The Anatomy of a Large-Load Rate Class

To understand a large-load tariff you have to understand the cost it is trying to recover, so let me separate the two things a customer buys from a utility. One is energy, the actual kilowatt-hours, which the utility largely passes through from the wholesale market or its own fuel. The other is capacity, the standing readiness to deliver power at the customer's maximum instantaneous demand, whenever the customer calls for it. For a household these two are bundled into a simple per-kilowatt-hour rate because a house's peak is small and predictable. For a load that can pull several hundred megawatts, the capacity cost dominates, and bundling it into an energy rate would either overcharge a steady customer or wildly undercharge a peaky one. So the tariff unbundles them.

The demand charge is the instrument for the capacity half. It bills the customer for its highest measured demand in a period, typically in dollars per kilowatt of peak per month, and it exists because the wires, transformers, and substation built for that customer must be sized to its peak whether or not the customer sits at peak the rest of the time. A data center with a high, flat load profile is, counterintuitively, a good citizen under a demand charge, because it uses the capacity it forces the utility to build. A load that spikes to full draw for a few hours and idles otherwise pays dearly under a demand charge, and rightly so, because it demanded expensive capacity it barely used. When I model a large-load bill, the demand charge and the customer's load factor, its average draw divided by its peak, usually matter more to the total than the energy price everyone quotes.

The minimum-take or take-or-pay provision is the instrument that protects the utility's investment. If a utility is going to build a dedicated substation and a new tie line for one customer, it needs assurance that the customer will actually consume enough to pay for that iron over its life. So the tariff sets a floor: the customer commits to purchase, or to pay for, a minimum quantity of demand or energy regardless of what it actually uses. This is what makes the utility willing to put shovels in the ground before the racks are installed. From the customer's side it is a real obligation, because a data center that ramps slower than planned, or that a tenant vacates, still owes the floor. The negotiation over where that floor sits, and how quickly it steps up as the load ramps, is often the most consequential number in the entire agreement.

The third structural element is the contract term. Large-load tariffs commonly require long minimum terms, frequently ten to fifteen years, precisely because the upgrades they trigger have long lives and the utility must be sure the customer will be there to pay them down. Bound up with the term are the exit provisions: what the customer owes if it leaves early, how the remaining unrecovered upgrade cost is handled, and whether any of it can be reassigned to a successor tenant. These clauses are unglamorous, but they are where a project's downside risk actually lives. A high, flat load with a long term and a fair exit clause is a financeable asset; the same load on a short term with a punitive exit clause is a trap.

Put these together and the logic of the rate class is coherent. The demand charge recovers the capacity the customer forces the utility to build. The minimum take guarantees the utility gets paid even if the customer underdelivers. The long term matches the customer's commitment to the life of the assets. And an explicit exit provision assigns the risk of early departure. A well-designed large-load tariff is simply these four ideas made specific to a number of megawatts, and reading one well means asking, at each element, whether the cost of serving this customer is being paid by this customer.