April 27, 2026 /SemiMedia/ — Rapid growth in AI servers is pushing a major shift in data center power design, with demand for power semiconductors rising sharply and lead times at some integrated device manufacturers stretching to around 30 weeks, according to industry sources.

As GPU clusters consume more energy, data centers are not only adding more chips but also upgrading how power is delivered. The focus of competition is expanding beyond processors, high-bandwidth memory and advanced packaging to include power conversion and management across the supply chain.

Several global suppliers have recently introduced 800VDC power solutions for data centers, highlighting a move toward higher-voltage architectures. These systems aim to reduce transmission losses and improve overall efficiency.Infineon has said the fast expansion of AI workloads is putting pressure on existing 54V power infrastructure, which is approaching its limits. The company sees centralized 800VDC systems as a way to improve efficiency and reliability. It also pointed to the combined use of silicon carbide, gallium nitride and silicon-based devices as key to future upgrades.

Texas Instruments’ latest results suggest demand is spreading into analog and power segments. The company reported quarterly revenue of $4.825 billion, up 19% year-on-year, and expects the current quarter to reach between $5 billion and $5.4 billion, both above market expectations.

On the supply side, strong demand and tight capacity at mature process nodes are limiting output growth. Industry sources said some power components are now in short supply, and suppliers may struggle to meet all orders in the near term.