A company called SemiQ just dropped a family of silicon carbide power modules built specifically for the motor drives inside data center liquid cooling systems. The QSiC Dual3 lineup includes six 1200V half-bridge SiC MOSFET modules in a 62mm x 152mm S4B1 package, with on-resistance options at 1mΩ, 1.4mΩ, and 2mΩ. Two of the modules hit 240 W/in³ power density. The target application: 250kW liquid chiller systems running active front ends and compressor drives.
That is a component-level announcement. Most people in this industry will scroll right past it. They should not.
Cooling systems consume 30 to 40 percent of total facility energy in a typical data center. In less-efficient enterprise facilities, that number climbs above 40%. The instinct is to attack that problem at the system level: better chillers, smarter controls, fancier coolant distribution units. All of that matters. But the efficiency ceiling of any chiller is ultimately set by the power electronics driving its compressor and pumps. If the inverter stage is burning watts converting DC to AC, or the active front end is hemorrhaging energy through switching losses, the rest of the system design is working against a handicap.
That is where silicon carbide enters the picture.
The vast majority of data center chiller systems still run on silicon insulated-gate bipolar transistors. IGBTs. They have been the default power semiconductor for industrial motor drives since the 1990s. They work. But they carry a fundamental physics penalty: a bipolar tail current at turn-off that dumps energy as heat every single switching cycle. At the 16kHz switching frequencies common in motor drive applications, those losses compound into real money.
SiC MOSFETs eliminate that tail current entirely. According to Wolfspeed's own testing, replacing IGBTs with SiC devices in the active front end and inverter stages of a 25kW compressor system yields a combined 2.4% efficiency improvement with a 50% reduction in total losses. At 16kHz, motor drive efficiency improves by more than 1.2%. Double the switching frequency to 32kHz and the passive components shrink, the magnetics get smaller, and you still hold above 98% conversion efficiency.
Scale that math to a 250kW chiller. A 2.4% efficiency gain on a quarter-megawatt system is 6kW of saved losses. Per chiller. Multiply across a facility running dozens of cooling units and the cumulative impact on PUE stops being theoretical.
SemiQ is based in Lake Forest, California, and has spent over a decade in the SiC device business. Timothy Han, the company's president, holds a PhD in electrical engineering from KAIST and has 30 years of experience in SiC power devices and module design.
The Dual3 family ships in three MOSFET-only configurations and three versions with a parallel Schottky barrier diode integrated into the package. The SBD variants matter for hard-switching topologies where the body diode of the MOSFET would otherwise conduct during dead time, generating reverse recovery losses. By co-packaging a SiC Schottky barrier diode, SemiQ eliminates that penalty. For thermal engineers designing chiller drive systems that run at elevated junction temperatures, the SBD versions should reduce derating requirements and simplify thermal management of the module itself.
Every MOSFET die in the Dual3 lineup goes through wafer-level gate-oxide burn-in screening above 1,450V. That detail will mean nothing to most readers. To a reliability engineer, it means SemiQ is stress-testing the gate oxide of each individual die before it ever reaches a module, catching early-life oxide failures that would otherwise show up as field returns. Gate oxide integrity is the single biggest reliability concern in SiC MOSFETs. The screening voltage sits well above the 1,200V rated blocking voltage, which puts real margin into the process.
The physical package is a 62mm x 152mm half-bridge configuration. SemiQ says the modules have low junction-to-case thermal resistance, which translates directly into smaller heatsinks and simpler mechanical design for the cooling system OEM. If you can pull heat out of the module faster, you need less aluminum and less airflow to keep junction temperatures in range.
SemiQ is positioning the Dual3 as a drop-in replacement for existing IGBT-based designs with minimal redesign. That framing matters because the biggest barrier to SiC adoption in industrial motor drives has never been the semiconductor performance. Everyone who has run the numbers knows SiC wins on efficiency. The barrier is the engineering cost of redesigning the power stage, the gate drive circuitry, the layout, the thermal stack.
"Rising AI-driven power and thermal demands in data centers are pushing the limits of traditional cooling and power systems," Han told Charged EVs. The statement is accurate, if understated. The thermal density roadmap set by Nvidia's GPU platforms has already forced a wholesale rethinking of cooling architecture. Rack power densities that sat at 10-15kW five years ago are now pushing past 100kW in AI training clusters. The cooling systems keeping those racks alive are drawing more power, running harder, and operating with less margin than ever before.
Any efficiency gain at the power conversion layer compounds across the entire thermal chain. A more efficient compressor drive means less waste heat generated by the cooling system itself, which means the facility's secondary cooling loop works less hard, which means less water, less energy, less mechanical stress on every downstream component.
SemiQ is not Wolfspeed. It is not Infineon. It does not have the brand recognition or the manufacturing scale of the tier-one SiC suppliers. But the company has been shipping SiC Schottky diodes and MOSFETs into demanding applications for years, and the Dual3 modules were debuted at APEC 2026, the biggest power electronics conference in the industry.
The real question is whether chiller OEMs will actually make the switch. Companies like Vertiv, Carrier, and Trane have enormous installed bases of IGBT-driven systems. Redesigning a power stage, even a "minimal" redesign, requires validation testing, thermal qualification, and field reliability data that takes months to accumulate. The OEMs who move first will capture an efficiency advantage that their competitors will spend a year or more trying to match.
For data center operators running liquid cooling at scale, the takeaway is simpler. The next time you spec a chiller for a new build or an expansion, ask your cooling vendor what power semiconductors are driving the compressor. If the answer is still silicon IGBTs, you are leaving kilowatts on the floor.