Reading the Queue:
What Utility Interconnection Data Actually Tells You

Utility investor decks now carry a chart that barely existed five years ago: gigawatts of new demand lining up for service, presented as evidence of a growth era. The numbers are usually accurate, and they are usually read wrong, because a pipeline is not a schedule.

Section 01What a queue actually measures

Every interconnection dataset, whether it tracks power plants seeking to sell or data centers seeking to buy, is the same underlying object: a ledger of applications photographed at one instant and sorted into stages. Stage names differ by utility and by program, but the ladder underneath is remarkably constant. A request has been filed. A study is underway or complete. An agreement has been executed. Construction has started. Power flows. Five rungs, and every public number you will ever see about a queue lives on one of them.

Each rung is a different legal and financial fact. A request is an expression of interest, priced at an application fee and a deposit. A study is engineering attention, the system's first honest look at what the project does to wires, transformers, and upstream capacity. An executed agreement is a contract, with milestones and real money attached. Construction is mobilized capital. Energization is the only rung that runs a server hall or a production line. Requested, studied, and contracted megawatts differ from one another the way job applications, interviews, and signed offer letters differ, and adding them together produces a number with the same analytical value as adding those three things together.

Yet that is what a headline pipeline figure usually is: the sum of every rung. By construction, its largest component tends to be its least committed rung, because the cheapest rung to stand on is the easiest one to crowd. One more distinction completes the toolkit. A pipeline number is a stock, the level in the funnel on a single date. Deliverability is a flow, the rate at which megawatts exit the bottom. Stocks are easy to publish. Flows are what an owner actually needs. Most disclosures publish the stock.

Section 02A worked example, in the utility's own numbers

Pacific Gas and Electric's second-quarter 2026 earnings presentation, filed with the U.S. Securities and Exchange Commission, contains one of the more instructive queue disclosures any large utility has published, on a slide titled "Enabling Affordable Load Growth."¹ It reports the company's data-center pipeline in four stages, and, unusually for the genre, it publishes the stage definitions, the fee thresholds that gate each stage, and the project counts behind the megawatts. The scope line matters and is easy to miss: the table counts applications of 20 MW or more to serve new data-center load, and nothing else.

All megawatt and project figures come from the filing; the share column is simple arithmetic on them.¹ Read as a funnel rather than a headline, the disclosure says the following. Just under two-thirds of the 12.7 GW sits on the first rung. Just under 4 percent of it holds an executed interconnection construction agreement, and about 1 percent is physically under construction. Between March and June the total more than doubled, from 5,090 MW to 12,710 MW, and nearly all of the growth arrived on the first rung, which went from 1,700 MW to 8,200 MW. Construction did not move. That pattern is not a failure; it is what a demand boom looks like in queue form. Funnels fill from the top, and the top is where new interest lands first.

Two features of this disclosure deserve genuine respect, because they are exactly what a careful reader should want. First, the definitions tightened during the quarter. The first rung now requires a paid $500,000 study fee to count at all, the second requires payments scaled to estimated project cost, and the March figures were recast to the new methodology so the quarter-over-quarter comparison holds. A funnel that charges admission is pruning itself of casual interest on your behalf. Second, the project counts let a reader do small-number arithmetic. Twenty-six applications share the 8,200 MW top rung, an average above 300 MW per application. Four projects make up the entire executed-agreement stage. When the committed end of a funnel is four projects, a single campus advancing or slipping moves the disclosed figure by hundreds of megawatts, so quarter-to-quarter changes at the bottom of the funnel are events, not trends.

The same slide models affordability: each gigawatt of new load actually served is projected, under stated assumptions, to reduce other customers' bills by one percent or more.¹ That is an argument addressed to regulators and investors about who benefits from load growth, and it is worth understanding on its own terms. It is not a service commitment to anyone inside the funnel, and the filing does not claim it is.

Section 03Why headline pipelines overstate near-term deliverability

The gap between a pipeline's headline and its committed base is not an accident of one utility or one quarter. It is produced by five mechanisms that operate on every queue, everywhere, and a reader who knows them can decompose any disclosure on sight.

Queue positions are options. For the applicant, a place in line is an option on future capacity, and the option is cheap relative to the facility behind it. Rational applicants therefore hold more options than they exercise: several candidate sites screened, more than one service territory approached, one project built. Fee gates raise the price of the option and thin the crowd, which is why the second-quarter rule changes in the worked example matter, but they do not change the underlying logic. Some fraction of every top rung is exploration, and the fraction is invisible from the aggregate.

The measured base rate is sobering. The longest public accounting of queue attrition comes from the generation side. Lawrence Berkeley National Laboratory's Queued Up series, which compiles interconnection requests across the seven organized markets and fifty non-ISO utilities, reports that over 2,060 GW of generation and storage capacity was actively seeking transmission interconnection at the end of 2025. Of the capacity that requested interconnection between 2000 and 2020, 13 percent had reached commercial operation by the end of 2025, and 75 percent had been withdrawn. For projects completed in 2025, the median time from request to commercial operation exceeded five years. Most striking for present purposes: 549 GW nationally holds a draft or executed interconnection agreement and is still not operating.² An executed agreement is a serious fact, and it is still upstream of electrons.

The base rate does not transfer cleanly, in either direction. A generator in a queue is a speculative seller exposed to prices, financing, and supply chains. A large load in a queue is often a funded buyer with a business need and a board mandate. There is no public dataset yet long enough to establish an attrition rate for large-load pipelines, and an honest reader neither borrows the 13 percent figure nor assumes load requests all convert. What does transfer is structural: wherever entering a queue costs less than finishing a project, the upper rungs mix the real with the exploratory, and no aggregate number can tell you the mixture.

Wherever entering the queue costs less than finishing the project, the top rungs mix the real with the exploratory, and the mixture is invisible from the aggregate.

Definitions are local and mobile. There is no accounting standard for the word pipeline. Stage names, entry gates, and scope lines are set by each publisher and revised at will, as the worked example's own second-quarter recast demonstrates. A disclosure that counts paid applications is not comparable to one that counts letters of interest, and a series that crosses a definition change is two series wearing one label. Comparing pipeline gigawatts across utilities without reading each utility's endnotes is not analysis; it is decoration.

A stock is not a schedule. Converting a funnel photograph into a delivery forecast requires the numbers that decks rarely publish: how long the median megawatt dwells on each rung, and what fraction converts to the next rung per period. Two funnels with identical stocks can imply entirely different futures, one filling faster than it drains and one draining steadily. Without dwell times and conversion rates, the only honest statements available from a stock are about the present shape of demand, not about anyone's energization date.

Section 04What the rule changes say about the data

The strongest independent confirmation that raw request volume overstates reality is that regulators keep rewriting the rules to prune it. In July 2023 the Federal Energy Regulatory Commission issued Order No. 2023, its major reform of generator interconnection procedures. The order replaced first-come, first-served serial studies with a first-ready, first-served cluster process that studies projects in batches, and it attacked speculative volume directly with four instruments: larger study deposits, required demonstration of site control, a commercial-readiness deposit at each study stage, and withdrawal penalties for projects that exit late.³ Every one of those instruments exists because the regulator concluded that unpriced queue positions had filled the nation's queues with capacity that would never be built.

The same movement has reached the load side, on a younger clock. In July 2025 the California Public Utilities Commission approved interim implementation of an updated Electric Rule 30 framework for large transmission-level customers in PG&E's territory, the class that includes data centers, under which applicants agree to fund necessary transmission work up front, with refund treatment reserved for a future decision.⁴ The worked example's own fee gates, added in the second quarter of 2026, are the disclosure-level echo of the same idea, and the presentation notes that projects from the utility's 2026 cluster study are advancing into the large-load pipeline: batch-study mechanics, born on the generation side, arriving on the demand side.¹

For a reader of queue data, the reform era carries two practical lessons. Recent-vintage queue capacity is more meaningful per megawatt than older-vintage capacity, because each megawatt now has more money and more demonstrated readiness behind it. And any time series that spans a reform breaks at the reform, so the first task with any queue chart is to date its definitions. The general principle underneath both lessons is the most useful sentence in this paper: weight megawatts by the money behind them.

Section 05What the disclosure is for

None of this is a criticism of the utility's numbers. An investor presentation answers investor questions: is the demand real, is the capital program justified, what does load growth do to rates. On those questions the worked example is genuinely informative, and its hygiene, publishing the full funnel, defining the stages, stating the fees, flagging the recast, disclosing project counts, is better than the genre's standard. The misreading is supplied by readers who take the top of someone else's funnel as the delivery schedule for their own project.

Read for what it is, a 12.7 GW pipeline tells an owner something important: this territory is where enormous load wants to land. Utility engineering attention, upgrade programs, and contractor capacity will be allocated in that context, and whatever your project needs from the shared system will be requested alongside everyone else's. Competition for the machinery of delivery is real information, and it argues for starting earlier, not later. What the pipeline cannot tell you, at any level of detail, is when your site gets its megawatts. For that there is exactly one instrument: a written, dated, site-specific answer from the utility. Obtaining one is a discipline of its own, which is the subject of the next section.

Section 06What to ask your utility for, in writing

Nothing on this list is adversarial, and a capable utility account team can produce most of it in the ordinary course. These are the load-side equivalents of the readiness showings that generation developers are now required to make. The pattern across all eight is the same: convert conversation into dated documents, because only dated documents can anchor a capital plan.

1. The governing framework, by name.Which tariff rule, program, or study track applies to your request, and its current regulatory status. Interim frameworks and open proceedings move; process risk is schedule risk, and it starts here.
2. Your position on the utility's own stage ladder.The stage your application occupies, the stage's definition, and the specific gate, payment, or signature that exits it. If the utility publishes a funnel to investors, your project is on one of its rungs.
3. A dated capacity statement for your delivery point.What can be served at your substation and feeder today, and what portion of your request depends on upgrades not yet built. Territory-level headroom is not site-level headroom.
4. The upgrade scope your service depends on, and who pays.Named upgrade projects, estimated in-service dates, cost responsibility, and refund mechanics. An energization date is only as real as the upgrades beneath it.
5. The full fee and milestone ladder to an executed agreement.Every payment, its trigger, its refundability, and the consequences of withdrawal. The money schedule is the true process map, and it is knowable in advance.
6. An energization estimate with its dependencies stated.A dated estimate that names the conditions that would move it is worth more than a confident verbal range. An answer that cannot yet be dated should enter your plan as a risk, not a milestone.
7. Aggregate queue context at your constraint.How much requested capacity sits ahead of yours at the same substation or upgrade, in aggregate; other customers' identities are neither needed nor appropriate. A short line and a long line justify different plans.
8. The flexible and phased service menu, with terms.Curtailment windows, firmness, pricing, and the path to firm service later. A faster date with conditions attached can be the right answer, but only once the conditions are read as carefully as the date.

A refusal, or an honest "not yet knowable," is itself high-grade information: it means the date is not knowable, and the capital plan should say so out loud. That is the moment alternatives earn their study, each with its own honest ledger. Phased energization takes partial capacity now and leaves the balance exposed to the queue. A flexible connection buys an earlier date and gives up firmness. On-site generation, whichever technology class fits the site, buys schedule control and takes on capital, fuel exposure, permitting, and an operating burden that belongs in the comparison at full cost. Storage shifts energy across hours and does not create it. Relocation moves the problem to where power exists and pays for the move. Waiting spends no capital and consumes the scarcest input, time. Pricing that menu against a written grid answer, rather than against a headline, is what a defensible power decision looks like.

Section 07The reading discipline, condensed

Five rules cover nearly every queue chart you will encounter.

1. Find the rung. Before reacting to any number, identify which rung of the ladder it measures and what that rung legally and financially requires of the applicant.
2. Read the endnotes before the chart. Scope lines, entry gates, thresholds, and recasts live there. The footnote is where the publisher tells you what the number is; the chart only tells you how big it is.
3. Compare rung to rung and vintage to vintage. Requested capacity against requested capacity, contracted against contracted, and never across a definition change without a break in the series.
4. Weight megawatts by the money behind them. Fee-gated, deposit-backed, and contracted capacity deserves more belief per megawatt than free-entry capacity, on any queue, in any jurisdiction.
5. Use pipelines for context, never for your date. The market's queue tells you about competition for the shared system. Only a written, dated, site-specific answer tells you about your project.

The distance between 12.7 GW and 490 MW is not a scandal, and treating it as one would be its own misreading. It is a market in the act of forming, disclosed with above-average candor by the utility that operates it. The queue is telling the truth. It is simply not answering your question. Your question is answered in writing, at your delivery point, with a date, and until it is, it has not been answered at all.