AI Data Centers Fuel Power-Chip Boom

Photo Image
전력반도체. (사진=게티이미지)

Artificial intelligence (AI) data centers are emerging as a major new growth engine for the power semiconductor industry. As AI accelerators become more powerful and power consumption per server rack rises, data-center bottlenecks are expanding beyond the availability of graphics processing units (GPUs) to include power delivery and conversion efficiency.

According to industry sources on Aug. 9, major power semiconductor companies are strengthening their AI data-center businesses by expanding production capacity and securing long-term supply agreements with major customers.

Infineon, the world's largest power semiconductor supplier, said it has either signed or is negotiating multiyear agreements worth billions of euros with key AI data-center customers. Onsemi also prioritized some production capacity and products for AI data-center customers over automotive and industrial clients in the second quarter.

As AI accelerators consume more power, the importance of rack-level power delivery and power-conversion efficiency is increasing. Power-delivery architectures are also moving toward higher voltages, which is expected to accelerate the adoption of compound power semiconductors such as silicon carbide (SiC) and gallium nitride (GaN).

Investment in the sector is growing as well. In February, Infineon raised its annual investment plan by EUR 500 million, from EUR 2.2 billion (3.6 trillion KRW) to EUR 2.7 billion (4.4 trillion KRW). The additional investment will be used to expand manufacturing capacity for power-supply solutions designed for AI data centers.

STMicroelectronics has also signed a multiyear agreement worth billions of dollars with Amazon Web Services (AWS) to support cloud and AI data-center infrastructure.

DB HiTek, a South Korean power semiconductor foundry, plans to invest about 2 trillion KRW over the next five years to expand high-voltage power semiconductor capacity and commercialize SiC and GaN technologies.

The company aims to broaden its portfolio into compound power semiconductors to address demand from AI data centers and electric vehicles.

Market research firm TechInsights forecasts that the AI data-center power semiconductor market will grow from USD 5 billion this year to USD 14.3 billion by 2031, representing a compound annual growth rate of 24.6%.

· This article was translated using AI and was published after final review by the reporter.