Contract types for industrial load, what each one really guarantees, and why the cheapest kilowatt hour is usually the one you can lose.
A gap between two quotes for the same fleet looks like margin or negotiating room. Often it is neither. It is the difference between power you are guaranteed to receive and power you are guaranteed to lose sometimes, and the contract that spells out which one you bought usually matters more than the rate printed at the top of it.
Two baseline forms, and what each one promises
Firm power is the simplest promise a hosting agreement can make: the host delivers the contracted capacity continuously, and any failure to do so is a breach rather than a scheduled event. Interruptible power promises the opposite in one respect, that delivery can be reduced or stopped at defined points, in exchange for a rate that reflects the risk you are taking on. Between those two sits most of what gets signed: capacity that is firm most of the time but interruptible under stated conditions, or interruptible by default with a firm floor beneath it that the host will not touch.
Reading a contract as purely firm or purely interruptible is usually a mistake, because most agreements are a mix, and the mix is where the real terms live. Ask which portion of your contracted capacity carries which label, rather than accepting a single word applied to the whole agreement. The mechanism that reduces your load once a clause like this is triggered is covered in more detail in our piece on curtailment, what it costs and how to cap it, which looks at how much of that reduction can happen and whether it is billed.
Who calls the interruption, and how much notice you get
The label firm or interruptible tells you what can happen. It does not tell you who decides when it happens, and that question matters as much as the label itself. An interruption can be called by a grid operator responding to network conditions, by the underlying power supplier reacting to a wholesale price, or by the host itself making an operational or economic decision about the site. Each of those sits with a different party, on a different clock.
Ask what notice you receive before an interruption starts, and whether that period is written into the contract or left to the host's discretion. Notice is what turns an interruption from a loss into a plan: with warning, a fleet can ramp down in order and the team can schedule around the gap, and without it the first sign is a row of machines going quiet. Why the notice should differ with the party calling the event is set out in our curtailment piece.
What the discount buys, and what it costs in practice
Interruptible capacity is priced lower than firm capacity because you are absorbing risk the host would otherwise carry, and that trade can be a good one. The part to work through before signing is what an interruption costs your operation when it happens, not just what the rate saves you when it does not. A fleet that can shed load cleanly and resume without manual intervention loses only the output during the interruption itself. A fleet that needs a manual restart to bring machines back online, or one running firmware that does not recover cleanly from an abrupt cut, loses more than the interruption window, because recovery time is added on top of it.
Ask directly how an interruption is delivered at the equipment level: a signal that lets machines ramp down and resume on their own, or a hard cut with no coordination at all. That single detail can decide whether the rate discount on interruptible capacity is a fair trade or a cost shifted somewhere you were not expecting to carry it. It also helps to ask whether an interruption is confirmed with a timestamp and expected duration, or whether you are left to infer both from the moment the machines went quiet.
Matching the contract type to how you run
The right mix of firm and interruptible capacity depends on what an interrupted hour costs your business, and that is not the same question for every operator. A fleet running against a fixed obligation of its own, where an interrupted hour has to be made up somewhere, has less room to absorb interruptible capacity than one where output simply varies with whatever the site delivers. A newer fleet still proving out its hardware may prefer firm capacity while it establishes a baseline, then shift some of that capacity to interruptible once it has a clear sense of what an interruption costs it.
There is no single right split, and an operator who tells you there is one answer for every fleet has not looked closely at yours. What matters is that the mix in your contract reflects a deliberate choice rather than whatever the host happened to have available when you signed. Our piece on how industrial power pricing works belongs alongside this one, because the discount attached to interruptible capacity should trace back to a specific layer of that rate rather than sit on the page as an unexplained number.
What to ask before you sign
Go back to the contract on the table and separate the capacity into what is firm and what is interruptible, rather than trusting a single label applied to the whole figure. For the interruptible portion, find who can call an interruption, how much notice the contract requires, and whether that notice period differs depending on who is calling it. Ask how the interruption is delivered at the machine level, and whether your fleet can resume without someone stepping in manually. If any of those answers is missing from the document in front of you, ask for it in writing before you rely on the rate that comes with it.
The interruption clause described above and the curtailment clause often sit in different parts of the same agreement, but they are asking the same underlying question: how much of your contracted capacity you get over the course of a year. The curtailment piece goes through the annual cap and billing side of that question in more detail, and our hosting rate page lists these questions alongside the others that make two offers comparable.
