On May 22, the 2026 World AI Server Power Supply Conference was held in Shenzhen. Centered on the theme of “Green Computing Power Supply Technology Innovation and Full-Industry-Chain Collaboration,” the conference showcased cutting-edge technologies and innovative achievements in AI computing infrastructure, server power supplies, and high-efficiency power delivery.
Hangzhou Boco Electronics Co., Ltd. was invited to attend and delivered a keynote presentation titled “High-Power, High-Efficiency AI Server Power Supply Solutions,” sharing the company’s technological innovations and application practices in addressing key challenges of AI-era server power supplies, including high power density, high efficiency, and dynamic response performance.
With the rapid growth of AI large-model training, GPU cluster deployment, and intelligent computing centers, server power supplies are evolving from traditional “stable power delivery” toward higher power density, higher efficiency, and faster dynamic response capabilities. Compared with traditional CPU servers, AI servers place much higher demands on power systems. GPU power consumption continues to rise, driving rapid increases in power density within nearly unchanged rack dimensions, while highly dynamic AI workloads and complex operating conditions such as EDPP require significantly enhanced transient response performance.
To address these industry trends, Boco Electronics continues to advance innovation in high-power-density, high-efficiency, and high-dynamic-performance power supply technologies. Its CRPS 3200W AI server power supply adopts a Titanium-level efficiency design with a power density of 100W/in³. Based on the existing platform, the company can further upgrade the output power of 3200W 12V CRPS power supplies within the same form factor. For 54V architectures, the platform can be upgraded from 2650W to 5500W, achieving a power density of 150W/in³ to meet the growing power demands of AI computing systems.

In hardware design, Boco Electronics utilizes a highly integrated modular architecture that deeply couples the high-voltage and low-voltage sides, optimizing both power density and thermal performance. By integrating advanced semiconductor devices such as silicon carbide (SiC) and gallium nitride (GaN), the company further improves switching frequency and system efficiency, providing stable and efficient power support for AI computing platforms.
To overcome the challenges posed by highly dynamic AI server workloads, Boco Electronics has also achieved breakthroughs in control technologies. Traditional LLC voltage-mode control solutions often struggle to respond effectively to rapid GPU load fluctuations. In response, the company introduced an innovative “charge control + current-mode feedforward control” technology, which incorporates real-time output current into the control system. Through current compensation and dynamic optimization, the solution significantly improves loop bandwidth and transient response performance, effectively reducing voltage fluctuations under EDPP conditions and enhancing power stability under high-dynamic-load scenarios. Currently, this technology has been successfully applied to single-phase LLC and high-power three-phase LLC architectures and has already entered commercial deployment.
Boco Electronics has now established a comprehensive AI server power supply solution portfolio covering multiple platforms, including 12V and 54V architectures. These solutions are widely applicable to high-density GPU clusters, AI training servers, intelligent computing centers, and next-generation data centers, delivering more efficient, stable, and reliable power support for global AI computing development.
As AI computing infrastructure continues to evolve, server power supplies are becoming a critical factor influencing data center efficiency and operational stability. Looking ahead, Boco Electronics will continue to deepen technological innovation in high power density, high efficiency, and high dynamic response, accelerating the iteration and large-scale deployment of AI server power supply products.