Energy discussions often revolve around categories.

Batteries versus thermal.

Short duration versus long duration.

Electrify versus don’t electrify.

But systems are not defined by categories. They are defined by constraints.

Every energy system has a binding constraint — the variable that ultimately determines what architecture is viable.

Sometimes it is fuel logistics.

Sometimes temperature.

Sometimes capital.

Increasingly, it is peak electrical power rather than total energy consumption.

Yet many energy discussions still center on annual kWh, fuel substitution, and efficiency. Those metrics matter — but only once the binding constraint is understood.

This distinction tends to surface when electrification moves from concept to implementation.

Consider an industrial facility looking to electrify part of its heat demand. On an annual basis, the energy numbers make sense. The emissions case is clear. The efficiency argument is compelling.

But the site may already be operating near its transformer capacity. Increasing that capacity would require a transformer upgrade and upstream reinforcement — a capital project in its own right, often comparable to the electrification investment itself.

At the same time, demand charges may already represent a meaningful portion of the electricity bill. Electrifying directly would increase peak kW, pushing both the physical and economic limits of the site.

In that context, the constraint is no longer annual energy consumption. It is peak kW and grid access.

When peak power becomes binding, the design space changes. Solutions that look optimal in energy terms may not address the limiting factor at all.

There is a common assumption that if the constraint is electrical, the solution must also be purely electrical. That assumption holds when energy is the dominant variable.

When power is the constraint, the relevant questions shift:

Does the solution reduce peak exposure?

Does it operate within existing connection limits?

Does it ease the transformer bottleneck?

If not, it may be improving the wrong variable.

Before selecting a technology, define the constraint.

Everything else is downstream.