Navitas, Great Wall Launch AI Data Center DC-DC Converter

Navitas Semiconductor and Great Wall Power have introduced a 2.5 kW DC-DC converter designed for the high-density power needs of modern AI data centers. The module uses Navitas’ GaNSense NV6169 power IC to deliver high power density and efficiency in a compact quarter-brick package. It supports the shift toward 400 V DC cabinet architectures, where data centers need to pack more computing into smaller footprints while reducing energy consumption.

Navitas’ 2.5kW GaN-based DC-DC converter achieves record power density and efficiency. Image used courtesy of Navitas Semiconductor

Compact Design with High Efficiency

The converter delivers a power density of 92.36 W/cm³, significantly higher than traditional silicon-based designs, by a factor of up to eight. This allows operators to reduce the space allocated to power delivery and free up more room for compute hardware. The quarter-brick module supports a wide input voltage range from 320 to 420 VDC, which suits 400 V DC cabinet designs now gaining traction in data centers built for AI workloads. Half-load efficiency reaches 97.9%, aligning with current Open Compute Project standards and helping reduce total energy usage in high-throughput environments.

The converter is designed to operate across a range of use cases, including AI servers, telecommunications infrastructure, and industrial equipment. Its high efficiency reduces heat generation, which lowers cooling requirements and simplifies thermal management across rack and cabinet systems. The compact footprint also helps support more modular, scalable power architectures.

GaNSense NV6169: Integrated, Fast, and Efficient

At the design’s center is the NV6169, a 650 V, 45 mΩ GaN power IC with GaNSense technology. The device integrates drive, protection, and lossless current sensing in a single 8x8 mm PQFN package. Lossless current sensing eliminates the need for external shunt resistors, improving efficiency and reducing board hotspots. The IC also features built-in short-circuit protection that responds in just 30 ns, six times faster than discrete alternatives.

The NV6169 is designed for AI data centers and other high-performance applications. Image used courtesy of Navitas

For transient handling, the NV6169 has an 800 V peak rating. Its low profile (0.85 mm) and large thermal pad improve cooling and reduce inductance. Compared to earlier solutions using external current sensing, the NV6169 reduces total on-resistance from 95 mΩ to 45 mΩ in equivalent high-frequency power factor correction circuits. This results in a 0.2% gain in overall system efficiency and simplifies layout and thermal design.

Supporting the Transition to 400 V Architectures

AI data centers increasingly rely on 400 V DC independent cabinet architectures to support rising compute demands. These setups allow higher GPU densities and shorter power paths but require power modules that are both space-efficient and high-performance. This converter supports that transition by delivering 2.5 kW of power in a form factor small enough for dense rack systems with minimal thermal overhead.

While AI is the main driver, the converter also suits telecommunications and industrial sectors, where space and energy budgets are tight. Its design is flexible enough to fit a variety of applications requiring high efficiency and dense power delivery.

The collaboration between Navitas and Great Wall brings together semiconductor innovation and system-level engineering, with a focus on meeting next-generation infrastructure requirements.