The artificial intelligence (AI) semiconductor market is shifting from high-bandwidth memory (HBM) to power semiconductors. As AI data centers grow to gigawatt (GW) scale, the demand for silicon carbide (SiC) and gallium nitride (GaN) power semiconductors is rapidly increasing to reduce power loss. Industry analysts predict that, following memory, power semiconductors will emerge as another beneficiary industry in the AI era.
According to the semiconductor industry on July 10, the expansion of investments in AI data centers by global tech giants is accelerating the growth of the power semiconductor market. AI servers are equipped with thousands of GPUs, requiring several times more power than traditional servers. To ensure stable power supply while minimizing losses, the adoption of high-efficiency power semiconductors is essential.
SiC and GaN are particularly efficient in high-voltage and high-temperature environments compared to traditional silicon (Si) power semiconductors, generating less heat. Their application is expanding in data center power supply units (PSUs), power conversion devices, uninterruptible power supplies (UPS), and cooling systems, positioning them to directly benefit from increased investment in AI infrastructure.
Market research firms forecast that the global power semiconductor market will continue to grow at double-digit rates in the coming years due to the proliferation of AI data centers. While electric vehicles have driven the SiC market thus far, AI data centers are expected to become a new growth axis.
Domestic companies are also accelerating their responses. DB HiTek is pushing to establish an 8-inch SiC process and expand its power semiconductor business. LX Semicon is preparing to enter the automotive and industrial power semiconductor markets, moving beyond display driver chips. YesTech is also expected to benefit as it expands its thermal processing equipment and semiconductor equipment business necessary for power semiconductor manufacturing.
The semiconductor equipment industry believes that the expansion of AI investment will not be limited to memory. If production line expansions for power semiconductors and new fab investments occur, there is a strong possibility that the equipment and materials market will also grow.
Industry insiders anticipate that future AI competitiveness will extend beyond GPU performance to include power efficiency. The more power GPUs consume, the more crucial it becomes to convert and supply that power efficiently, which will impact data center operating costs. Jensen Huang, CEO of NVIDIA, has recently emphasized the importance of power infrastructure alongside computing in the era of AI factories.
An industry official stated, "In the early stages of AI investment, securing GPUs and HBM was key, but moving forward, power efficiency will be a determining factor in data center competitiveness. SiC and GaN power semiconductors are likely to emerge as new core components of the AI ecosystem."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.