The interconnection queue is the only stage of battery development where you commit years and millions before anyone tells you the price. Everything else in the lifecycle quotes first and bills second: land has an option fee, permits have application schedules, an EPC gives you a contract price. The queue inverts it. You file, you post deposits, you wait through a study cycle you do not control — and the number that decides whether your project lives arrives at the end, with the land already under option and the clock already years in.
That inversion is why, on the projects I have worked on, this is the stage that ends them. Not because the answer is usually bad — most studies do not kill the project — but because the answer is late, and binary.
This article walks the machine itself: what you file, how the study actually runs, where the money sits while you wait, and why the calendar resets. For where the queue sits in the whole lifecycle, the project development hub has the map; for the timeline across every stage, see how long development takes.
What you are actually asking for
An interconnection application is a request for two things at once: permission to connect at a specific point of interconnection, and a study of what the grid needs built before it can accept you. The second is the one that matters. A battery is a large machine that can swing from full import to full export in seconds, and the grid operator’s question is not whether your equipment is safe — that comes later — but whether the wires, transformers and breakers between you and everyone else can carry what you propose to do.
Filing is not a form; it is a posture. The application fixes your requested capacity, your technology, your point of interconnection and your in-service date, and it attaches money: study deposits and readiness security that scale with project size — typically single-digit millions of dollars across the process on a utility-scale project. Some of it returns if you proceed; the point of it is that it does not return casually. Queues drowned for a decade under applications that cost almost nothing to file, and the deposits exist to make your seriousness legible to the operator — and to everyone queued behind you.
The cluster study: judged as a group, on the grid’s calendar
In most US regions the era of one-at-a-time studies is over. Applications are batched into a cluster: a group of projects studied together against the same snapshot of the network, on a fixed cycle with fixed entry windows. Miss the window and you wait for the next cycle — the calendar is the operator’s, not yours.
Being studied as a group changes what your result is. The study answers what the network needs if everyone in the cluster connects, and allocates the cost of those needs across the projects that trigger them. Your bill therefore depends on your neighbours: who filed in your cluster, at what size, ahead of you at which buses. Two identical batteries at the same substation can draw materially different assignments in different cluster years, because the group around them differed.
This is also why the queue rewards study of the queue itself. Before filing, a developer reads the operator’s public queue like a fisheries chart: what is ahead, what is stalled, which buses are crowded, where withdrawals are likely. The interconnection entry covers the study phases in detail; what matters here is the strategic fact — your outcome is a function of the group, and the group changes.
The bill: network-upgrade allocation
The study’s output is an engineering answer with a price attached: the network upgrades your cluster triggers, and your allocated share. When upgrades bite, assignments commonly land around 10–20% of total project cost. But the honest range is wider in both directions — from roughly nothing on an uncongested bus to well past that on an unlucky one, where a single triggered upgrade such as a transformer replacement or a new line segment can rival the equipment cost.
The published averages look far worse than that band, and the reason is the most useful thing in this article. Berkeley Lab’s interconnection-cost work puts total interconnection cost for storage in MISO at about $248/kW — requests from 2000–2020, counted irrespective of status. Against a four-hour battery at $200–400/kWh, or $800–1,600 per kW of capacity, that is a fifth to a third of project cost. On its face, ruinous.
Then look at the same lab’s PJM numbers, split by what happened to the project. Over 2020–2022, storage and storage-hybrid projects that were still active or had withdrawn averaged about $337/kW. Projects with completed interconnection studies averaged about $4/kW.
That is not a rounding difference. It is two orders of magnitude, in one market, over one period — and it is the mechanism of the queue in a single pair of numbers: a large assignment usually does not get paid, it gets withdrawn from. The projects carrying $337/kW are disproportionately the ones that died of it; in MISO, withdrawn projects averaged about $388/kW of network upgrades beyond the interconnecting substation, roughly 85% of their total interconnection cost. What a built project pays is the small number, because the expensive assignments select themselves out of the population.
Two consequences you can act on. Do not budget from a published average: it describes the distribution you are drawing from, not the bill you will pay. And the distance between $337 and $4 is the measurable value of the walk-away discipline below — projects that treat a bad assignment as a verdict end up in the $4 population, and projects that treat it as an opening position end up funding everyone else’s average.
Two properties make this number dangerous. It arrives late — years after site control, when your sunk development spend is real. And it is binary — the project either carries the bill or it does not; there is no negotiating a transformer down by a third.
The discipline that survives this is deciding the walk-away number before the study answers. A developer who knows in advance what assignment kills the project can read the study result as a verdict and act on it. A developer without that threshold tends to keep paying to hold a queue position that has quietly stopped making sense — which is how “we are in the queue” becomes a way of not deciding for two more years.
The restudy: where the years actually go
The band for this stage is wide — commonly 12 to 48 months to an executed agreement, and in congested markets the queue alone commonly runs two to five-plus years. Most of the spread inside one market is a single mechanism: the restudy.
A cluster’s answer is only valid for the cluster that was studied. When projects ahead of you withdraw — and they do, in numbers — the network the study assumed no longer matches the network being requested, and the operator re-runs the analysis. Costs reallocate across the survivors. Sometimes your bill falls, because a crowded bus emptied. Sometimes it rises, because you inherit a share of an upgrade that a withdrawn project would have carried. Either way, the calendar resets while you wait to find out.
So the practitioner’s habit: watch the queue above you, not just your own application. The withdrawal of a project you have never spoken to, at a substation you share, is the event most likely to move your date and your bill. Attrition in these queues is not an edge case — historically most of what enters never reaches operation, and every exit reshuffles the survivors. Berkeley Lab’s Queued Up series, which tracks the queues nationally, found that of all the capacity requesting interconnection between 2000 and 2019, only about 13% had reached commercial operation by the end of 2024, while roughly 77% had withdrawn. Four of every five megawatts that enter a queue leave it without being built, and every one of those exits reshuffles the survivors’ costs and dates.
What the US queue reform changed
The reform that produced today’s process — FERC’s Order 2023, adopted across most US regions — replaced first-come-first-served serial studies with the cluster machinery above, and re-priced the queue’s front door: higher readiness deposits, demonstrated site control at entry, and penalties for withdrawing late. The stated aim was to drain speculative positions out of queues that had stretched toward five years in the largest markets.
For a battery developer the practical reading is: the queue now pays for readiness. A project that enters with land secured, a real point of interconnection and money it is prepared to leave on the table moves with the cluster; a project that enters to reserve an option now pays for the privilege and pays again to leave. That is a better trade for serious projects than the old world, but it moves spend earlier — more of the at-risk money is now committed before the study, not after it.
Texas runs its own version of reality: ERCOT sits outside FERC’s order and operates its own connection process, historically faster and with less network-upgrade allocation — one reason so much storage connects there. And outside the US the institutions change entirely: Great Britain and the EU run their own connection regimes on their own clocks. Great Britain’s was rebuilt in 2025 for the reason the US queues were: Ofgem approved NESO’s TMO4+ reform on 15 April 2025 and it went live that June, replacing “first come, first served” with “first ready and needed, first connected”. The reformed queue is phased explicitly across 2026–2030 and 2031–2035 — the plainest available statement that a connection date can sit a decade out.
The shape — apply, be studied, carry an allocated cost, wait on your neighbours — travels better than any of the numbers do.
Surviving the queue
Everything above compresses to four practitioner rules:
- File ready, not early. Under cluster rules, readiness is the currency; an unready filing burns deposits to hold a place it cannot use.
- Set the walk-away number before the study answers, and treat the result as a verdict against it — not as an opening position.
- Watch the queue above you. A neighbour’s withdrawal is the leading indicator of your own restudy, in both directions.
- Budget the stage as risk capital, not fees. Deposits, security and study costs are money placed at risk against an unquoted bill — size them the way you would size any other position, and know what total loss looks like before you enter.
Where these numbers come from
The practitioner ranges here — durations, deposits, the 10–20% band — are from projects I have worked on. The published figures are not mine and are cited with their scopes so you can check them: Berkeley Lab’s Queued Up series for the queue completion and withdrawal rates (2000–2019 requests, as of end-2024); its MISO interconnection-cost analysis for the $248/kW storage average (2000–2020 requests, all statuses, total interconnection cost) and the $388/kW withdrawn-project network-upgrade figure; and its PJM analysis for the $337/kW active-or-withdrawn and $4/kW completed figures (2020–2022). Those scopes are load-bearing: the same lab’s numbers say very different things depending on which population they count. The Great Britain reform is Ofgem’s approval of NESO’s TMO4+, 15 April 2025. Every one of them is a snapshot of a moving system; check the vintage before you quote it in a board paper.
The queue is not hostile. It is a machine for allocating a scarce thing — network capacity — under honest uncertainty about what each connection costs. But it is a machine that answers slowly and charges for the question. The developers who do well in it are the ones who treat the years and the deposits as the price of one number, and who know, before it arrives, exactly what they will do with it.
FAQ
How long does the interconnection queue take for a BESS project? Commonly 12 to 48 months from application to an executed interconnection agreement, and in congested markets the queue alone commonly runs two to five-plus years. The spread within one market is mostly restudies: when projects ahead of you withdraw, the cluster is re-run and the calendar resets.
Who pays for network upgrades? In most US regions the study assigns network-upgrade costs to the projects that trigger them. The assignment commonly lands at 10–20% of total project cost when upgrades bite, but the honest range runs from roughly nothing to well past that — a single triggered upgrade such as a transformer replacement or a new line segment can rival the equipment cost.
What is a cluster study? Instead of studying applications one at a time in filing order, the grid operator studies a group of them together against the same network snapshot, on a fixed cycle. Your result therefore depends on who else is in the group — and on who leaves it.
What deposits are at risk in the queue? Study deposits and readiness security, posted at application and at each phase — typically single-digit millions of dollars in total on a utility-scale project. Under the US queue reform, withdrawing late also carries penalties, so the exit itself has a price.