AI Power Demand Accelerates 800V HVDC, GaN Adoption

Rapid growth in generative AI and high-performance computing is pushing data-center power density beyond the limits of conventional low-voltage distribution. As individual rack demand moves from tens or hundreds of kilowatts toward 1 megawatt, higher currents increase conduction losses, cooling requirements and the amount of copper needed. An 800V high-voltage direct-current architecture addresses those constraints by raising voltage, reducing current and potentially eliminating conversion stages, making it a key building block for next-generation AI factories.

The transition is prompting suppliers to redesign power supplies, conversion modules and semiconductor components around 800V HVDC. Gallium nitride, or GaN, offers faster switching and higher power density, while silicon carbide, or SiC, is suited to high-voltage and high-power stages. Their adoption could shift industry value away from conventional silicon devices toward advanced power semiconductors, packaging and thermal-management systems, creating new opportunities across the AI power ecosystem as rack capacity approaches the megawatt scale.