800 VDC Data Center Power: Siemens

Introduction

On August 14, 2026, Siemens and Reinhausen announced the joint development and planned industrialization of a modular solid-state transformer.
The proposed system would convert grid input at up to 36 kV directly into a stable 800 VDC output.
Fewer conversion stages could provide a simpler power path, but they also raise new protection and maintenance questions.
In addition to efficiency, decision-makers must assess fault tolerance, bypass options, redundancy and the ability to perform interventions while the facility remains operational.

According to a statement issued by Siemens and Reinhausen on August 14, 2026, the two companies are jointly developing and planning to industrialize a modular solid-state transformer for powering AI data centers. The concept would provide a stable 800 VDC output directly from an input voltage of up to 36 kV, reducing the number of conversion stages found in conventional multistage AC/DC architectures.

The announcement does not constitute full operational validation of a complete, production-ready data center reference architecture. From an engineering perspective, however, it indicates an important direction: medium-voltage input, power-electronic conversion and high-voltage DC distribution may become increasingly integrated.

Fewer conversion stages do not, by themselves, demonstrate higher availability. Protection, redundancy and maintainability must be assessed across the entire power path as one system.

Who is this analysis for?

This article is intended for data center operators, facility managers, electrical designers, technical procurement teams, investors, colocation providers and decision-makers responsible for critical infrastructure.

At 800 VDC, protection assumptions inherited from AC systems cannot be applied automatically. DC interruption, fault selectivity and arc risk require separate validation.

The technical focus is not limited to theoretical conversion efficiency. The central question is how the power path of an 800 VDC data center can be aligned with the required availability model, how faults can be localized and under what conditions the system can be operated safely throughout its lifecycle.

What does the announced architecture mean?

The technology can be evaluated most safely in a controlled pilot environment with documented fault tests and a predefined recovery plan.

Conventional data center power systems may contain several successive levels of transformation, switching and AC/DC conversion. The announced approach aims to generate an 800 VDC output directly from a grid input of up to 36 kV through a modular solid-state transformer.

In this context, a solid-state transformer is a system based on power-electronic conversion. It should therefore not be treated as a conventional transformer: converter modules, controls, cooling, protection logic and any backup energy sources collectively determine its actual operating behavior.

Reducing the number of conversion stages may decrease the complexity of the power path. However, this does not automatically make the overall system simpler. Instead of using separate devices, more functions may be concentrated within an integrated platform, making it necessary to assess common failure points and maintenance dependencies.

Why does this matter for data centers now?

High-density AI infrastructure is driving larger amounts of electrical power closer to IT loads. In this environment, every conversion stage adds space requirements, heat load, monitoring points and maintenance tasks.

For this reason, 800 VDC distribution is not merely a new voltage level. It may affect switchgear, busbars or cables, rack-level conversion, metering, the grounding concept, fire risk assessments and operator workflows.

Technical decision-makers must determine where the 800 VDC power path should end, where rack-level conversion should take place and how far unified vendor or system integrator responsibility can reasonably extend.

What does this mean from an operational perspective?

The most important question concerns post-fault behavior. Operators need to understand which load circuits would be affected by the failure of a module, a DC section, a control unit or a cooling component. Modularity delivers a genuine availability benefit only if the failed unit can be isolated and the remaining capacity has been demonstrated to continue supplying critical loads.

The operational assessment should cover at least the following areas:

module failure and automatic load redistribution;
control or communication failure;
loss of auxiliary power or cooling;
a DC-side short circuit and its selective isolation;
loss of redundancy during planned maintenance;
startup, shutdown and de-energization sequences;
return to normal operation after a partial or complete outage.

Condition monitoring is also essential in a power-electronic system. Alarms should not merely indicate a fault; they should also support root-cause identification and safe intervention.

Protection, selectivity and DC interruption

DC fault handling differs from established AC practices. Direct current has no natural current zero in each cycle, so interruption, arc extinction and adequate isolation distances require specific design attention. Protection devices must be suitable for the applicable voltage, expected fault current and polarity conditions.

The power electronics within a solid-state transformer may themselves influence the magnitude and time profile of the fault current. Consequently, selectivity cannot be verified using only conventional short-circuit assumptions. Designers need a coordinated assessment of manufacturer models, protection characteristics and every relevant system operating state.

The method of galvanic isolation, ground-fault detection, insulation monitoring, manual isolation and safe de-energization must be validated separately. Their availability and implementation cannot be inferred from the 36 kV-to-800 VDC conversion function alone.

Redundancy, bypass and maintainability

For an integrated converter platform, assessing only the number of modules is not sufficient. Common controls, cooling systems, busbars, input and output switchgear, and maintenance access must all be mapped. Any of these elements could limit the actual resilience of a system nominally designed around an N+1 or another redundancy model.

Bypass design may be particularly complex when the normal power path converts medium-voltage AC to 800 VDC. The design must define which alternative power path can provide a compatible voltage, what transfer time is acceptable and how protection coordination will be maintained during bypass operation.

Maintainability must also be verified at the physical level. It is not enough for a module to be replaceable in principle. The system must provide suitable access, a lifting route, isolation points, protection against backfeed and clear work instructions.

Common mistake

A common design mistake is to equate a shorter conversion chain automatically with lower risk. Fewer devices may indeed mean fewer interfaces, but greater integration can expand the impact of an individual component or shared subsystem.

It is also risky to transfer AC protection logic directly to the DC side or to evaluate rack-level compatibility using nominal voltage alone. Transients, the grounding system, connection sequences and manufacturer interfaces are equally important.

Recommended next step

The technology should first be assessed through a project-specific feasibility study. The initial task is to define the complete single-line diagram and the system boundaries of the power path, including backup supply, energy storage and rack-level conversion.

This can be followed by a failure mode and effects analysis, a selectivity study, modeling of maintenance operating states, and verification of protection and control interfaces. For a new architecture, integrated site testing should be planned in addition to factory acceptance testing, particularly for module failure, DC short circuits, loss of communication and bypass operation.

In an existing live data center, phased deployment may reduce transition risk. An isolatable load block or pilot environment can provide an opportunity to validate operation, monitoring and maintenance processes without modifying the entire critical power path at once.

Conclusion

Based on the Siemens and Reinhausen announcement, direct conversion from medium voltage to 800 VDC represents one possible new direction for data center power systems. Fewer conversion stages are an attractive design objective, but the decision cannot be based solely on the length of the power path.

Digital Technologies’ engineering approach is to assess such architectures as complete systems—from protection and redundancy to cooling, monitoring, bypass and safe maintainability. The true project value depends on whether the new system can demonstrably support the required availability, controlled fault recovery and sustainable operation throughout its lifecycle.

Sources

Siemens press release, August 14, 2026.
Reinhausen announcement concerning the development partnership, August 14, 2026.
Transformer Magazine report on the Siemens–Reinhausen development.