Siemens and Reinhausen Develop 36 kV Solid-State Transformers for AI Data Centers

Siemens and Reinhausen have entered into a strategic partnership to develop and manufacture a next-generation solid-state transformer (SST) tailored for AI data centers. The collaboration addresses the critical need for more efficient power delivery as artificial intelligence workloads push traditional electrical infrastructure to its limits.

The new SST is engineered to handle grid voltages of up to 36 kilovolts AC, converting them into a stable 800-volt DC output delivered directly to rack-level infrastructure. This architecture is designed to support high-density AI racks by reducing the number of conversion stages in the power chain. Reinhausen is contributing specialized expertise in power electronics, voltage regulation, and power quality to ensure the system can handle the rigorous demands of modern compute clusters.

The AI Power Challenge

This development comes as AI workloads create unprecedented power demands, characterized by extreme rack densities and sudden load spikes. Traditional power infrastructure often relies on multiple AC-to-DC conversion stages, which introduce significant energy loss and increase the physical bulk of the equipment. Reinhausen noted that data centers supporting AI are among the industry's most demanding consumers, where abrupt spikes in power usage can strain not only the servers and storage but also the cooling systems.

Breaking the Power Wall

As AI compute continues to scale, the industry is facing a "power wall" where electrical delivery becomes a primary bottleneck for growth. By shifting toward an 800 VDC standard and utilizing solid-state transformers, operators can significantly increase power density and overall efficiency. This transition reduces the physical footprint of power equipment and lowers operational costs (OPEX) by minimizing heat waste. Such improvements are considered critical for the long-term sustainability and scalability of massive AI clusters, which require immense amounts of energy to maintain performance.

Future Outlook

Industry observers are now watching how this 800 VDC architecture will be integrated into existing data center designs and whether it will drive a broader shift toward DC-native power distribution. While the partnership focuses on the transformer and rack-level delivery, the broader impact on the energy efficiency of AI training and inference remains a key metric for the industry. Further details on the deployment timeline and specific efficiency gains are expected as the manufacturing phase progresses.