Overview of the Navitas and NVIDIA Collaboration

Navitas Semiconductor and NVIDIA have announced a significant breakthrough in AI data center infrastructure with the development of a 24 kW, 800 VDC-to-12 VDC power delivery board (PDB). Unveiled on October 15, 2026, this advanced platform leverages Navitas’ GaNFast technology to optimize power delivery for next-generation artificial intelligence workloads. The solution is designed to seamlessly integrate with the NVIDIA MGX ecosystem, simplifying the complex power requirements of high-performance computing processors, commonly referred to as xPUs. This collaborative effort highlights a strategic push toward more efficient, high-density power architectures, and the platform is scheduled to be showcased at the 2026 OCP Global Summit in NVIDIA’s AI Factory MGX Wall partner display area.

Technical Architecture and Single-Stage Conversion

At the core of this innovation is a highly efficient single-stage conversion architecture. Unlike traditional designs that rely on a conventional 50 V-to-12 V intermediate bus converter (IBC) stage, the new PDB converts 800 VDC directly to 12 VDC. This architectural shift fundamentally reduces the number of conversion steps, thereby minimizing power losses and decreasing the overall component count. The resulting 12 V output feeds directly into the voltage regulator module (VRM), which subsequently provides the precise point-of-load voltages required by the xPUs. By eliminating the intermediate stage, the design not only improves overall system efficiency and reliability but also significantly reduces board space, streamlining the power distribution network within the compute tray.

GaNFast Power Stage and Component Specifications

The primary-side power stage of the 24 kW PDB is built upon advanced gallium nitride (GaN) technology, specifically utilizing 16 NV6034 GaNFast FETs. Each of these devices is rated at 650 V with an on-resistance of 18 mΩ, delivering the high switching performance necessary for compact, high-power conversion systems. To manage the thermal challenges associated with such high power densities, the FETs are arranged in a stacked half-bridge topology and housed in Navitas’ dual-cooled 8 mm × 8 mm DFN package. This specialized packaging and topology ensure optimal heat dissipation, supporting the rigorous thermal and electrical demands of next-generation AI infrastructure while pushing the boundaries of power density.

Form Factor and Integration for AI Compute Trays

Physical integration is a critical factor in modern AI data center design, and the new PDB addresses this by strictly following NVIDIA’s 400 mm × 60 mm form factor. This standardized footprint allows the board to be integrated directly onto the compute motherboard, positioned in close proximity to the xPUs. Placing the power delivery system near the processors drastically reduces power-delivery losses and parasitic impedances, which is essential for supporting the extreme current demands of next-generation AI chips. Furthermore, the platform targets a remarkable full-load efficiency of 97% and a power density exceeding 2,000 W/in³, simplifying high-power compute tray designs and maximizing the available physical space within the server rack.

Market Context and the Shift to 800V DC Distribution

This new 24 kW board extends Navitas’ growing portfolio of 800 VDC power solutions, building upon the previously announced 20 kW, 800 V-to-6 V power delivery board also developed with NVIDIA. Both platforms underscore a broader industry transition toward high-voltage DC distribution within AI data centers. As AI workloads scale exponentially, the limitations of traditional lower voltage distribution become increasingly apparent, leading to higher current levels and corresponding I²R losses. By adopting 800V DC architectures and leveraging wide-bandgap semiconductors like GaN and silicon carbide (SiC), data center operators can drastically reduce cable sizing, minimize transmission losses, and improve the overall energy efficiency of the facility from the grid interface down to the rack level.

Future Outlook for AI Data Center Power Delivery

Looking ahead, the 800 V-to-12 V PDB adds critical flexibility to Navitas’ broader AI power portfolio. It empowers system designers to select intermediate-bus voltages and voltage regulator architectures tailored to their specific point-of-load requirements. This modularity is vital as AI processor architectures continue to evolve at a rapid pace. The platform strongly supports Navitas’ strategic objective to replace conventional silicon power devices with GaN and SiC technologies across the entire power chain. As the industry moves toward increasingly dense and powerful AI factories, innovations like this 24 kW GaN-based PDB will be instrumental in ensuring that power delivery scales seamlessly alongside compute performance, defining the next generation of sustainable and high-performance data center infrastructure.