Does an interconnection survive a change of load?

A common and expensive assumption is that an interconnection is a quantity — a number of megawatts, attached to a point on the network, available to whoever owns the site. It is not. It is a contractual position held by a named customer, studied against a described load, and the description is part of the agreement rather than context around it.

What was described generally includes more than the peak. It includes the shape of the demand, the voltage and point of delivery, the equipment behind the meter, the protection scheme, and — the item that matters most here — what happens when the system is short. A load that the operator can reduce or drop under defined conditions was studied as such. The studies, the upgrade scope, the cost allocation and frequently the rate treatment all followed from that.

So the question a conversion raises is not whether the megawatts are still there. It is whether the load the buyer intends to run is the load the position describes. Those are different questions and only the second one is answered in the interconnection file.

This is a step beyond the verification exercise in [how to verify a power claim](verifying-a-power-claim). There the issue is whether the document is real and what it is conditioned on. Here the document is real, executed and energized — and it may still describe somebody else's load.

The distinction sits at the centre of this and it is worth stating in the utility's terms rather than the buyer's.

An interruptible or curtailable load is one the system can reduce when it needs to. From the operator's side it is close to a resource: it can be counted on to go away at the moment the system is most constrained, which means it can frequently be served without building for it as though it were always present. That treatment is why such loads are typically served on terms not available to ordinary customers, and it is a large part of why sites designed to be interrupted could be connected at scale and at speed.

A firm load is the opposite commitment. It expects to be served continuously, it must be planned for at its peak, and it contributes to the capacity the system has to hold. A training or inference tenant contracts for availability and passes that requirement straight through to the facility, which passes it to the utility. There is no version of a high-availability data-center tenancy that can be satisfied by a supply arrangement the operator may curtail at its discretion.

The conversion therefore does something specific: it removes the flexibility that was part of the consideration for the original position. Whatever the site received in exchange for being interruptible — speed of connection, a scope of upgrades sized for a droppable load, a rate structure, an avoided cost allocation — was received against a commitment the new use cannot honour.

That is the mechanism. Everything below is what follows from it.

The honest position is that the consequences are operator-specific and jurisdiction-specific, and that they are being actively revisited in most markets as large loads become a planning question rather than a marginal one. A page that told a buyer what will happen would be wrong somewhere and would be wrong soon.

What can be stated durably is the list of things that may move, and it is short enough to work through with a utility in a single meeting.

The recurring exposure is material modification. Interconnection processes generally distinguish between changes that can be absorbed into an existing position and changes significant enough to require re-examination — and a change from a droppable load to a firm one is squarely the kind of change that gets tested against that threshold. Where it crosses, the consequences run from a re-study of the same position through to a fresh application, and the difference between those two outcomes is measured in years.

The last row of that table deserves its own treatment, because it is the one most often missing from a data room.

Sites built around interruptible load frequently sit inside a web of arrangements that reward flexibility: demand-response enrolments, curtailment programs, load-management provisions inside the supply contract, and sometimes bilateral arrangements with the utility or a retailer. These are commitments, not options. They generally carry performance obligations, measurement, and consequences for failing to deliver when called.

A firm AI load cannot deliver them. Which raises four questions a buyer should ask before, not after, agreeing a price:

• What is the site actually enrolled in, and on what terms? Enrolment paperwork is frequently held by an operations team rather than in the corporate file, and it does not always appear in a disclosure schedule unless asked for by name.
• What does exit cost? Programs commonly have term commitments, notice requirements, and clawback of benefits already received. The exit cost is a real number in the acquisition, and it is one nobody has usually priced.
• Does the current commercial arrangement depend on them? Where the rate treatment or the supply contract is conditioned on continued participation, leaving the program changes the economics of the site rather than merely ending a side arrangement.
• Can the site remain partially flexible? This is the underexplored answer. A phased conversion, a portion of load retained on interruptible terms, or on-site generation covering the curtailment obligation can in some cases preserve part of the arrangement. Whether that is available is a conversation with the utility, and it is worth having before the assumption is made in either direction.

Reduced to its core: the buyer is paying for a position in a queue that is now years long in most serious markets, and the conversion is an event that could put that position at risk. Everything else in the transaction is replaceable at a price. The position is not.

The events that put it at risk are consistent enough to list, and they compound because a conversion frequently involves several at once:

• The change in the character of the load, as above.
• A change in requested capacity, in either direction. Reducing the request is not automatically safe; in some processes it is as much a modification as increasing it.
• A change to the studied parcel or the point of delivery, which can follow from a redevelopment plan that moves the building.
• A change of the customer of record, which is a transfer to the utility whatever the parties call it — the entity question worked through in [who holds the interconnection position](/structures/who-holds-the-interconnection-position).
• Missed milestones and unpaid obligations during a conversion period when the site is producing nothing. Positions are lost to unmet deposit and milestone requirements more often than to substantive disputes.

So the practical sequence, and the reason this page exists, is that all of this is established with the utility before the site is priced, not during closing. The buyer's questions are: does the existing position support a firm load; if not, what is required and how long does it take; is the position preserved through that process or does it reset; what changes in the service arrangement and the rate; what does exiting the flexibility commitments cost; and what credit support will be required, from which entity.

Six answers, from one counterparty, in writing. A site priced on its existing interconnection without them is priced on an assumption that the utility has not agreed to — and the utility is the only party whose view of this matters.