EPC unveils two GaN white papers for AI data centers and motor drives

Efficient Power Conversion released two technical white papers Sept. 14 on GaN power conversion for 800 VDC AI data centers and integrated 100 V motor-drive platforms. The papers show how GaN is being used to cut losses, shrink power stages and simplify multi-axis robotics and other precision-motion systems.

Why it matters: - AI data centers are pushing power delivery toward 800 VDC as racks scale from tens of kilowatts to hundreds of kilowatts and megawatt-class clusters. - High-efficiency GaN architectures can reduce conductor size, I²R losses and cooling needs. - Robotics, drones and precision-motion systems need motor-drive platforms that are smaller, more integrated and easier to scale across many axes.

What happened: - Efficient Power Conversion (EPC) published two new technical white papers on Sept. 14, 2026. - The first paper focuses on GaN-based power delivery for AI data centers, from 800 VDC input to point-of-load rails. - The second paper covers EPC23108/09/10/11 100 V integrated GaN power stages for multi-axis motor-drive design. - EPC also released the papers in English and Chinese.

The details: - The AI-data-center paper examines three conversion architectures for 800 VDC to 50 V, 12.5 V and 6.25 V rails. - The paper puts special focus on modular Input-Series, Output-Parallel, or ISOP, LLC converters. - At 800 VDC, a 200 kW rack needs one-sixteenth the current of a 48 V bus. - EPC says that lower current reduces conductor size, I²R losses and cooling requirements. - The paper says dividing the high-voltage input among multiple modules lets low-voltage GaN FETs achieve high-ratio conversion at high switching frequencies. - Practical results in the paper include an 800 V-to-50 V implementation with 98.6% peak efficiency. - The EPC91123 800 V-to-12.5 V converter reaches 98.2% peak efficiency in an 8 mm profile. - A 6 kW 800 V-to-6.25 V architecture reaches 98% peak efficiency and 97% full-load efficiency. - EPC says Gen 7 GaN plus modular ISOP architectures offer a practical path from an 800 VDC backbone to the low bus voltages and very high currents needed by next-generation AI processors. - The motor-drive paper addresses humanoid robots, drones, surgical robotics and precision-motion systems. - EPC’s EPC231xx family integrates high-side and low-side eGaN FETs, gate drive, level shifting, control interface and protection functions into a monolithic 100 V ePower Stage IC. - The EPC91128 through EPC91131 three-phase BLDC/PMSM inverter platforms give engineers a test environment for current capability, sensing, protection, thermal performance and PWM interfaces. - The platforms operate across a recommended 14-80 V range, with 48 V typical. - The platforms support switching frequencies from 20 kHz to 150 kHz. - Complementary PWM versions maximize control flexibility. - Single-PWM-input-pin versions reduce control-line count and wiring complexity. - EPC says that wiring reduction becomes more important as robotic systems scale to dozens of motors. - The company says the EPC231xx family and its evaluation platforms provide a validated starting point that can be replicated across many motor axes. - Jason Zhang, EPC vice president of DC-DC marketing and system engineering, said power conversion must move closer to the processor as GPU power rises. - Zhang said Gen 7 GaN combined with modular ISOP architectures provides a practical path from an 800 VDC backbone to the very low bus voltages and extremely high currents required by next-generation AI processors. - Marco Palma, EPC vice president of motor drive marketing and system engineering, said integration must also simplify control, protection, sensing, thermal management and wiring. - Palma said the EPC231xx family and evaluation platforms give engineers a replicated starting point across multiple motor axes.

Between the lines: - EPC is positioning GaN as a full-system enabler, not just a better switch. - The white papers target two markets where power density, thermal limits and integration pressure are rising at the same time. - The AI paper points to a broader shift away from 48 V distribution in the highest-power racks. - The motor-drive paper suggests the key challenge is moving from discrete power stages to repeatable, highly integrated platforms.

What's next: - EPC is steering engineers toward the two white papers as reference designs for AI infrastructure and multi-axis motion control. - The company’s publication strategy suggests more emphasis on architecture-level guidance as GaN adoption expands. - More information is available through EPC’s announcement and social channels, including the company’s LinkedIn page, Facebook page and X account.