic-ray tolerance address the higher bus voltages now common in AI-cluster power shelves and utility storage inverters.
How It Compares
Wolfspeed is not alone at 3.3 kV. Microchip released 3.3 kV mSiC modules in May 2026 pitched at halving device count in AI data-center power stages, and Navitas brought 3,300-volt SiC parts to market the same month. Wolfspeed’s pitch leans on the baseplate-less WolfPACK format for solid-state transformers and on its switching-loss numbers rather than device-count reduction. The company has not published list pricing; samples go to select customers through its direct sales representatives, with both families demonstrated at PCIM 2026 in Nuremberg in June.
Critical Perspective
Wolfspeed’s claim of up to 42% lower switching losses holds only at the conditions it specifies, 125 degrees Celsius on a 1.8 kV bus, and the comparison is against other silicon-carbide modules rather than the silicon IGBTs most installed converters still use. The company published no list pricing and is shipping samples only to select customers, so the efficiency claim cannot yet be weighed against cost. Wolfspeed is also not alone at this voltage: Microchip released 3.3 kV mSiC modules in May 2026 pitched at halving device count, and Navitas brought 3,300-volt SiC parts to market the same month, so the 3.3 kV class is becoming crowded rather than a Wolfspeed exclusive. If three suppliers reached the same 3.3 kV class within weeks of each other, how durable is a switching-loss lead that competitors can answer with their own packaging before any of these parts ship in volume?
Why It Matters
Switching losses set the cooling bill and the power density of every converter between the grid and a server rack. A 42% loss cut at 3.3 kV lets designers push more megawatts through the same enclosure, or run cooler at the same load, the constraint now binding AI data-center buildouts and grid-scale storage. The 3.3 kV class also underpins solid-state transformers, the medium-voltage building block several developers are betting on to replace bulky line-frequency iron in data-center and charging infrastructure.
Sources
Related Coverage
| This articleWolfspeed 3.3 kV WolfPACK / baseplate | Microchip 3.3 kV mSiC | Navitas 3,300 V SiC | |
|---|---|---|---|
| Format | Baseplate-less + baseplate | Module | Module |
| Headline claim | Up to 42% lower switching loss | Halves device count | 3,300 V + 650 V GaN |
| Target | AI data centers, SST, storage | AI data centers | AI / industrial |
