Why Data Center Power Plants Are Built From More, Smaller Turbines

Unit Sizing, Geared F-Class Machines and System Strength — Worked Through a Recent 624 MW Texas Order, and the Engineering a Decision Like This Actually Creates

1. Executive Summary

In July 2026, Ansaldo Energia announced a contract with Pacifico Energy for eight AE64.3A gas turbines in a sixty hertz configuration, with associated generators, supporting a data center power project in Texas. First deliveries are scheduled for 2027. For Ansaldo it is a return to the United States new-build generation market after more than thirty years.

The coverage has focused on the return, which is the right headline for a trade publication and the least interesting part of the story for an engineer. Three things in the announcement carry more information.

The first is the unit count. Eight machines at a simple-cycle rating of about seventy-eight megawatts gives roughly six hundred and twenty-four megawatts. The same capacity could have been bought as two large combined-cycle blocks. Choosing eight mid-size units instead is a deliberate trade of thermal efficiency for granularity, redundancy and delivery time — and losing one unit costs twelve and a half percent of the plant rather than half of it.

The second is that the machine is gearbox-coupled for sixty hertz service rather than direct-drive at synchronous speed. That is unusual among F-class frames in the American market and it has real consequences for the mechanical package, the auxiliaries and the maintenance model.

The third is what the order reveals about why a manufacturer absent for three decades can win work now. The established suppliers have order books extending years out. Delivery slot availability has become a competitive variable, and in some cases the dominant one. That is a market condition rather than a technology shift, and it will not last indefinitely.

This paper works through what was actually announced, the machine and its configuration, the engineering logic behind the unit count, the behind-the-meter versus grid-connected question that determines most of the regulatory workload, what six hundred megawatts of synchronous generation does for a grid that is losing it, and the engineering an order like this creates — including one risk that is routinely underpriced.

2. What Was Announced

Stated plainly, from Ansaldo’s own release and the trade coverage that followed:

• Eight AE64.3A gas turbines in a sixty hertz configuration, together with the associated generators.
• Customer is Pacifico Energy, a California-based energy infrastructure developer.
• The project supports major data center infrastructure in Texas.
• First deliveries scheduled to begin in 2027.
• Ansaldo describes the machine as suited to modular and redundant architectures, for both grid-connected and behind-the-meter applications, and as capable of supporting cogeneration.
• The company describes this as a foundation for further collaboration.

At the commonly cited simple-cycle rating of about seventy-eight megawatts per unit, eight machines represent approximately six hundred and twenty-four megawatts of nominal capacity before any steam cycle is added.



Several things were not announced and should not be assumed. The site has not been publicly identified. Whether the plant is simple cycle, combined cycle or cogeneration has not been stated. Whether it connects behind the meter or to the grid has not been confirmed. Nor has the contract value. Commentary linking the order to any particular development in the developer’s portfolio is speculation until confirmed, and this paper treats it that way.

3. Why a Thirty-Year Absence Ended

The explanation is not that the machine improved or that the manufacturer discovered the American market. It is that demand outran supply.

Ansaldo’s own reporting puts final sales in the broader sixty hertz gas turbine market above sixty-one gigawatts in 2025, described as the second-best result since 1980, driven particularly by United States demand for data center applications. The established suppliers have absorbed that demand into order books that now extend years into the future, and delivery slots have become scarce.

When lead time becomes the binding constraint rather than price or efficiency, the competitive landscape changes shape. A buyer who needs capacity operating in a defined window will consider a supplier they would not otherwise have shortlisted, provided the machine is proven somewhere. The AE64.3A is: more than seventy units installed worldwide with over four million equivalent operating hours. It is not a new design being introduced to a new market; it is an established design entering a market it has been absent from.

Two implications follow for anyone planning a project. Delivery slot availability should be evaluated as a first-order commercial variable alongside price and heat rate, not as a schedule detail discovered after selection. And the window in which a less-established supplier is competitive on that basis is a market condition, so a procurement strategy built entirely on it carries timing risk of its own.

4. The Machine

The AE64.3A is the small F-class machine in Ansaldo’s portfolio. The published characteristics relevant to a project team:

One caution on the output figure. A gas turbine rating is a reference condition, and a simple-cycle nameplate in a press release is not the capacity a plant will deliver in August in Texas. Ambient temperature reduces output substantially on a machine of this type, which is why inlet air conditioning — evaporative cooling, chilling or fogging — is a standard consideration on hot-climate sites and a real capital and water decision.

5. The Gearbox Is the Interesting Part

Ansaldo states that the AE64.3A is suited to sixty hertz applications specifically through its gearbox-based configuration and dedicated generator package. That is worth explaining, because it is a genuine architectural difference from most F-class machines in American service.

A synchronous generator on a sixty hertz system with two poles turns at 3,600 revolutions per minute. Most large frame machines sold into sixty hertz markets are designed to run at that speed and couple directly to the generator. Machines designed primarily for fifty hertz markets run at 3,000 revolutions per minute, and a fifty hertz frame scaled to sixty hertz is a different machine, not a rewired one.



The alternative is to let the turbine run at whatever speed is aerodynamically optimal for its size and use a reduction gearbox to drive the generator at synchronous speed. That is the approach here, and it has consequences worth understanding.

• It decouples turbine design from grid frequency. One machine serves both fifty and sixty hertz markets with a different gear ratio, which is why the same design can be offered into a market it has been absent from without a redesign.
• It permits a higher rotational speed for a smaller machine, which is aerodynamically favourable at this output. Small direct-drive machines at 3,600 revolutions per minute are working against their own scaling.
• It adds a component. The gearbox is a highly loaded piece of rotating machinery with its own lubrication, cooling, alignment, vibration monitoring and maintenance requirements, and it is a serviceable item in the train.
• It changes the torsional picture. A geared train has different torsional natural frequencies from a direct-coupled one, and torsional interaction with electrical phenomena — grid faults, out-of-phase synchronisation, subsynchronous conditions where series compensation is present — has to be assessed on the actual train, not by analogy.

None of this is a criticism. Geared gas turbine and generator trains have a long and successful operating record, and the arrangement is standard in the aeroderivative and industrial-frame world. It is simply a different machine to engineer around, and a project team whose experience is entirely with direct-drive frames should recognise that.

6. Why Eight Mid-Size Units Instead of Two Large Ones

This is the decision that tells you the most about what the plant is for.

Six hundred and twenty-four megawatts could be procured as two large combined-cycle blocks, three intermediate frames, or eight mid-size machines. The efficiency argument favours fewer, larger units in combined cycle. The reliability argument favours more, smaller ones. For a data center, the reliability argument wins, and the reason is what happens when a unit trips.

Four further advantages follow from granularity, and one significant cost.



• Load following. Eight units can be staged so that running machines sit near their efficient loading point rather than all running at part load. Gas turbine efficiency falls off at part load, so a fleet that stages well can outperform a smaller number of larger machines operating below their design point.
• Maintenance without capacity loss. Major inspections on one unit remove an eighth of the plant. On a two-unit plant, a hot gas path inspection removes half of it for weeks.
• Phased delivery and phased revenue. Units arriving over a period allow capacity to be commissioned progressively rather than waiting for a complete block, which matters when the load is arriving progressively too.
• Delivery slot availability. Mid-size machines from a supplier with capacity are obtainable when large frames from the established suppliers are not.
• The cost is efficiency and unit capital. Eight machines mean eight sets of auxiliaries, eight fuel skids, eight generators, eight step-up transformers, eight breakers and eight of everything else, and simple-cycle F-class heat rate is well short of a modern large combined cycle. Whether that trade is right depends on the duty, and for a plant whose job is high-availability supply to a critical load rather than lowest-cost energy, it frequently is.
  

7. Simple Cycle, Combined Cycle, or Cogeneration

The announcement covers turbines and generators. It does not state the plant configuration, and the three options have very different implications.



• Simple cycle. Lowest capital, fastest to build, best start and ramp flexibility, worst heat rate. Suits a plant whose value is availability and response rather than fuel efficiency, and suits phased delivery because each unit is independent.
• Combined cycle. Add heat recovery steam generators and a steam turbine and efficiency improves dramatically — but the plant becomes more capital intensive, slower to build, slower to start, and the steam turbine introduces a new large single unit whose loss is significant. On an eight-unit plant, the steam block sizing and how many gas turbines feed it becomes an availability question in its own right.
• Cogeneration. Ansaldo specifically notes the machine supports cogeneration, and on a data center site the obvious application is absorption chilling driven by exhaust heat — turning waste heat into cooling capacity for the load the plant exists to serve. That is a genuinely attractive coupling and it is engineering-intensive, because it ties the thermal plant and the cooling plant into one system with shared failure modes.

The choice is not purely technical. It depends on the offtake structure, on whether the plant sells into the market when the data center does not need it, on emissions permitting, and on how much site area and water are available. Announcing turbines before announcing configuration is normal — the long-lead equipment is ordered first because it has to be.

8. Behind the Meter or Grid Connected

This is the question that determines most of the regulatory and engineering workload, and it has not been answered publicly. Ansaldo notes the machine suits both.

The engineering difference is largest in the islanding case. A plant that must pick up and carry a large data center load without the grid behind it faces load rejection, block loading, frequency and voltage control on an isolated system, and protection coordination without an infinite bus — all substantially harder than the grid-parallel case. Fast-start capability, which is cited as a strength of this machine, is directly relevant to that duty.



The co-location question more broadly — large loads sited next to generation, and what obligations attach — is under active examination by regulators and system operators, and the answers are still moving. A project team should treat the regulatory position as a live design input rather than a settled background condition.

9. What 624 MW of Synchronous Generation Does for the Grid

There is a system-level point here that the market coverage does not make, and it matters more than it looks.



These are synchronous machines. Unlike inverter-based resources, a synchronous generator is physically coupled to system frequency through a rotating mass, and it provides three things that converter-based generation does not provide inherently.

• Inertia. Rotating mass resists frequency change, slowing the rate of change of frequency after a disturbance and buying time for governor response and protection. The declining inertia of modern systems is a recognised reliability concern.
• Short-circuit current. A synchronous machine contributes substantial fault current, which raises system strength at its point of connection. Converter-based resources are current-limited and contribute comparatively little, which is the root of the weak-grid control interaction problems now dominating interconnection study work.
• A voltage source behind an impedance. The machine presents a genuine voltage reference that other equipment can synchronise to, rather than tracking a measured phase.

So a six-hundred-megawatt synchronous plant is not simply capacity added next to a large load. Depending on where and how it connects, it is a system strength asset in a region that has a great deal of inverter-based generation and a documented interest in system strength. That is a genuine reliability contribution and it is one of the more compelling arguments for gas generation co-located with large loads, distinct from the energy argument.

10. The Engineering Work an Order Like This Creates

Eight gas turbines and generators is a long-lead equipment order. It is a small fraction of the engineering the project requires.

• Electrical balance of plant. Eight generator step-up transformers, generator circuit breakers or high-side breakers, a collector bus arrangement, and a switchyard. The bus configuration — how many sections, how they tie, what a bus fault removes — is a reliability decision equivalent in weight to the unit count decision.
• Excitation, governing and control. Automatic voltage regulation, power system stabilisers where required, governor droop and frequency response settings, and the plant-level control that coordinates eight units.
• Protection. Generator protection for each machine, transformer differential, bus protection, and coordination through to the interconnection. On a plant with islanding capability, an additional set of schemes for the transition.
• Studies. Load flow, short circuit, protective coordination, arc flash, motor starting, transient stability, and torsional analysis appropriate to a geared train. If the plant islands, dynamic simulation of load rejection and block loading.
• Auxiliary and station service. Eight machines have substantial auxiliary loads, and the station service design determines what happens to the plant when the grid is not there — which is the whole point on a behind-the-meter arrangement.
• Grounding, lightning protection and insulation coordination for the plant and switchyard.
• Fuel supply. Gas pressure, conditioning, metering, and the pipeline interconnection, which has its own long lead time and its own permitting.
• Interconnection engineering, application support and the study process with the transmission provider and system operator.

The ordering of long-lead equipment ahead of the balance of the design is normal and correct, because the turbines gate the schedule. It also means that a number of design decisions are now constrained by equipment already ordered, which is an argument for having the electrical and interconnection engineering underway in parallel rather than sequentially.

11. Models: The Risk Nobody Prices

This deserves its own section because it is the failure mode most specific to a manufacturer new to a market, and it is routinely discovered late.



Every interconnection process in North America requires validated dynamic models of the generating equipment, in the formats the transmission planner and system operator specify — positive-sequence models for stability studies, and increasingly electromagnetic transient models as well. Those models must represent the machine as configured, at the parameters actually deployed, and must be supplied with documentation adequate for the study engineer to use them.

For equipment with a long American service history, those models exist, are familiar to the reviewing engineers, and have been used in dozens of prior studies. For a machine returning to the market after thirty years, several questions arise that a project team should ask before the purchase order is signed rather than after.

• Do validated models exist in the required formats, at the required fidelity, for this machine in this configuration?
• Do they represent the gearbox-coupled train, the excitation system and the governor as they will actually be supplied and set?
• Does the developer have the right to use and share those models with the transmission provider, the system operator and their consultants?
• If the plant may be required to demonstrate ride-through, fast frequency response or islanding, do the models support demonstrating it?
• Who provides modelling support when the study engineer has a question, and in what time zone?

The consequence of getting this wrong is not a technical dispute, it is a schedule loss. An interconnection study cannot proceed without models, and a study that stalls waiting for a model package pushes every downstream date. Model deliverables, formats, usage rights and support obligations belong in the equipment purchase specification, with delivery dates tied to project milestones — which is the same lesson the inverter-based world learned expensively over the last several years.

12. Fuel, Emissions, and Permitting

Eight F-class machines in Texas carry a permitting workload that runs in parallel with the engineering and frequently governs the schedule.



• Air permitting. Nitrogen oxide and carbon monoxide emissions drive combustion system selection and post-combustion treatment. Dry low-emissions combustion, water or steam injection, and selective catalytic reduction are the levers, and each has capital, efficiency, water and space consequences. The annular combustion chamber and its emissions performance across load range is directly relevant here.
• Load-range emissions. A plant that starts frequently and operates at varying load has a different emissions profile from a baseload plant, and permits are written against operating scenarios. A flexibility-driven duty needs a permit that accommodates it.
• Water. Inlet air cooling, any steam cycle, and emissions control can all consume water, and water availability in parts of Texas is a real constraint that shapes plant configuration.
• Fuel supply. Pipeline capacity, delivery pressure, gas quality and its variability, and the interconnection agreement with the pipeline. Fuel flexibility is cited as a machine strength, and where an alternative fuel is intended, its storage and handling become part of the permit.
• Noise and siting. Eight turbines is a substantial acoustic source, and the surrounding land use determines how much attenuation is required.
  

13. What a New OEM Entry Means for a Project Team

None of these are objections. They are the ordinary diligence that any first-of-a-kind supplier relationship warrants, and a developer who does them properly is better positioned than one buying familiar equipment without asking.

14. The Broader Signal

Set aside the specific order and the pattern is clear enough.



Data center load growth has moved the constraint on new generation from capital and permitting to manufacturing capacity. When the binding constraint is a factory slot, the market reorganises: suppliers who were not competitive on price or efficiency become competitive on availability, buyers accept configurations they would not previously have chosen, and equipment decisions get made earlier and with less design definition behind them.

Three consequences follow that are worth naming.

• More projects will be configured around what is deliverable rather than what is optimal. That is a rational response to the constraint, and it means more plants built from more, smaller units — with the redundancy benefit and the efficiency penalty that implies.
• More equipment will be new to the American market. Which raises the model, service, parts and standards questions in Sections 11 and 13 across a growing share of projects.
• More synchronous capacity will be added near large loads. Which, if it is connected and controlled well, is a genuine benefit to systems that have been losing system strength for a decade.

The window will close. Manufacturing capacity is being added, and at some point lead times normalise and the competitive basis reverts to price, efficiency and service. Projects being configured now are being configured inside an unusual market, and the plants will operate for thirty years in a normal one.

15. Keentel Engineering Services for Thermal Generation Projects

Keentel Engineering provides the electrical power systems engineering behind thermal generation and the loads it serves — from interconnection through detailed design to commissioning and compliance.

15.1 Interconnection and System Studies

• Generation and large load interconnection engineering, application support, study-phase technical packages, and coordination with the utility, transmission provider and system operator.
• Load flow, short-circuit, protective coordination, arc-flash, motor starting and reactive capability studies across the plant and the interconnection.
• Transient stability analysis, and electromagnetic transient studies where system strength, control interaction or switching transients require them.
• Islanding and load rejection analysis for behind-the-meter and campus arrangements, including block loading, frequency and voltage control on an isolated system.
• System strength and grid contribution assessment, quantifying what synchronous generation actually delivers at a given point of connection.

15.2 Models and Compliance

• Model deliverable, format, fidelity and usage-rights requirements written into equipment purchase specifications, with delivery dates tied to project milestones.
• Positive-sequence and electromagnetic transient model development, review, cross-validation and submittal packages prepared to the transmission planner’s format.
• Model verification and validation against staged test data and disturbance records.
• Reliability standard applicability assessment and compliance programme support for registered generation, with evidence structured during design and commissioning.

15.3 Plant Electrical Design

• Generator step-up transformer and generator breaker specification, switchyard and collector bus arrangement, and medium-voltage distribution design.
• Generator, transformer, bus and interconnection protection design, settings and coordination, including schemes for grid-parallel and islanded operation.
• Excitation, voltage regulation and governor settings coordination, and plant-level control architecture across multiple units.
• Auxiliary and station service design, direct-current systems, grounding and lightning protection, and insulation coordination.
• Equipment specification with the behavioural, test and model deliverable requirements that make a specification enforceable.

15.4 Owner’s Engineer, Commissioning, and Diligence

• Technical due diligence on equipment selection, including first-of-a-kind supplier assessment, service and parts capability, standards compliance and interface definition.
• Bid normalisation so that competing proposals from different suppliers can be compared on equivalent terms.
• Design review of EPC and vendor submittals, and QA/QC of third-party studies and model packages.
• Commissioning specification and integrated test programme development, including load rejection, islanding transition and protection verification.

Keentel Engineering holds a Florida Certificate of Authorization and maintains offices in Tampa, Austin, Sacramento, and Baltimore, supporting projects across the interconnections.

16.References and Further Reading

Facts about the order in this paper are taken from the manufacturer’s public announcement and contemporaneous trade coverage. Market conditions and project details change; check for more recent information before relying on any of it.

The Announcement

• Ansaldo Energia press release, "Ansaldo Energia returns to the USA market with an order for eight AE64.3A gas turbines for a Texas data center project", July 2026, and the associated equipment pages  —  Ansaldo Energia
  https://www.ansaldoenergia.com/about-us/media-center/power-generation-news-insights
• Trade coverage of the order and of the gas turbine supply constraint, including POWER Magazine, Gas Turbine World, Gas to Power Journal and Offshore Technology  —  Various trade press
  https://www.powermag.com/ansaldo-us-gas-turbine-market-equipment-crunch/
• Ansaldo Energia 2025 consolidated reporting, source of the sixty hertz gas turbine market figure cited in this paper  —  Ansaldo Energia
  https://www.ansaldoenergia.com/

Machine and Plant Engineering

• ISO 3977 series, Gas turbines — procurement, covering rating conditions, and ASME PTC 22 for gas turbine performance testing  —  ISO and ASME
  https://www.asme.org/
• API 616 for gas turbines for petroleum and industrial service, API 613 and API 677 for special-purpose and general-purpose gear units, and API 670 for machinery protection systems  —  American Petroleum Institute
  https://www.api.org/
• IEEE Std C50.13 for cylindrical-rotor synchronous generators, IEEE Std C37.102 for generator protection, IEEE Std C37.101 for generator ground protection, and IEEE Std C37.013 for generator circuit breakers  —  IEEE Standards Association
  https://standards.ieee.org/
• IEEE Std 421 series for excitation systems, and the IEEE and CIGRE guidance on subsynchronous and torsional interaction applicable to geared turbine-generator trains  —  IEEE Standards Association and CIGRE
  https://standards.ieee.org/

Interconnection, Compliance, and Studies

• NERC Reliability Standards — including the MOD series covering modelling data and model verification, the PRC series for protection and disturbance monitoring, and the FAC series for facility ratings and interconnection  —  North American Electric Reliability Corporation
  https://www.nerc.com/pa/Stand/Pages/ReliabilityStandards.aspx
• ERCOT Planning Guide, Nodal Protocols and Nodal Operating Guide, and the ERCOT resource integration and large load interconnection materials  —  Electric Reliability Council of Texas
  https://www.ercot.com/mktrules/guides/planning
• IEEE Std 3002.3 for short-circuit studies, IEEE Std 3002.7 for motor starting studies, IEEE Std 1584 for arc-flash calculation, and IEEE Std 399 legacy guidance on industrial and commercial power system analysis  —  IEEE Standards Association
  https://standards.ieee.org/
  

17. Frequently Asked Questions

Notice and Disclaimer

This document is original technical content prepared by Keentel Engineering LLC for general professional information. It is not project-specific engineering advice, investment advice, or a recommendation regarding any equipment, supplier, developer or project, and it does not constitute a design, a study, or an evaluation of any installation.

Facts regarding the order described are taken from the manufacturer’s public announcement and contemporaneous trade press as at the date of publication. Where this paper notes that a matter has not been publicly confirmed — including plant configuration, site, interconnection arrangement and contract value — that reflects the state of public information and not any private knowledge. Analysis, inference and commentary are Keentel Engineering’s own engineering assessment and should not be attributed to any party named.

Equipment ratings, outputs, configurations and performance characteristics are as published by the manufacturer at reference conditions. Site performance differs and must be established from project-specific conditions and manufacturer data. Nothing here should be relied upon for equipment selection.

Keentel Engineering LLC is an independent engineering consultancy. Reference to any manufacturer, developer, standard, code, industry organisation, regulator, or market operator in this document does not imply affiliation with, endorsement by, or sponsorship from any such organisation, nor any commercial relationship with any party named.