Artificial intelligence is driving a fundamental transformation of data center infrastructure. The industry is moving away from traditional power supply architectures toward more efficient and scalable solutions capable of supporting the next generation of AI applications. Solid-state transformers are a key component of this transformation. They enable higher efficiency and direct conversion of AC voltage to DC voltage—from the medium-voltage grid to the server rack. They also take up less space and are more cost-effective. Thanks to their ability to respond to load changes within milliseconds, they can seamlessly compensate for the extreme power fluctuations caused by AI training applications. This helps prevent downtime and reduce costs.

A medium-voltage-capable SST offers a compact and efficient alternative to conventional multi-stage AC/DC conversion architectures, thereby enabling a more efficient power supply for AI infrastructures with high power density.

Reinhausen is recognized as a global leader in voltage regulation technology for power transformers and has decades of experience in the areas of power quality and grid stability. The company has also developed extensive expertise in medium-voltage power electronics and related components, which form the core of modern SST architectures.

Siemens combines extensive expertise in the fields of power electronics, electrical power supply solutions, grid protection, automation, and digitalization. When implementing SSTs, Siemens leverages its expertise in grid integration, power conversion, and protection and control technology to deliver next-generation energy systems for AI data centers.

Key features of the solution include modularity, scalability, and high availability. The design enables operation at voltages of up to 36 kilovolts and, thanks to its modular structure, can also be used for lower grid voltages. This allows the architecture to be flexibly adapted to different grid connection conditions. The SST provides the 800-volt direct current (VDC) output voltage that will be required in AI data centers in the future.