A 100 MW data-center requirement is the beginning of a sizing study, not a turbine specification. Before comparing equipment, establish whether that number describes IT demand, the whole facility or power delivered at a particular electrical bus. Then define the outage the plant must withstand.
The purchase decision depends on the result. A configuration with more installed megawatts can leave less usable capacity after one unit trips. A package that meets demand in mild weather can require a different unit count at the site's limiting condition.
And a turbine ready to ship can still arrive before its site is ready to operate it. The operating date must come from the complete project schedule.
This guide develops the load boundaries, capacity calculations, phased operating plan and sourcing brief needed to compare real offers. Every worked scenario is hypothetical and explicitly identified.
Key Takeaways
- Define whether the project number describes IT load, facility demand or delivered generation.
- Compare net output at the required site conditions and electrical boundary.
- Check capacity after the specified outage, then assess controls, transient response and shared infrastructure separately.
- Select a turbine configuration alongside fuel, permitting, electrical and commissioning work.
Turn the capacity target into a decision sequence#
A useful sizing study moves through three decisions in order: the demand to serve, the events the plant must withstand, and the equipment that can satisfy both. Reversing that order creates avoidable redesign. Buying an attractive turbine first and asking the electrical team to make it fit later transfers the uncertainty into the rest of the project.
Create a project basis that identifies the facility bus, the operating phases and the weather cases used for capacity planning. Give that document a revision number. Every vendor should quote against the same revision, and every later change should be visible in the comparison. An increase in cooling demand or a change in the required outage criterion can invalidate an earlier shortlist even if the IT target stays unchanged.
For each phase, distinguish committed load from potential expansion. If future buildings are uncertain, ask for an expansion concept rather than treating all future demand as an immediate procurement requirement. Reserve the necessary access, electrical interfaces and site space in that concept. This preserves an option without confusing it with capacity already purchased and commissioned.
The project basis should also state what counts as successful operation.
Is all campus load protected, or may selected loads be shed? Must the facility continue through scheduled generation maintenance? Does the design require survival of a second event during that maintenance? These are owner and engineering decisions; an equipment supplier should not choose them implicitly by offering a particular unit count.
Establish what the 100 MW includes#
Start with a load schedule that names each boundary. IT equipment, cooling, electrical losses and other facility services may sit inside the data-center demand figure. Fuel compression and generation auxiliaries may sit outside it. Counting a load twice is as misleading as leaving it out.
For an illustrative early screen, suppose 100 MW is the IT demand and the design team assumes a facility-to-IT power ratio of 1.25 at the chosen operating point. The resulting facility demand is 125 MW. This ratio is an assumption for this example; an annual energy-based PUE value should not automatically be used as a peak-load sizing factor.
If the engineer instead specifies 100 MW at the campus supply bus, begin with that requirement. Obtain an explicit statement of which losses and auxiliary loads must be supplied before power reaches that bus.
The EIA's explanation of net generation distinguishes electricity supplied from electricity consumed operating a power plant. Apply the same discipline to each offer: the label “net” is useful only when its boundary is defined.
Calculate surviving capacity before selecting a model#
For identical units in a simplified single-unit-outage screen:
Surviving capacity = (installed units − 1) × dependable net output per unit.
This simplified expression assumes identical units can each deliver the stated net output in the outage case. Shared auxiliary loads, changes in losses and other common constraints must be accounted for separately unless already included in those ratings.
Assume a requirement of 125 MW delivered to the campus bus. The following figures are hypothetical site-net ratings, not specifications for named products.
| Configuration | Installed site-net capacity | Capacity after one unit is unavailable | Meets the 125 MW capacity screen? |
|---|---|---|---|
| Five units at 32 MW | 160 MW | 128 MW | Yes, with 3 MW capacity headroom |
| Four units at 43 MW | 172 MW | 129 MW | Yes, with 4 MW capacity headroom |
| Three units at 60 MW | 180 MW | 120 MW | No |
The last option has the most installed capacity but fails this particular outage test. This is why total nameplate MW is a poor substitute for configuration analysis.
A passing calculation does not establish uninterrupted service. The engineering study must determine whether the remaining machines are already online, how they respond to a unit trip, what bridges the transient, and whether an outage of shared switchgear, fuel equipment or controls defeats the arrangement. Maintenance-plus-failure criteria may require another configuration.
Examine the systems shared by multiple units#
Review the capacity table alongside a simplified electrical single-line diagram and a fuel-system diagram. Mark which components are dedicated to a turbine and which are common to several units. Include startup supply, fuel compression, cooling, water treatment where applicable, controls communications and the connection to the campus distribution system.
A practical review question is: “Which outage could prevent two or more otherwise healthy turbines from supplying the load?” Record the answer as an engineering action. It may lead to separation, additional equipment, alternate operating procedures or a different accepted risk. Counting turbines alone cannot answer it.
Also investigate a loss of the preferred fuel supply. Do not infer an alternate-fuel capability from the turbine family name. The offered configuration, storage, treatment, controls, operating approvals and test plan must support the proposed arrangement. If an alternative is not defined, record the fuel interruption as an unresolved dependency.
This review should include the owner, electrical engineer, mechanical engineer and controls specialist. Each sees a different part of the failure path. The result should identify the load preserved, the duration of any interruption and the actions required to restore normal operation. It should not merely report that the installed megawatts exceed demand.
Model the first phase as carefully as the final campus#
A plant selected for a mature campus may spend its early years serving much less demand. Ask how many units will operate during each phase, what load each will carry, and how the configuration preserves the required reserve.
Include a low-load operating case, a normal case, the seasonal peak and the defined outage case. Record when another turbine, transformer or bus section enters service. Your part-load performance assessment should use the selected configuration's data rather than its full-load efficiency alone.
Also distinguish reserve capacity from reserve response. A machine that can start later may help restore capacity, but it does not by itself support a load during the intervening seconds or minutes. Coordinate the generation plan with UPS, storage and load-management requirements.
Work through a phased-campus example#
Consider an invented project with delivered facility demand rising from 45 MW to 85 MW and eventually 125 MW. Assume each candidate unit can deliver 32 MW at the specified limiting site condition. The purpose is to test the arithmetic and expose decisions, not to assign a rating to an LM2500 or any other model.
| Demand phase | Units installed in this example | Total dependable capacity | Capacity after one unit is unavailable | Headroom after that outage |
|---|---|---|---|---|
| 45 MW | 3 | 96 MW | 64 MW | 19 MW |
| 85 MW | 4 | 128 MW | 96 MW | 11 MW |
| 125 MW | 5 | 160 MW | 128 MW | 3 MW |
This sequence passes a simplified single-unit-outage capacity screen at each phase. It does not say that every installed unit runs continuously, or that an offline spare can instantly replace a tripped machine. The dispatch and transient studies must determine which units run, their loading and the response available from other systems.
The table also exposes how little headroom remains at maturity. An additional 4 MW of previously omitted common plant demand would take the mature requirement to 129 MW and defeat the 128 MW surviving-capacity result. The buyer should resolve that load before committing to the configuration, rather than describing the initial three-megawatt margin as comfortable reserve.
Now change the owner's requirement to surviving one forced outage while another unit is already unavailable for maintenance. With five units installed, only three remain: 3 × 32 = 96 MW. The same plant that passed the first screen now falls short of 125 MW. This is an illustration of why shorthand such as “N+1” should be accompanied by a written event definition and a load boundary.
Build a shortlist around verified packages#
The LM2500, LM6000 and SGT-800 are useful families to investigate for modular generation strategies. Their inclusion is an invitation to compare identified configurations, not a claim that any available unit fits the example above.
GE Vernova describes modular expansion and both aeroderivative and heavy-duty options in its data-center power FAQ. The procurement task is to translate that architectural flexibility into documented output, service and installation commitments for your project.
Ask each supplier for the same information:
| Decision | Evidence to request |
|---|---|
| Can it meet the required load? | Site-specific net output and heat-rate curves |
| Will it match the electrical system? | Frequency, voltage, generator and protection details |
| What happens after a trip? | Operating philosophy, load-response data and study assumptions |
| Can it use the available fuel? | Approved composition, pressure and conditioning requirements |
| What is included? | Itemized package and balance-of-plant scope |
| Can it meet the date? | Equipment release, construction and commissioning milestones |
Use the existing LM2500, LM6000 and SGT-800 comparison to organize model-level questions.
Compare model families without assigning them invented site ratings#
Use OEM information to establish which configurations merit further study. GE Vernova's LM2500 material distinguishes the family and its modular XPRESS offering. The LM6000 product page identifies its current configurations and reference performance. Siemens Energy's SGT-800 page presents the industrial turbine's available applications and ratings.
These sources support a model shortlist. The project's engineers then need an applicable performance case for each offered package. A table that simply divides 125 MW by an OEM headline rating skips the two steps most likely to change the purchase: adjusting the rating to the site and evaluating the defined outage.
Ask for separate responses for equipment available now and equipment that would require manufacture, refurbishment or conversion. Keep the evidence of commercial availability separate from the evidence of technical suitability. A suitable model does not establish that a suitable asset is for sale, and a confirmed asset does not establish that it fits this campus.
Connect the capacity study to fuel and operating economics#
Once the unit count is credible, use the operating schedule to calculate fuel demand. As a hypothetical example, 125 MW at a plant-net heat rate of 10 MMBtu/MWh implies 1,250 MMBtu/h of fuel energy while serving that load. This is an assumed energy requirement, not a volumetric gas-flow specification. Converting it into a required flow requires the fuel's stated heating value and consistent reference conditions.
Ask the gas-supply team to assess that requirement against pressure, available flow and service terms. Keep the required maximum separate from expected annual consumption. A contract can have costs associated with reserving capacity even when actual generation is lower; use the proposed commercial terms in the model.
EPA's combustion-turbine technology characterization discusses ambient and part-load performance effects. Its technical background reinforces why one full-load reference point is insufficient for phased demand. Its historical cost figures are not current project estimates.
For the commercial comparison, ask for annual generation and fuel use by phase, planned service events and the cost of keeping additional capacity available. Avoid assigning value to every installed megawatt as though it will produce saleable electricity throughout the year. Reserve capacity serves an operational purpose even when it produces little energy.
Put fuel and site work on the same schedule#
Before reserving equipment, identify the projects that make it usable: the gas connection, any compression and treatment, foundations, exhaust system, emissions controls, electrical distribution and startup power.
Review gas turbine fuel requirements and black start and load acceptance with the responsible engineers. Record who owns each interface and which milestone controls the operating date.
An equipment-ready date and an energized-campus date should appear as separate entries in the procurement schedule. A turbine can be ready for shipment while an essential site system remains unresolved.
Make the procurement handoff specific#
The output of early sizing should be a short sourcing brief supported by the engineering basis. Include the required delivered load by phase, the exact outage criterion, the limiting ambient case, acceptable unit counts, mandatory electrical interfaces and the planned operating date.
Attach a responsibility schedule for information still missing. The performance assessment might sit with the equipment supplier, the gas connection with the utility or gas provider, and the load-step study with the electrical engineer. Assigning those tasks makes a conditional shortlist useful; leaving them unowned makes it fragile.
Before purchase, reconcile the shortlisted turbine configuration with the project's latest load schedule. Confirm that the offered serial numbers, enhancement systems and scope still match the evaluated proposal. A last-minute substitution should trigger an explicit review of output, service, interfaces and dates rather than inheriting the previous package's approval.
Frequently Asked Questions#
How many gas turbines does a 100 MW data center need?#
There is no universal count. Establish the load boundary, site-net output per unit and required outage condition, then evaluate the resulting configurations and their dynamic behavior.
Is a larger turbine always more economical?#
No. Compare total installed scope, expected operating profile, fuel use, maintenance and the consequence of losing that unit. Fewer machines can simplify some systems while increasing the capacity represented by a single outage.
Can a turbine plant eliminate the need for UPS?#
Do not assume so. Generation capacity, startup capability and ride-through perform different functions. The required power-quality and interruption performance must be demonstrated for the complete electrical architecture.
Does N+1 establish data-center Tier compliance?#
No. A unit-count calculation answers a specific capacity question. It does not establish a facility certification, the behavior of its distribution system or the continuity of the complete power architecture. State the required events and service outcomes, and have the responsible specialists assess the design against the owner’s requirements.
What information should be fixed before requesting firm equipment offers?#
Define the load boundary, required delivered MW by phase, electrical requirements, site-performance case and outage criterion. Provide the fuel information and required calendar dates. If any input remains open, identify it and request conditional proposals rather than treating an early model shortlist as a completed design.
Prepare your sourcing request#
Start with SecondWatt's gas turbine equipment pages, then use Get Quote → I'm looking to buy equipment to submit the site, required net MW, frequency, operating date, load phases and redundancy requirement. Include the load boundary so equipment discussions begin with the same definition of usable power.