Interconnection agreement: what an executed LGIA commits

An interconnection agreement is the executed contract between an interconnecting customer and the party responsible for the network it connects to — a transmission owner, a distribution utility, or a regional system operator, depending on the market and the voltage.

The document class carries different names in different places. In much of the United States the generation-side form is a large-generator interconnection agreement, commonly written LGIA, with small-generator and load-side equivalents alongside it; elsewhere the equivalent instrument sits inside a connection agreement or a grid access contract. The names are not interchangeable and the forms are not identical. What is common across all of them is the job the document does, which is the only durable thing to write down: it converts a studied position into allocated obligations between named parties.

That is what distinguishes it from everything below it on the ladder. A will-serve letter records an expectation. A completed study produces a number. An executed agreement produces a party who must build something and a party who must pay for it. The distinction is not one of degree — it is the point at which the power position stops being diligence and starts being contract.

Two qualifications belong in the definition rather than in a footnote, because both are load-bearing.

It is not a power supply contract. An interconnection agreement governs the connection — the physical and contractual arrangements by which the site attaches to the network. What electricity costs, who supplies it and on what terms is a separate arrangement under a tariff or a supply contract. A project holding an executed interconnection agreement and no supply arrangement has a door and nothing coming through it.

It is not a delivery date. The agreement allocates who constructs the interconnection facilities and the network upgrades. Construction then has to happen, on a schedule controlled by whoever is obligated to perform it, and the gap between execution and energization is where projects most often lose time they had already committed elsewhere.

An LGIA is a generation-side agreement: it is used for a generating facility connecting so that electricity can be injected into the network. A data center is principally a load seeking to take electricity from the network. Its binding paper may instead be called a large-load interconnection agreement, electric-service agreement, facilities agreement, construction agreement, connection agreement, or a combination of documents under the utility's tariff. The label and document stack differ by market, utility, voltage and project configuration.

That distinction matters because generation and load can impose different studies, facilities, operating limits and payment obligations. It matters even more when a data-center campus includes on-site generation or storage. The campus may then have one set of documents for serving the load and another for interconnecting equipment that can operate in parallel with, or export to, the network. Holding a generation-side agreement does not by itself establish the amount, firmness or timing of electric service available to the load.

For underwriting, the correct response to unfamiliar terminology is not to force the document into an LGIA template. It is to identify the operative agreements and follow the obligations across them. A short electric-service agreement may incorporate a tariff, a facilities exhibit, a construction schedule and separate credit-support terms; the commercially important provisions may therefore sit outside the signature document.

The signed agreement rarely contains the whole commercial position. It may incorporate a tariff by reference, attach an earlier facilities study, leave the construction schedule in a later exhibit, and place deposits or credit support in separate instruments. Amendments and utility correspondence can also change the dates or scope without restating the original document. Reviewing only the signature agreement is therefore capable of producing a precise answer to the wrong version of the transaction.

The need to identify the operative stack is heightened by a moving large-load rule set. In June 2026 FERC directed all six regional grid operators under its jurisdiction to justify or reform the tariff rules governing how data centers and other large loads connect to the transmission system. (Federal Energy Regulatory Commission, as of September 18, 2026) That action does not create one national large-load agreement, and it should not be read as one. It is evidence that the governing process and document labels remain market-specific and subject to change. The reviewer has to establish which tariff, procedure and form applied when the position was created, and whether a later amendment or transition rule affects it.

A complete request follows the obligation rather than the filename. The facilities scope should reconcile to the latest studies and engineering exhibits. The current schedule should reconcile to milestone notices and utility correspondence. Cost responsibility should reconcile to invoices, deposits, security and contribution agreements. Service capacity and operating limits should reconcile to the tariff and electric-service paper. Assignment should reconcile to any consent already obtained and to change-of-control language across the stack.

The practical output is a one-page obligations schedule naming each requirement, the document and clause that creates it, the responsible party, the due date, the amount or capacity, and evidence of current compliance. That schedule is more useful to capital than a folder marked "executed interconnection agreement" because it shows whether the position is alive today and what has to be funded to keep it alive tomorrow.

Reduced to its structure, the document assigns each of the items below to somebody. The value of reading it as a finance document rather than a technical one is that every row is a cost, a timing exposure, or a credit requirement wearing engineering vocabulary.

A first-time reader looks for the megawatts. A credit reader looks for the obligations, because the rights in this document are largely fixed and the obligations are where the project can fail.

The upgrade cost, and whether it is capped. Network upgrade obligations are the item most capable of moving a project from viable to not. What matters is not the estimate but its status: whether the number is a study output or a contractual figure, whether it can be revised, what triggers a revision, and whether anything reimburses the customer for work that benefits the wider network. An uncapped, revisable upgrade obligation is an open position on the wrong side of the balance sheet, and it is often the single largest one in a layer-1 budget.

The security package running the other way. The customer typically posts security to the network party, escalating as the work proceeds. That is capital committed to a counterparty that is not financing the project, on a schedule the project does not control, and it has to be funded from somewhere before anything is earning. Where the posting entity is thinly capitalised, support has to come from a parent — which quietly re-couples layers a structure was built to separate.

The milestone schedule as a termination risk. The agreement is a set of conditions the customer must keep meeting. Positions are lost to unmet deposits and unmet milestone dates far more often than to substantive disagreements, and a delay elsewhere in the project can cascade into a default here — which is why the milestone dates belong in the same schedule as the construction programme rather than in a separate file.

Assignment and change of control. The security package over a power position is generally indirect: capital secures the entity that holds the agreement rather than the agreement itself. That works only where a change in the entity's ownership is not treated as an assignment requiring consent. It is a drafting point with an outsized commercial consequence, and it is the whole subject of who holds the interconnection position.

The operating obligations, read as availability. What the facility has committed to do when the system is constrained determines what availability it can sell. A tenant contracting for continuous operation cannot be served out of a connection the operator may reduce at its discretion — the repricing mechanism set out in curtailment.

And the description of the load or the generation. The agreement was studied and executed against a described thing. Where the project's plans change materially, the description may no longer match, and the consequences of that mismatch are specific rather than general. The questions to ask, and who has to answer them, are worked through in interconnection under a load change.

An executed agreement is the strongest power document most projects will ever hold, and it is routinely oversold on the strength of that. Four things it does not settle.

It does not make the power available. Execution allocates the work; the work then has to be done. Where the network party is building the upgrades, the schedule belongs to them, and the customer's remedies for delay are whatever the agreement provides and generally no more. A project whose energization date is an assumption about somebody else's construction programme should say so out loud in its own model.

It does not fix the cost. Estimates in an executed agreement are frequently subject to true-up against actual construction cost. The direction of travel is not usually favourable, and a budget carrying the estimate as though it were the price has understated the layer-1 capital requirement by an amount nobody has sized.

It does not make the position pledgeable. The rights under it are generally restricted on assignment and encumbered with obligations, which is why direct security over the agreement is uncommon and awkward. The workable answer is security over the holder, and it has to be put in place before it is needed rather than negotiated during enforcement.

And it does not answer whether the site is financeable. The power position is one gate of several. Site control has to outlast the timeline, entitlement has to permit the use, and the demand behind the megawatts has to be contracted rather than indicated. The full set is in what makes a site financeable, and a genuine, executed, well-negotiated interconnection agreement clears exactly one of them.

The useful summary for a buyer, a broker or a lender's analyst reading a data room: an executed interconnection agreement moves the power question from *whether* to *how much and by when*. That is a large move. It is not the end of the diligence, and the paragraph in the marketing pack describing it as "fully powered" is describing something the document does not say.