The Rise of 800V HVDC Architecture Drives Growing Demand for Third-Generation Semiconductors

2026/06/17

TechNews
June 15, 2026
Author: Su Tzu-yun

As the scale of AI data centers continues to expand, industry competition is no longer limited to the number of GPUs or the level of computing power. It is gradually evolving into an efficiency race focused on “how much AI computing power can be generated per unit of electricity.” Episil noted that as the power consumption of a single data center moves toward the scale of hundreds of megawatts, or even gigawatts, reducing energy loss during power conversion has become a key issue in AI infrastructure.

From the recently discussed 800V high-voltage direct current (HVDC) architecture to compound semiconductor technologies such as silicon carbide (SiC) and gallium nitride (GaN), these are all seen as important keys to improving energy efficiency.

At a power semiconductor forum hosted by SEMI, Episil stated that traditional data centers require multiple stages of power conversion between the power grid and GPUs. Each stage results in energy loss, and most of the lost energy is eventually converted into heat. Additional electricity must then be consumed by cooling systems such as air conditioning and liquid cooling to reduce temperatures.

For AI data centers facing power limitations, Episil gave the example that if overall power conversion efficiency improves by about 10%, it may become possible to deploy more GPUs under the same power supply conditions, further increasing AI token output capacity. Therefore, reducing power loss and improving energy efficiency have become important development directions for next-generation AI infrastructure.

Episil stated that as AI infrastructure continues to scale up, market demand for high-efficiency power components is also increasing. The industry is currently moving toward 800V HVDC architecture to reduce the number of power conversion stages and lower transmission losses, while SiC and GaN are key materials supporting the next generation of power architecture.

Among them, GaN is suitable for high-frequency, high-efficiency power conversion applications, while SiC has advantages in high-voltage and high-power fields. These materials are expected to be widely used in AI data centers, electric vehicles, humanoid robots, low-Earth orbit satellites, industrial power supplies, and other markets.

In addition, Episil held its shareholders’ meeting on the 12th. Chairman Hsu Chien-hua stated that with the rapid development of AI infrastructure, generative AI, and AI agents, demand for high-performance computing, power management, and high-speed transmission is rising simultaneously. The company’s three major business segments—silicon epitaxy, compound semiconductors, and silicon photonics—are all expected to benefit.

Among them, the compound semiconductor business has gradually entered the supply chains of international customers. In silicon photonics, germanium-silicon (GeSi) epitaxy products have entered the volume production stage. Their revenue contribution is expected to reach 5% by the end of 2026, with the potential for multiple-fold growth in 2027.

【Disclaimer】
The content of this article only represents the author's personal views and has nothing to do with Creating.
The content, text description and originality have not been verified by this site. This site does not make any guarantee or commitment to this article and all or part of the content, authenticity, completeness, and timeliness. It is for readers' reference only. Please verify the relevant content yourself.

Creating Nano Technologies, Inc.
59 Alley 21 Lane 279, Chung Cheng Road, Yung Kang City, Tainan, TAIWAN
TEL:886-6-2323927 FAX:886-6-2013306 URL: http://www.creating-nanotech.com

BACK TO LIST