Data centers have always required stable and continuous power, but the rapid growth of AI computing is changing the electrical load profile behind that requirement. Higher rack densities, larger GPU clusters, expanding cooling systems, UPS equipment, and increasingly dynamic computing loads can place greater demands on the upstream electrical system.
For this reason, selecting a power transformer for a data center should not begin and end with a capacity such as 2500kVA, 10MVA, or 40MVA. The transformer needs to fit the utility connection, medium-voltage distribution architecture, redundancy strategy, expected load profile, harmonic environment, installation conditions, and future expansion plan.
At HONLE, we manufacture oil-immersed and dry-type transformer solutions across different voltage levels and capacity ranges for power distribution, industrial facilities, infrastructure, and higher-voltage applications. For data center projects, the appropriate transformer depends on where it sits in the electrical system, how much load it carries, and how the facility is expected to expand.
The important question is therefore not simply:
“What transformer capacity do we need?”
It is:
What does the complete data center power system require from the transformer today, and how will those requirements change as computing capacity expands?
Why Are AI Computing Loads Changing Data Center Power Requirements?
Traditional data center loads are already demanding because servers, storage equipment, network infrastructure, cooling systems, pumps, fans, UPS systems, and auxiliary equipment may operate continuously.
AI computing adds another layer.
GPU-based computing clusters can concentrate significantly more electrical demand into a limited area while introducing faster changes in load than many conventional computing environments.
For the power transformer, this affects several areas:
- total required capacity;
- expected operating load;
- short-term load variation;
- thermal performance;
- voltage regulation;
- harmonic conditions;
- redundancy;
- future expansion.
This is why the transformer specification should be developed together with the broader electrical design rather than selected independently.
A data center that expects major AI capacity expansion should communicate that expansion plan to the power transformer manufacturer before the transformer design is finalized.
Where Does the Power Transformer Sit in a Data Center Electrical System?
The exact architecture varies by project, utility connection, facility size, redundancy level, and local electrical standards.
A simplified arrangement may look like:
Utility/Grid → HV or MV Power Transformer → MV Switchgear → Distribution Transformer → UPS/Power Distribution → IT and Cooling Loads
In a large campus, a higher-voltage power transformer may form part of the main grid connection or primary substation.
Further downstream, medium-voltage and distribution transformers can supply different buildings, electrical sections, cooling systems, or IT loads.
Dry-type transformers may also be installed inside electrical rooms where indoor installation, fire planning, ventilation, maintenance access, and space conditions become more important.
This distinction matters because one data center can contain several transformer levels with very different responsibilities.
A transformer serving the main substation should not be selected according to exactly the same criteria as an indoor transformer supplying downstream facility distribution.
The project therefore needs to define:
- grid connection voltage;
- internal MV voltage;
- LV distribution voltage;
- transformer quantity;
- transformer location;
- critical and non-critical load separation;
- redundancy arrangement.
Once this architecture is known, individual power transformer specifications become much easier to define.
How Much Transformer Capacity Does a Data Center Actually Need?
Capacity planning should begin with the load profile rather than an arbitrary transformer rating.
The electrical team normally needs to account for IT loads together with cooling, pumping, ventilation, lighting, controls, auxiliary equipment, and other facility loads.
Future growth also matters.
A facility designed around today’s IT capacity may face expensive electrical modifications if rack density or AI computing capacity increases significantly several years later.
The following is only an illustrative planning example, not a fixed sizing standard:
| Item | Phase 1 | Phase 2 | Future |
| IT Load | 3MW | 5MW | 8MW |
| Cooling + Auxiliary | 1MW | 1.5MW | 2MW |
| Total Active Load | 4MW | 6.5MW | 10MW |
| Assumed Power Factor | 0.95 | 0.95 | 0.95 |
| Reserve | 20% | 20% | 20% |
| Calculated Capacity | 5.1MVA | 8.2MVA | 12.6MVA |
The actual installed capacity may differ substantially once N+1 or 2N redundancy, transformer loading strategy, maintenance conditions, and distribution architecture are considered.
This is exactly why simply telling a supplier that a project needs “a 10MVA transformer” is not enough.
A power transformer manufacturer should understand how the figure was reached and what load the transformer is expected to carry during both normal operation and contingency conditions.
Future AI Expansion Should Be Discussed Before Transformer Production
One of the most important questions for a new data center is:
How much additional computing capacity could be installed without rebuilding the upstream electrical system?
Imagine a campus beginning with a 4MW total facility load but planning to move toward 8MW or 10MW as additional AI halls are commissioned.
If the first phase is designed only around today’s demand, later expansion can affect more than the transformer.
It may also require changes to:
- MV switchgear;
- busbars;
- cables;
- protection settings;
- circuit breakers;
- cooling systems;
- generators;
- UPS capacity;
- utility connection agreements.
Transformer planning therefore needs to consider the complete expansion path.
This does not mean every transformer should simply be oversized.
Excessive oversizing can increase initial investment and may not provide the most suitable operating point for the system.
Instead, the project can consider whether future demand will be supported through larger initial transformer capacity, additional transformer bays, modular expansion, parallel transformers, or another planned electrical architecture.
Providing this information early allows us to discuss the power transformer around both present demand and realistic future operating conditions.
Does a Data Center Need an Oil-Immersed or Dry-Type Transformer?
Both can be used within data center infrastructure, but usually at different locations and for different purposes.
For main substations and larger outdoor electrical systems, an oil-immersed power transformer is often considered where higher voltage levels and larger capacities need to be handled efficiently.
Depending on the grid connection and project architecture, this may include medium- or higher-voltage transformer solutions. For example, larger facilities may use 35kV or 110kV equipment at the upstream level, but these are only examples rather than fixed requirements for every data center.
Dry-type transformers are more commonly considered in indoor electrical rooms and building distribution areas where installation environment, fire planning, maintenance access, ventilation, and oil containment are important considerations.
HONLE also manufactures SC(B) series dry-type transformers for indoor distribution applications where this type of construction is more suitable.
The decision should therefore be based on where the transformer is installed rather than treating oil-immersed and dry-type designs as competing options for every position.
A large data center campus may use both.
How Do UPS Systems and Nonlinear Loads Affect the Power Transformer?
Modern data centers contain substantial amounts of power electronic equipment.
Depending on the architecture, this can include:
- UPS systems;
- server power supplies;
- rectifiers;
- variable-frequency drives;
- cooling equipment;
- electronic power conversion systems.
These devices can introduce nonlinear current characteristics and harmonic distortion into the electrical network.
For transformer specification, this means the project should provide more than total kVA.
It is useful to understand:
- proportion of nonlinear load;
- expected harmonic spectrum or THDi where available;
- UPS topology;
- major power electronic loads;
- cooling-system drive loads;
- expected power factor;
- load balance.
Harmonic currents can contribute to additional transformer losses and heating, so their effect needs to be considered together with the expected load profile.
The solution is not automatically to increase transformer capacity.
Oversizing may help with some thermal effects, but it does not by itself solve every harmonic or voltage distortion issue.
Instead, transformer thermal design, impedance, system configuration, filtering, UPS characteristics, and overall electrical distribution should be considered together.
Why Does Transformer Impedance Matter in a Data Center?
Transformer impedance is sometimes treated as a line on the datasheet, but it directly affects how the surrounding electrical system behaves.
Among other factors, impedance influences:
- prospective short-circuit current;
- voltage drop;
- protection coordination;
- downstream equipment ratings;
- parallel transformer operation.
A lower impedance can help reduce voltage drop but may allow higher fault current.
A higher impedance can limit fault current but can increase voltage variation under load.
That means specifying a power transformer with the lowest possible impedance is not automatically the right approach.
For large data centers, the relationship between transformer impedance and switchgear fault ratings can become particularly important.
A transformer specification therefore needs to be coordinated with the system short-circuit study rather than chosen independently.
If transformers will operate in parallel, their voltage ratio, vector group, impedance, tap position, and other parameters must also be properly coordinated.
What About Voltage Regulation When AI Loads Change Quickly?
Stable downstream voltage matters throughout a data center.
The required transformer regulation arrangement depends on the grid, transformer position, load variation, and overall voltage-control strategy.
Possible arrangements can include:
- fixed ratio;
- off-circuit tapping;
- on-load voltage regulation.
For sites where upstream voltage or operating load changes significantly, an on-load tap-changing arrangement may be considered as part of the system design.
HONLE manufactures voltage-regulated oil-immersed transformer solutions for applications where the output voltage needs to be adjusted while the transformer remains energized. An SZ11-M series transformer is one example that may be considered when the project specifically requires on-load regulation.
However, the need for an on-load tap changer should be determined from actual voltage conditions and system studies.
It should not be specified simply because the facility contains AI servers.
The important information for the power transformer manufacturer includes incoming voltage variation, required output voltage range, system regulation limits, transformer loading profile, and how other voltage-control equipment operates within the network.
N+1 and 2N Redundancy Change Transformer Planning
Redundancy is one of the defining features of critical data center infrastructure.
Common terminology includes N, N+1, and 2N architectures.
In simplified terms:
- N provides the capacity required for the load;
- N+1 adds additional equipment or capacity;
- 2N provides two complete capacity paths.
However, having two transformers does not automatically create a fully redundant power system.
The complete path matters.
For example:
Utility → Transformer → Switchgear → UPS → PDU → Server
If two transformers feed a common downstream component that becomes a single point of failure, transformer duplication alone does not provide complete end-to-end redundancy.
Transformer planning should therefore ask:
- Can one transformer carry the required critical load if another is unavailable?
- Are transformers connected to independent bus sections?
- Can maintenance be performed without interrupting the required load?
- What happens during transformer or switchgear isolation?
- Is the redundancy target N+1, 2N, or another architecture?
- How much spare transformer capacity exists under contingency operation?
These answers affect both transformer quantity and rated capacity.
Why Do Transformer Losses Matter More in 24/7 Data Centers?
A data center may operate continuously throughout the year.
That makes transformer losses more important than they may appear when only the purchase price is considered.
Two main loss categories should be considered:
No-load loss occurs whenever the transformer remains energized.
Load loss varies with electrical loading.
For equipment that remains energized continuously, even relatively small efficiency differences can accumulate over long operating periods.
The project should therefore evaluate transformer losses together with expected loading rather than considering efficiency only at rated capacity.
For example, a transformer operating at moderate load for most of its service life will have a different loss profile from one that frequently operates close to its rated output.
This is another reason why we prefer to understand realistic operating conditions before finalizing a power transformer specification.
Cooling and Temperature Rise Cannot Be Separated from the Site
Transformer thermal performance depends on more than the transformer itself.
Installation conditions matter.
For indoor equipment, the transformer room needs sufficient ventilation and appropriate clearance.
For outdoor installations, ambient temperature, altitude, solar exposure, enclosure arrangement, and local environmental conditions may affect the design.
For data center projects, cooling deserves additional attention because the surrounding facility already contains substantial heat-generating equipment.
When specifying the power transformer, the project should provide:
- maximum ambient temperature;
- minimum ambient temperature where relevant;
- altitude;
- indoor or outdoor installation;
- ventilation conditions;
- enclosure requirements;
- expected continuous loading;
- possible overload conditions.
A dry-type transformer placed inside a poorly ventilated electrical room should not be evaluated using transformer nameplate capacity alone.
The surrounding thermal environment is part of the transformer application.
Fire Safety and Installation Location Also Affect Transformer Selection
Indoor transformer installation introduces different considerations from an outdoor substation.
The project may need to review:
- fire protection strategy;
- transformer room arrangement;
- ventilation;
- access and maintenance clearance;
- noise;
- cable routing;
- enclosure protection;
- oil containment where applicable.
Dry-type transformers are frequently considered for indoor installations because they do not rely on insulating oil in the same way as conventional oil-immersed designs.
Oil-immersed transformers, meanwhile, remain well suited to many outdoor substations and larger upstream power applications.
Rather than asking which transformer type is universally better, the more useful question is:
Which transformer construction fits this particular point in the data center power system?
What Information Should Be Sent to a Power Transformer Manufacturer?
A good transformer inquiry should give enough information to understand both the electrical requirement and the application.
The following example shows how much more useful a structured inquiry can be than providing capacity alone.
| Parameter | Example* |
| Rated Capacity | 2500kVA |
| Voltage Ratio | 35/10.5kV |
| Frequency | 50Hz |
| Impedance | 8% |
| Tap Range | ±5% |
| Maximum Ambient | 40°C |
| Expected Load | 75% |
| Future Growth | 25% |
Illustrative values only. Final parameters must be confirmed according to the actual electrical design, applicable standards, and system studies.
Along with these numbers, we recommend providing information about:
- installation location;
- transformer quantity;
- redundancy architecture;
- load profile;
- UPS system;
- nonlinear load proportion;
- cooling loads;
- short-circuit requirements;
- vector group;
- insulation level;
- applicable IEC, IEEE, ANSI, or local standards;
- testing and documentation requirements.
The more complete the project information is, the easier it becomes to manufacture a power transformer that fits the surrounding power system rather than simply matching a capacity number.
How Does HONLE Support Data Center Power Transformer Projects?
Data center electrical systems can involve several voltage levels, transformer types, and stages of power distribution.
At HONLE, we manufacture a broad range of oil-immersed and dry-type transformers covering different voltage classes, capacities, and application requirements. Rather than limiting a data center project to one standard transformer series, we look at the transformer’s actual position in the electrical system and the operating conditions it needs to support.
For an upstream substation, the project may require a higher-voltage oil-immersed power transformer.
For facility-level medium-voltage distribution, another voltage ratio and capacity range may be more appropriate.
For indoor electrical rooms, a dry-type transformer may better match the installation environment.
Where voltage regulation is a specific system requirement, an on-load regulated transformer can also be considered.
This allows transformer selection to follow the complete electrical architecture instead of starting from a fixed product list.
Before production parameters are finalized, we normally need to confirm the incoming voltage, required secondary voltage, transformer capacity, redundancy strategy, load characteristics, installation environment, impedance, tap arrangement, temperature conditions, applicable standards, and future expansion plan.
As AI computing facilities become larger and more power-intensive, transformer selection increasingly becomes part of long-term infrastructure planning.
The right power transformer should support not only today’s connected load but also the way the data center is expected to operate, expand, and maintain power continuity over its service life.
FAQ
1. What type of power transformer is commonly used in a data center?
It depends on the transformer’s position in the electrical system. Larger outdoor substations may use oil-immersed power transformers, while dry-type transformers are often considered for indoor electrical rooms and downstream building distribution.
2. Can a data center use both oil-immersed and dry-type transformers?
Yes. A large campus may use oil-immersed transformers for upstream power distribution and dry-type transformers further downstream inside data center buildings.
3. Does every large data center need a 35kV or 110kV power transformer?
No. The required voltage level depends on the local utility connection, facility capacity, electrical architecture, and regional grid conditions. 35kV and 110kV are possible examples for larger projects, not universal requirements.
4. How should transformer capacity be calculated for an AI data center?
Capacity should consider IT load, cooling and auxiliary loads, power factor, operating margin, redundancy architecture, and future expansion. The final transformer rating should be verified against the complete electrical system rather than calculated from server load alone.
5. Should a transformer be oversized for future AI expansion?
Some capacity margin can support future growth, but simply oversizing the transformer is not always the most efficient approach. Projects can also plan additional transformer bays, parallel units, modular expansion, or future distribution sections.
6. Do GPU and AI computing loads affect transformer selection?
They can. AI computing environments may have higher power density and more dynamic load behavior, which makes load profile, thermal performance, voltage stability, and future capacity planning more important.
7. Do UPS systems create harmonics that affect transformers?
UPS systems and other power electronic loads can contribute to nonlinear currents and harmonics. Their effect depends on equipment topology and the complete electrical system, so harmonic information should be included when the power transformer manufacturer evaluates the application.
8. What transformer impedance is suitable for a data center?
There is no single impedance value suitable for every data center. Transformer impedance should be coordinated with fault current, voltage drop, switchgear ratings, protection settings, and any parallel transformer requirements.
9. Does a data center need an on-load tap changer?
Not always. On-load voltage regulation may be useful where supply voltage or operating conditions require adjustment without de-energizing the transformer, but the need should be confirmed from the actual network and voltage-control strategy.
10. How does N+1 redundancy affect transformer capacity?
Under an N+1 arrangement, the remaining equipment must still support the required critical load when one unit is unavailable. This can change both transformer quantity and individual transformer capacity.
11. Is 2N always better than N+1 for data center transformers?
Not necessarily. 2N offers a different level of redundancy but also increases equipment, space, complexity, and investment. The appropriate architecture depends on the required availability level and complete electrical system design.
12. Why are transformer losses important in data centers?
Data center transformers may remain energized 24/7. No-load and load losses therefore accumulate continuously and can affect long-term energy consumption and operating costs.
13. What information should be included when requesting a data center power transformer quotation?
At minimum, provide capacity, primary and secondary voltage, frequency, impedance requirements, vector group, tap range, installation environment, load profile, redundancy arrangement, applicable standards, and future expansion expectations.
14. When should a data center contact the power transformer manufacturer?
Ideally, transformer discussions should begin during electrical system planning rather than after every parameter has already been fixed. Early technical communication makes it easier to coordinate capacity, impedance, voltage ratio, redundancy, installation conditions, and future expansion requirements.