GE Vernova Announces a Medium-Voltage UPS: When Is MV-Level Protection Justified Over LV UPS Blocks?

Introduction

• On August 24, 2026, GE Vernova announced a medium-voltage UPS system designed for data centers.
• The system can be installed on the MV network between the power source and critical loads.
• The concept is intended to protect large load blocks and mitigate rapid power fluctuations.
• The key engineering considerations extend beyond capacity to fault boundaries, selectivity, bypass arrangements, and maintainability.

GE Vernova's announcement brings renewed attention to a key question for high-capacity data centers: how far should critical loads continue to be protected by low-voltage UPS blocks, and when might moving protection to the medium-voltage level be justified? According to the company, the solution integrates power electronics, controls, and energy storage, with deliveries expected to begin in mid-2027.

The announcement alone does not demonstrate that an MV UPS is the better choice for every large data center. A meaningful assessment requires comparison of the complete electrical architecture, operating states, and consequences of potential failures.

Who is this relevant to, and why?

The topic is relevant to data center operators, facility managers, electrical designers, project owners, technical procurement teams, colocation providers, and executives responsible for critical infrastructure. It may be particularly important for new high-capacity facilities and expansion projects where the number, footprint, and switching complexity of conventional LV UPS blocks are increasing rapidly.

The technical focus is not to evaluate a specific vendor's product, but to understand how system behavior changes when power conversion and energy storage become part of the medium-voltage distribution system.

What did GE Vernova announce?

According to the available announcement, the MV UPS is positioned on the medium-voltage network between the power source and critical loads. The system uses integrated power electronics, controls, and energy storage to protect downstream load blocks.

The objective is not limited to conventional ride-through capability. It also includes mitigating the rapid power fluctuations associated with AI infrastructure. This matters because changes in high-density computing loads can affect multiple parts of the electrical chain, from incoming supplies and switchgear to transformers and the controls for backup power sources.

As deliveries are reportedly expected to begin in mid-2027, the current period is primarily suitable for technical preparation, evaluation of reference architectures, and project-specific risk analysis.

MV UPS or LV UPS blocks?

In a conventional data center design, the UPS is often installed after voltage transformation on the low-voltage side and divided into several smaller blocks. This allows capacity to be aligned with data halls, technology units, or redundant A and B power paths.

By contrast, an MV-level approach can consolidate a larger load range upstream of the low-voltage feeders. This may simplify some higher-level power paths. However, as the protected block becomes larger, the potential impact area of a failure or planned intervention may also increase.

The two architectures differ meaningfully on five points.

Protected load. An MV UPS consolidates a single larger, centralized block, whereas an LV block design protects smaller, distributed units.

Fault boundary. At MV level the fault boundary may be wider; with LV blocks a failure is typically easier to localize.

Expansion. An MV UPS aligns with large capacity increments, while an LV block design supports more gradual, block-by-block expansion.

Maintenance. The MV solution requires advanced medium-voltage expertise; LV blocks mean more units to manage, but a smaller intervention area for each.

System coordination. At MV level, protection, transformation, and bypass must be coordinated together; on the LV side, more switching and paralleling points may arise.

This list is not a universal ranking. The appropriate architecture depends on the load profile, site topology, redundancy model, and operating organization.

What does this mean for operations?

Selectivity and fault boundaries

Integrating an MV UPS may change fault-current conditions and protection coordination. Engineers must verify that, in the event of a downstream fault, the protective device nearest to the fault operates while the rest of the load block remains energized wherever possible.

The assessment must cover normal operation, energy-storage operation, bypass mode, and operation from backup sources. Different sources can produce different fault levels and dynamic behavior, so a single static calculation may not represent every relevant operating state.

Bypass and maintenance strategy

A bypass is not merely an alternative conductor path. It determines how power can be maintained during UPS maintenance, a control failure, or intervention on the power electronics. The bypass source, switching logic, interlocks, protection, and the grid disturbances to which the load may be exposed in bypass mode must all be assessed.

A separate plan is required for complete isolation. If a main unit can only be disconnected through an extensive shutdown, nominal redundancy does not necessarily translate into maintainable redundancy.

Managing dynamic loads

When evaluating rapidly changing AI loads, recording peak power alone is not sufficient. The rate, frequency, and duration of load changes also affect the power electronics, energy storage, generator controls, and transformer loading.

The potential buffering role of an MV UPS should therefore be modeled and tested using realistic load profiles. The dynamics of computing workloads, the electrical system, and the cooling infrastructure should be assessed in a coordinated manner.

Common mistake: assuming that a larger block is automatically simpler

A common design mistake is to equate fewer UPS units with lower complexity. The equipment count may indeed decrease, but system risk does not disappear. It may shift to the MV switchgear, common controls, bypass system, energy storage, or downstream transformation.

It is similarly misleading to compare only efficiency or footprint. The decision should also account for the protection system, spare parts, service capabilities, testing requirements, isolation time, and the business impact of a potential outage.

When might MV-level protection be justified?

An MV UPS merits detailed evaluation when a project involves large, clearly defined load blocks, extensive medium-voltage distribution, and rapid power fluctuations. It may also be relevant where multiplying UPS units on the LV side would require substantial floor area, cabling, and switching infrastructure.

An LV block-based design may remain preferable where gradual capacity deployment, smaller fault boundaries, tenant-level separation, or the existing operating model supports that approach. In brownfield environments, MV UPS integration may also be constrained by available switchgear space, cable routes, protection systems, and permitted shutdown windows.

Recommended next step

The first step should be to develop one MV UPS concept and at least one LV UPS block concept against the same load and availability requirements. The comparison should include:

1. single-line diagrams for every significant operating state; 2. failure modes and effects analysis; 3. selectivity and short-circuit studies; 4. bypass, isolation, and maintenance scenarios; 5. expansion stages and temporary configurations; 6. the planned scope of FAT, SAT, and integrated system testing; 7. lifecycle costs, including service and spare parts.

A migration plan is also necessary when upgrading a live data center. In addition to the final configuration, every transitional switching state must be documented because temporary configurations during construction often operate with reduced redundancy.

Conclusion

GE Vernova's August 24, 2026 announcement indicates that, as data center loads become larger and more dynamic, uninterruptible power protection is becoming a strategic design issue at the MV distribution level. An MV UPS may provide integrated protection for large load blocks, but only if the fault boundaries, bypass paths, and maintenance conditions associated with centralization are properly managed.

Digital Technologies recommends that the choice be preceded by a vendor-neutral system assessment. The decisive question is not whether the MV or LV solution is more modern, but which architecture provides demonstrable selectivity, maintainability, scalability, and operational continuity throughout the facility's lifecycle.

Sources

The factual news basis for this article is GE Vernova's August 24, 2026 announcement and related reporting by Data Center Dynamics and Switchgear Magazine.