Data Center Power Terminology: What the Words Mean

Most of the words used to describe data center power supply are already defined precisely but these definitions are located in various places. This page describes those varied terms into a single point of reference.

Standby, prime, and continuous are equipment rating classifications which come from an international standard for generating sets, and each one sets hard limits on annual operating hours, average load, and overload capability. “Firm” is a market and tariff term with a definition that varies by system operator. Grid-parallel and islanded are electrical operating modes with protection requirements attached to them.

None of these words were invented to describe a procurement position however they are still being to describe procurement positions. A term with an engineering definition in one domain has been borrowed into data center power specification, where it carries commercial meaning it was never built to relate to. The result can be confusion with two parties agreeing on a word and disagree on what has been bought.

This matters because the label is not descriptive but it is constraining. A rating word chosen in a specification determines how many hours the equipment may run in a year, what average load it may carry, what maintenance interval applies, and whether the warranty and the air permit remain valid. Choose the wrong one and the asset is either oversized for its duty or operating outside the conditions it was sold under.

What follows is a working definition set. Where a standards definition exists, it is cited. Where one does not exist, that is stated plainly and a working definition is offered instead.

What are generator set ratings, and why does the rating matter more than the megawatt number?

A rating is a duty cycle classification, not a size. The same physical machine can be sold at several different ratings, and the megawatt figure on the nameplate changes with the rating. The rating tells you what the manufacturer will stand behind.

The International Organization for Standardization publishes ISO 8528-1, which covers generating sets driven by reciprocating internal combustion engines. It defines four ratings. Each combines a limit on annual operating hours with a limit on average load, expressed as a percentage of the rated output.

What is Emergency Standby Power (ESP)?

Emergency Standby Power is the rating for equipment that runs only when the normal supply has failed. The standard limits it to approximately 200 operating hours per year, at variable load, with an average load over any 24 hour period well below the nameplate figure. It carries no overload allowance. It is the highest nameplate number the machine can be sold at, and the most restrictive duty behind that number.

What is Limited-Time Running Power (LTP)?

Limited-Time Running Power is the rating for equipment expected to run for a defined number of hours per year at full load. The standard limits it to approximately 500 operating hours per year. It permits operation at 100 percent of the rated output for those hours, with no overload allowance. It is the rating that fits interruptible tariff arrangements, peak shaving, and load curtailment programs, where the equipment runs in parallel with the utility supply for known periods.

What is Prime Rated Power (PRP)?

Prime Rated Power is the rating for equipment operating as the main supply, for an unlimited number of hours, under variable load. The unlimited hours come with a load constraint. Average load across a 24 hour period is typically capped at around 70 percent of the prime rating. A 10 percent overload is usually permitted for short periods. This is the rating for genuinely off-grid operation, and the 70 percent average load rule is where most specification errors occur, because it means a machine rated at a given output cannot carry that output as an average.

What is Continuous Operating Power (COP)?

Continuous Operating Power is the rating for equipment running at constant load for unlimited hours. It permits sustained operation at high load with no overload allowance. It is the rating that applies to combined heat and power installations, base load generation, and any application where the machine is expected to sit at a fixed output indefinitely.

What is a Data Center Continuous rating?

A Data Center Continuous rating, sometimes written as DCC or DCP, is a manufacturer rating rather than a standards rating. It does not appear in ISO 8528-1. Individual manufacturers introduced it to describe equipment that must carry the varying load of a data center for unlimited hours, sitting between the prime and continuous definitions. Because it is a manufacturer classification, its precise limits differ between suppliers, and a specification that names it without naming the supplier has not specified anything.

What happens if a machine is operated outside its rating?

Operating outside the rating shortens engine life, invalidates the warranty, and can put the site outside the terms of its air permit. Standby-rated equipment run as a prime source is the common case. The machine will run. The consequences appear in the maintenance interval, the overhaul date, and the compliance file.

What is bridge power, and how is it different from temporary power?

Neither term has a standards definition. Both entered the vocabulary from a procurement need rather than an engineering one, and both describe intent and duration rather than capability. That is why they are used inconsistently. Working definitions:

Bridge power is on-site generation deployed to serve load during the period between site energization and the delivery of the contracted utility supply. Its defining characteristic is that it is scheduled against a known future event. It exists because the grid connection is late, not because the grid connection is absent.

Temporary power is on-site generation deployed for a defined short duration with no expectation of a permanent alternative supply arriving. Construction power and event power are the standard cases. It is usually rented rather than owned.

The commercial distinction that matters is not duration. It is whether the asset is expected to leave, and who holds the residual value if it does not.

Bridge power arrangements routinely outlast their stated bridge. Interconnection timelines move. When they move, the question of what happens to the asset becomes a question about who owns it, whether it can be redeployed, whether it can be reclassified to a different rating for a different duty, and whether it has a second-life value in another market. A rented fleet leaves and the question does not arise. An owned fleet does not leave, and the answer determines whether the original investment case still holds.

Specifying bridge power without specifying the exit is specifying the equipment and leaving the economics open.

What do grid-parallel, islanded, and behind-the-meter actually mean?

These describe electrical operating modes and physical connection points. Imprecision here has a protection and safety consequence rather than a commercial one, which is why the definitions are firmer.

Grid-parallel means the on-site generation operates synchronized with, and connected to, the utility supply. Power can flow in either direction depending on the interconnection agreement. The utility cares about fault current contribution, protection coordination, and whether export is permitted.

Islanded means the on-site generation supplies the load with no connection to the utility supply. The site is electrically separate. Frequency and voltage are controlled entirely by the on-site equipment.

Intentional islanding is a planned, permitted transition to islanded operation, covered by the interconnection agreement.

Unintentional islanding is the fault condition where generation continues to energize a section of the network after the utility supply has been lost. The Institute of Electrical and Electronics Engineers standard IEEE 1547 governs the interconnection of distributed energy resources and requires the resource to detect this condition and stop energizing the island within a defined time, commonly cited as two seconds. This is a life safety requirement, because a de-energized line that is still live is a hazard to line workers.

Open transition transfer disconnects the load from one source before connecting it to the other. There is a brief interruption.

Closed transition transfer briefly parallels the two sources so the load never sees an interruption. Closed transition requires utility agreement, because for that brief period the on-site equipment is connected to the network.

Behind-the-meter describes generation connected on the customer side of the utility revenue meter. It serves site load directly and reduces metered import.

In front of the meter describes generation connected on the network side, selling into the market rather than serving a specific site.

Behind-the-meter is a connection topology, not a permission. It does not by itself establish whether the equipment may run in parallel, whether it may export, or how many hours it may operate. Those come from the interconnection agreement, the tariff, and the air permit, and they can each impose a different limit.

What is the difference between firm capacity, firmed capacity, and dispatchable capacity?

This is the group where the same word carries the most weight and the least precision.

Firm describes power or capacity intended to be available at all times during the period of a guaranteed commitment to deliver, including under adverse conditions. That is broadly the definition used by the US Energy Information Administration. The important word is commitment. Firm describes a contractual obligation, not a physical property of an asset. The exact terms, and the conditions under which the commitment can be curtailed, are set by the tariff and the market rules, and those differ between system operators.

Firmed describes variable generation that has been contractually backed by something else so that the combination can be offered as firm. A wind or solar resource paired with storage, or with a contracted backstop, is firmed. The underlying resource has not changed. The obligation has been layered on top of it.

Dispatchable describes a capability. A dispatchable resource can be instructed to change its output, within a response time that varies enormously between technologies. Dispatchable is not the same as available, and it is not the same as entitled. A resource can be technically dispatchable and contractually unavailable.

The practical consequence for a data center is direct. A site that holds firm capacity on paper and has no generation on site has bought a contractual position. The position is only as good as the counterparty, the curtailment terms, and the conditions under which the obligation can be suspended. That may be entirely adequate. It is a different thing from having generation on site, and the two should not be compared as though they were the same product.

Why this matters

The words in a power specification are doing procurement work they were not designed for. Rating classifications were written to let a buyer compare two machines from two manufacturers on the same basis. Market terms were written to describe obligations between counterparties on a network. Neither was written to describe a data center's power strategy, and both are now being used that way.

The gap between the engineering definition and the commercial one is where cost overruns live. A machine specified at the wrong rating is either too large or operating outside its warranty. A bridge arrangement with no defined exit becomes a permanent asset with no owner for its residual value. A firm capacity position with no on-site generation is a commitment, and commitments have conditions.

The fix is not better vocabulary. It is asking, for each term in the specification, where the definition comes from and what it actually constrains.

External Links

International Organization for Standardization

Institute of Electrical and Electronics Engineers

US Energy Information Administration

A published technical explanation of the rating classifications

Five Nines and Fast Power

Making Better Decisions in AI Energy Infrastructure Investments

Questions and Answers

Q: What is the difference between standby and prime power? A: Standby and prime are duty cycle ratings, not sizes. Standby is limited to roughly 200 operating hours per year and applies only when the normal supply has failed. Prime permits unlimited hours but caps average load at around 70 percent of the rated output. The same machine can be sold at both ratings with a different nameplate figure for each.

Q: What is bridge power? A: Bridge power is on-site generation deployed to serve load between site energization and delivery of the contracted utility supply. It has no standards definition. Its defining characteristic is that it is scheduled against a known future event, which means the important question is not how long it runs but what happens to the asset if the event moves.

Q: What is the difference between bridge power and temporary power? A: Bridge power fills a gap before a permanent supply arrives. Temporary power serves a defined short duration with no permanent alternative expected. The commercial difference is whether the asset is expected to leave the site, and who holds the residual value if it does not.

Q: What does behind-the-meter mean? A: Behind-the-meter describes generation connected on the customer side of the utility revenue meter, serving site load directly. It is a connection topology and nothing more. It does not establish whether the equipment may run in parallel with the grid, whether it may export, or how many hours it may operate.

Q: What is the difference between grid-parallel and islanded operation? A: Grid-parallel means the on-site generation runs synchronized with and connected to the utility supply. Islanded means it supplies the load with no utility connection, controlling frequency and voltage on site. Unintentional islanding is a fault condition that safety standards require the equipment to detect and stop.

Q: What is firm capacity? A: Firm capacity is power or capacity intended to be available at all times under a guaranteed commitment to deliver, including under adverse conditions. It describes a contractual obligation rather than a physical property of an asset. The curtailment conditions attached to that obligation vary by system operator.

Q: What is the difference between firm and firmed capacity? A: Firm describes a delivery commitment. Firmed describes variable generation that has been contractually backed by something else so the combination can be offered as firm. The underlying resource is unchanged. The obligation has been layered on top of it.

Q: Is dispatchable the same as firm? A: No. Dispatchable describes a technical capability to change output on instruction. Firm describes a contractual commitment to deliver. A resource can be dispatchable and contractually unavailable, or contracted as firm without being physically on site.

Q: What is a Data Center Continuous rating? A: A Data Center Continuous rating is a manufacturer classification rather than a standards one. It does not appear in ISO 8528-1. Its limits differ between suppliers, so a specification naming it without naming the supplier has not specified a duty cycle.

Q: What happens if a generator is operated outside its rating? A: Operating outside the rating shortens engine life, invalidates the manufacturer warranty, and can place the site outside the terms of its air permit. The most common case is standby-rated equipment operated as a prime source.