The reference design was presented at the Open Compute Project summit this week and is intended for next-generation infrastructure including Nvidia's Vera Rubin. While Navitas stated the board is a collaboration rather than a volume order, the company noted that a 1.5 percentage point efficiency increase could reclaim approximately 15 MW of power for every 1 GW of AI capacity to support additional compute.
Tuesday, October 20, 2026
Story
Navitas and Nvidia Build 24 kW Board to Cut AI Power Stage
AI8 days ago
A new 800V to 12V direct conversion board developed by Navitas and Nvidia aims to reduce electricity waste by eliminating an intermediate power stage in AI racks. The 24 kW board, designed for the MGX ecosystem, removes the standard 50V intermediate bus to bring power closer to the GPU and minimize heat loss, according to the Navitas CEO. Built with 16 NV6034 GaN FETs, the 400 by 60 mm hardware targets 97% efficiency at full load and a power density exceeding 2,000 W/in³.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.