Efficient Power Conversion (EPC), the US-based semiconductor company specialising in gallium nitride (GaN) power devices, has developed a new power converter designed to accelerate the shift to 800-volt direct current (VDC) distribution systems in artificial intelligence data centres.
The company’s latest innovation, a 6 kW 800 V-to-12.5 V converter using an Input Series Output Parallel (ISOP) topology, aims to meet the growing power demands of so-called “AI factories”, which are expected to require megawatt-scale rack power delivery systems.
EPC said the GaN-based design offers higher power density and efficiency than silicon-based alternatives. Occupying less than 5,000 mm² and standing only 8 mm high, the device is suited to space-constrained AI boards. The converter efficiently steps down 800 VDC to 12.5 VDC close to the load, reducing bussing losses and improving overall system performance.
By moving from alternating current (AC) to 800 VDC at the rack level and converting directly to 12.5 V at the board, EPC’s technology removes multiple conversion stages. The company said this simplification improves scalability and energy optimisation for future data centre infrastructure.
“GaN is an essential technology for the 800 VDC ecosystem,” said Alex Lidow, Chief Executive of EPC. “Our collaboration with NVIDIA is focused on developing compact, efficient, and cost-effective board-level conversion to power future AI factories at gigawatt scale.”
EPC, which positions itself as a leader in enhancement-mode GaN (eGaN) power management, produces transistors and integrated circuits that outperform traditional silicon-based metal–oxide–semiconductor field-effect transistors (MOSFETs). Its technologies are used in applications ranging from DC–DC converters and lidar systems to electric mobility, robotics, drones, and low-cost satellites.
