You Signed for 500 Kilowatts. How Many of Them Can You Actually Use?

For much of my career in manufacturing, I bought equipment. Machines, components, and systems that had to run continuously in environments where tolerances were tight and the cost of being wrong was real.

Every one of them arrived with a rated capacity - i.e., a number on a spec sheet describing what the equipment could do.

I learned early that the rated number and the number you actually get can be two different things.

The rating is accurate. It is accurate under a particular set of conditions: ambient temperature, duty cycle, input quality, how hard you are willing to run the machine. Change the conditions, and you can get a different number.

Nobody misled anyone. The gap lives in what the spec sheet does not say.

I think about data center capacity commitments the same way.

• • •

In a data center transaction, power has become an increasingly important part of what the buyer is actually purchasing. The scarce asset is not simply floor space. It is usable power, in the right place, under the operating conditions that matter.

That makes it worth being precise about what is being purchased.

A contract tells you what you are owed. It does not tell you what you can use.

• • •

Suppose the number in the contract is 500 kilowatts. Before signing, the questions worth asking are more specific than the total:

• Does 500 describe the IT load you can actually draw, the rating of the breakers installed, capacity reserved somewhere upstream, or total facility power with cooling included?
• What happens when an entire distribution path is out of service?
• Can your equipment actually use both feeds?
• How much capacity remains if either feed has to carry the cabinet by itself?
• Where in the hall does that power land, and can enough of it be concentrated where the dense racks will go?
• Are the phases balanced?

None of these are exotic questions. They are ordinary electrical questions. They also have a habit of becoming important only after somebody tries to add another rack.

• • •

Two of those questions have the biggest financial consequence.

The first is simply what the number means.

A quoted figure can represent usable IT load, breaker rating, capacity reserved upstream, or total facility power including cooling. Those are very different numbers for the same deployment.

The distinction can matter directly in billing. In many installations, continuous load is planned at no more than 80 percent of breaker rating unless the equipment is specifically rated otherwise. If you are paying against the breaker rating rather than against what you can actually use, that difference is real.

It is cheap to establish that basis in writing and expensive to discover it later.

The second is what redundancy costs you in usable capacity.

Two 30-amp feeds do not give you 60 amps of usable redundant capacity. If the load is balanced across A and B and either feed must be able to carry the entire cabinet after a failure, the cabinet's continuous load is limited by what one feed can safely carry - about 24 amps, or roughly five kilowatts at 208 volts.

And the redundancy has to continue all the way to the equipment.

Two feeds help a server only if it is dual-corded with one cord on each side, or if a transfer switch sits somewhere in the rack. A single-corded device connected to only one rack feed cannot take advantage of the cabinet's A/B redundancy on its own.

• • •

I want to be careful here - none of this means the provider is misleading anyone.

Colocation contracts are written honestly, and the numbers in them are real. The provider is committing to something specific and, in my experience, generally delivering it.

The gap is in what goes unstated. A contract describes an obligation, but it does not necessarily describe every operating condition that determines how much of that capacity can actually be deployed.

That is not very different from the equipment I bought in manufacturing. The spec sheet was accurate and incomplete at the same time. Asking the second set of questions was my job.

• • •

Gaps like this tend to surface at the worst possible moment.

Usually it is during an expansion. The deployment is planned, budgeted, and scheduled. The hardware has arrived.

Then somebody looks at the actual distribution system and finds that the capacity is committed on paper and constrained in practice. One phase is heavily loaded. The panel with apparent capacity is not the one serving the cabinet that matters. The A and B paths are carrying very different loads.

At that point every option costs something: move the load, buy capacity you may not otherwise need, rework the deployment, or wait.

• • •

At five or eight kilowatts per cabinet, differences like these were easier to absorb. There was enough room in many designs that being approximately right often worked.

Density has taken that room away.

When a single cabinet draws 60 or 100 kilowatts, a 20 percent difference between contracted capacity and usable capacity stops being a rounding error.

On a 500-kilowatt commitment, 20 percent is 100 kilowatts. Today, that can be an entire rack.

That is why I think this belongs in the purchasing conversation rather than the troubleshooting one.

• • •

I am starting to see better conversations move in this direction.

Buyers ask for operating data earlier rather than relying only on nameplate figures. They ask what is deliverable with a path down, not just what the headline number says.

Capacity planning moves forward in the process, before equipment is on the dock.

That is the healthy version of this. The buyer gets a realistic picture. The provider gets credit for capacity it can genuinely deliver. And both sides stop negotiating against an estimate.

You cannot answer every capacity question from nameplates alone. At some point, the estimate has to give way to measured operating data.

You signed for 500 kilowatts.

It is worth knowing, before you sign, how many of them you will actually be able to use.

The Bigger Picture is a monthly series from Packet Power CEO Nate Nomeland exploring the operational and strategic challenges facing facility leaders.