Data center capacity is one of the most frequently quoted—and increasingly misunderstood—metrics in the infrastructure industry. Operators announce gigawatt-scale campuses, hyperscalers reserve hundreds of megawatts, and developers describe multi-gigawatt pipelines. Yet those figures do not necessarily represent infrastructure that can run customer workloads today.
A megawatt can describe capacity that has been announced, planned, contracted, supplied by a utility, physically constructed, energized, commissioned, equipped with IT hardware, or placed into production. Those stages are not interchangeable, and moving from one to another can take months or years.
The distinction is becoming more important as AI drives larger infrastructure commitments. Customers, investors, utilities, and technology leaders need to understand exactly what a capacity figure represents before using it to compare providers or assess how much AI compute can realistically be deployed.
Clear definitions improve major investment decisions.
Why Data Center Capacity Has Become Harder To Compare
Historically, data center capacity could often be discussed using relatively straightforward measurements such as square footage, rack count, or available IT load.
AI has pushed the industry toward megawatts and gigawatts because power has become one of the clearest constraints on deployment scale.
That shift makes sense, but it creates another problem.
Developers, operators, utilities, customers, and investors can all use the word “capacity” to describe different things.
A developer may describe the maximum potential capacity of a site. A utility may discuss the amount of power approved for connection. A colocation operator may report contracted capacity. A customer may care only about capacity available for immediate deployment.
All of those numbers can be valid while describing very different realities.
Power Secured Requires Its Own Definition
Power has become such a critical constraint that phrases such as “power secured” increasingly appear in project announcements.
But that phrase also needs qualification.
A utility agreement, reserved capacity, completed substation, energized feeder, and power physically available to the IT load represent different stages of electrical readiness.
A project may have an agreement covering future electricity supply while still waiting for transmission infrastructure, transformers, switchgear, substations, or utility construction.
For AI projects in particular, the difference can determine whether accelerators can be installed next quarter or several years later.
Contracted Capacity Measures Commercial Demand
Contracted data center capacity is an important metric because it indicates that customers have committed to infrastructure under commercial agreements.
It is therefore much stronger evidence of demand than a speculative development pipeline.
But contracted capacity is still not necessarily operational capacity.
Customers frequently reserve infrastructure before construction is finished. In some cases, securing an anchor tenant is precisely what gives a developer enough commercial certainty to finance and build the next phase.
This model is particularly relevant to large AI projects, where customers may need to reserve substantial blocks of future power well before the compute hardware arrives.
Firmus Shows Why The Distinction Matters
Firmus’ September 8 announcement illustrates the issue clearly.
The company says it has surpassed 900MW of contracted capacity after signing a multi-year agreement to provide OpenAI with dedicated AI compute capacity from two sites in Malaysia.
Firmus also says its portfolio consists of seven AI Factory sites across Australia, Singapore, Indonesia, and Malaysia.
Two are operational, while five are under development with ready-for-service dates targeted over the next 24 months.
The 900MW figure therefore demonstrates substantial contracted customer demand. It does not mean that more than 900MW of AI infrastructure is currently operating.
For customers and industry analysts, both pieces of information are necessary to understand the company’s actual position.
Energized Capacity Is Much More Meaningful
Energization represents an important transition because electrical infrastructure is connected and capable of receiving power.
For power-constrained markets, this can be one of the most important development milestones.
However, energized capacity still should not automatically be equated with usable IT capacity.
The facility may need further testing. Cooling systems may still be undergoing validation. IT halls may not be fitted out. Network connectivity may not be complete.
Energization proves that a major infrastructure barrier has been crossed, but it does not necessarily mean customers can deploy immediately.
Commissioning Separates Construction From Operation
Commissioning is one of the most important and least appreciated stages in the data center capacity lifecycle.
Mission-critical commissioning verifies that systems operate as designed, that equipment works together correctly, and that the facility can respond appropriately to normal and abnormal operating conditions.
Testing can include utility failures, generator starts, UPS transfers, cooling failures, control-system behavior, redundancy scenarios, alarms, and other events that would be difficult or dangerous to discover for the first time after production workloads arrive.
Uptime Institute emphasizes that commissioning should establish performance baselines, verify operating procedures, and determine the facility’s actual operating limits.
For customers, commissioned capacity is therefore considerably more meaningful than capacity represented only by a completed building.
Operational Capacity Is The Number Customers Ultimately Need
Operational capacity is the closest measure to infrastructure actually performing useful work.
Even here, definitions can vary.
An operator may classify a data hall as operational when it is available for customers, while an AI provider may focus on the amount of compute already installed and serving workloads.
Customers should therefore ask whether a quoted operational megawatt figure refers to available facility capacity, occupied IT load, or actual power being consumed by production equipment.
The distinction becomes particularly important when comparing colocation operators with vertically integrated AI infrastructure providers.
How CIOs Should Read Megawatt Announcements
When a provider announces a large data center capacity figure, CIOs and infrastructure teams should first identify the stage represented by that number.
- Announced Capacity: The scale the developer says it intends to build.
- Pipeline Capacity: Potential capacity across projects at different development stages.
- Power-Secured Capacity: Capacity associated with some form of power commitment, subject to the precise definition used.
- Contracted Capacity: Capacity committed to customers under commercial agreements.
- Constructed Capacity: Physical infrastructure that has substantially been built.
- Energized Capacity: Infrastructure that has crossed the electrical connection milestone.
- Commissioned Capacity: Infrastructure tested and validated for operation.
- Installed IT Capacity: Capacity with compute and associated IT hardware deployed.
- Operational Capacity: Infrastructure available for or already supporting production workloads.
- Utilized Capacity: The portion of available infrastructure actually occupied or consuming IT load.
Questions Buyers Should Ask Providers
A large headline capacity figure should be the beginning of due diligence rather than the end of it.
- How much capacity is operational today?
- How much has been commissioned?
- How much has been energized but is not yet commissioned?
- How much is currently under construction?
- What portion of the pipeline has firm utility power commitments?
- Which phases depend on future transmission or substation work?
- How much capacity is already contracted to other customers?
- What capacity is available for the buyer’s required delivery date?
- What does the provider mean by “ready for service”?
- Which construction and equipment dependencies remain?
- What happens contractually if delivery is delayed?
- Does the quoted megawatt figure refer to facility power or usable IT load?
Frequently Asked Questions
What Is Data Center Capacity?
Data center capacity generally describes the amount of infrastructure available or planned to support IT equipment, often expressed in megawatts of electrical load. The exact meaning depends on whether the capacity is planned, contracted, built, energized, commissioned, operational, or utilized.
What Is Contracted Data Center Capacity?
Contracted capacity is infrastructure that customers have committed to under commercial agreements. It demonstrates demand but does not necessarily mean the data center capacity has already been constructed or placed into operation.
What Is Energized Data Center Capacity?
Energized capacity has reached the stage where the relevant electrical infrastructure can receive power. Additional commissioning, IT installation, networking, and customer deployment work may still be required before production workloads can run.
Why Is Commissioning Important?
Commissioning verifies that electrical, cooling, controls, safety, and other mission-critical systems operate together as designed. It helps identify problems before production workloads depend on the facility.
Why Are AI Companies Contracting Capacity Before It Is Built?
Large AI deployments require substantial power, cooling, networking, and compute infrastructure with long development and procurement lead times. Reserving capacity early can secure future access to infrastructure that may otherwise be unavailable when the compute is required.
Conclusion
This distinction improves planning and procurement.
Data center capacity is becoming too important to describe with one headline megawatt number.
A 1GW development pipeline, 1GW of contracted capacity, and 1GW of commissioned infrastructure can all be legitimate figures, but they communicate fundamentally different information.
The distinction matters most to customers trying to deploy real workloads. Planned capacity demonstrates ambition. Contracted capacity demonstrates demand. Energized capacity demonstrates progress toward electrical readiness. Commissioned capacity demonstrates that the infrastructure has been tested. Operational capacity is what ultimately allows compute to run.
As AI pushes data center developments toward unprecedented scale, CIOs should increasingly evaluate providers on the path from announcement to operation rather than headline megawatts alone.
The most useful capacity question is therefore no longer simply, “How many megawatts do you have?” It is, “How many megawatts can you deliver, at this site, for production workloads, by the date we need them?”