Network Upgrades and Withdrawal Penalties: Cost Allocation, the Reimbursement Nobody Discounts, and a 25% Cliff

Two things about network upgrades are almost universally modelled wrongly.

The first is that the cost is treated as a cost. It usually is not: an interconnection customer that funds network upgrades is generally entitled to repayment, in the form of credits against future transmission service. The money comes back. What the project actually bears is the time value of a very large interest-free loan to the transmission system — which on a $60m upgrade is roughly half of face, not all of it.

The second is the withdrawal decision. Under FERC Order 2023 the penalty for leaving is waived where the assigned network upgrade cost has risen by 25% or more in the most recent cluster study report, or by 100% or more in the individual facilities study report. That is a cliff. At a 24% increase you pay to leave; at 25% you walk away free.

This post covers who pays for upgrades, how they are allocated inside a cluster, what the reimbursement is worth, and how to model a withdrawal decision that turns on a threshold set by someone else's study.

ℹ️ Note: Deposit and penalty structures are set in each region's tariff and differ materially from the Order 2023 pro forma. Figures below are labelled by source and should be checked against the applicable tariff.

Who Pays for Network Upgrades?

The interconnecting generator, on a "but for" basis — and then gets most of it back. ISO-New England, NYISO and PJM allocate upgrade costs to the generator whose interconnection caused them, and the customer is generally entitled to repayment through credits against future transmission service.

That two-part structure — pay now, recover later — is the whole economics of the line item, and collapsing it into a single capital cost is the error worth fixing first.

It also explains why network upgrades interact with the capital structure rather than just the budget. The upfront payment has to be funded — usually from the construction facility, occasionally from sponsor equity — while the credits arrive over many years against transmission charges. That is a classic asset-liability mismatch: a large, early, certain outflow against a long, slow, contingent inflow. Lenders treat the two halves very differently, and it is common for a facility to fund the payment while giving little or no credit to the receivable, which is a defensible position given how much has to go right for the credits to be realised in full.

The refinements that matter are the term and the usability. A twenty-year reimbursement discounted at a project's cost of capital is worth roughly half its face; a five-year reimbursement is worth most of it. And credits are worth nothing at all to a party with no transmission service charges to offset them against — which is a real question in some structures and is worth asking before the schedule is built.

How Are Costs Allocated Inside a Cluster?

By a mix of load-ratio share and power-flow modelling, and the mix depends on what is being built.

PJM's approach under its Regional Transmission Expansion Plan is representative. For regional facilities and the necessary lower-voltage facilities that go with them, cost is split evenly: 50% across the system on a load-ratio share basis and 50% assigned to specific customers by solution-based distribution factor — the DFAX method, which models how much of each party's power flow actually uses the facility. Lower-voltage projects use DFAX alone. Smaller reliability upgrades and local enhancements are allocated 100% to the transmission zone where the facility sits.

The practical consequence for a developer is that the number arriving from a cluster study is not a bill for a piece of equipment. It is the output of a power-flow allocation across every party in the study, and it moves when any of those parties moves. A project that withdraws reallocates its share to the rest; a project that is added dilutes everyone's. That is why cluster studies restudy, and why the withdrawal penalty regime exists at all.

The Self-Funding Problem

A structural issue worth knowing, because it changes who earns what on the same upgrade.

In several regions, transmission owners have had the right to unilaterally elect to fund network upgrades themselves and earn a regulated return on that funding. Where they do, the interconnection customer does not advance the capital — it pays the return, through charges, over the life of the asset.

FERC issued an order on 13 June 2024 requiring PJM, MISO, SPP and ISO New England to show cause as to why tariff provisions granting that unilateral right are not unjust, unreasonable and unduly discriminatory or preferential.

The reason it matters commercially is that the two arrangements are not economically equivalent, and the customer does not choose between them. Self-funding removes a large upfront capital requirement — which for a developer with constrained working capital can be genuinely valuable — and replaces it with a permanent obligation to pay somebody else's cost of capital on an asset the customer would otherwise have owned the credits to. Which is better depends on the developer's own cost of capital against the transmission owner's allowed return, and a developer with cheap capital is straightforwardly worse off.

A model that carries "network upgrades" as a capital cost without establishing which funding arrangement applies is modelling a different transaction from the one in the agreement.

What Are the Deposits at Risk?

Region-specific, and larger than the Order 2023 pro forma suggests. Indicative structures:

Region	Study deposits	Additional security
PJM	$75k (≤20 MW) · $150k (20–100) · $250k (100–300) · $400k (300+)	$4,000/MW readiness deposit at Decision Point 1
CAISO	$150k–$500k+	IFS ladder: 10% of assigned upgrades after Phase I, more after Phase II, balance at IA execution
MISO	$60k–$300k	Milestone deposits keyed to capacity and estimated upgrade cost
ISO-NE	$50k–$250k	Per-MW readiness deposits; separate FCM penalties where a capacity obligation exists
NYISO	$50k–$200k	Facilities study deposits tied to estimated upgrade cost
SPP	$50k–$200k	Attachment V-R posting based on allocated upgrade costs

On a 200 MW PJM project that is a $250,000 study deposit plus an $800,000 readiness deposit at Decision Point 1 — over a million dollars committed before the facilities study begins, and total withdrawal exposure in a congested PJM zone that can exceed $2m in deposits alone, before any network upgrade cost allocation.

CAISO's structure is the one to study for shape. The Interconnection Financial Security ladder steps from 10% of assigned network upgrade costs after Phase I, through a further posting after Phase II, to the full allocation at interconnection agreement execution. It is the same option structure the previous post described, with the exercise price revealed in instalments.

How Do Withdrawal Penalties Work?

Tiered by phase, floored at the study deposit, capped at what was collected, and waived where the withdrawal does not hurt anyone.

The Order 2023 framework sets penalties by the phase of the process reached, equal to the greater of the customer's study deposit or a tabulated amount. Three limits apply:

• Penalties cannot exceed the dollar amounts collected from interconnection customers.
• Penalties are not assessed where the withdrawal does not have a material impact on any interconnection request in the same cluster.
• The funds are used to fund cluster studies, and once those are complete, to offset net increases to network upgrade cost assignments experienced by the remaining customers in the same cluster.

That last provision is elegant. The harm a withdrawal causes is precisely that it reallocates upgrade cost onto the projects that remain, and the penalty is applied to exactly that harm. It is one of the better-designed pieces of the Order.

The 25% Cliff

And here is the exemption that decides real money.

FERC defines "significant, unanticipated increases" — which release a customer from the penalty — as:

• an increase of 25% or more of the assigned network upgrade costs in the most recent cluster study report; or
• an increase of 100% or more of the assigned network upgrade costs in the individual facilities study report.

The logic is sound. A developer should be able to leave when the deal it entered has materially changed, and should not be able to leave for free when it simply changed its mind.

The consequence is a discontinuity. Consider a project assigned $60m of upgrades at cluster study. A restudy at $73m is a 21.7% increase — the project is worse, potentially badly, and leaving costs the full penalty. A restudy at $76m is a 26.7% increase — worse by a further $3m, and leaving is free.

Between those two numbers the developer's entire legal position changes, and nothing it does affects which side of the line it lands on. The threshold is computed from someone else's study of someone else's cluster.

There is a practical instruction that follows, and it is one of the few genuinely actionable things in this post: compute the exemption threshold in dollars the moment the cluster study lands, and put it on the front page of the project's risk register. A developer that knows the number is $75.0m knows exactly what to watch for in the restudy. One that does not will be negotiating the withdrawal decision in the same week it first learns the threshold exists.

The asymmetry underneath the cliff is worth naming as well, because it explains why the threshold is set where it is. A developer's incentive to leave rises smoothly with the upgrade cost — every dollar makes the project worse. The right to leave without penalty arrives in a single step. Between the point where withdrawal becomes economically rational and the point where it becomes free, there is a band in which the developer wants out, cannot get out cheaply, and is exposed to further increases while it decides.

That band is where most withdrawal disputes live, and it is why the penalty is best understood not as a fine but as the price of an option the developer implicitly wrote when it entered the cluster. It sold the right to leave, in exchange for being studied. The 25% and 100% thresholds are the strike prices at which that option is handed back.

What Happens When Someone Else Withdraws?

The flip side, and it is the reason a cluster study result is never final.

When a project leaves a cluster, the network upgrades it was driving do not necessarily disappear — some of them were serving several projects, and the DFAX allocation simply redistributes across whoever remains. A project that took a cluster study result at $60m can find itself at $75m after a restudy without any change to its own configuration, its own site, or its own request.

That produces a cascade risk the serial queue was notorious for and clustering reduces without eliminating. One withdrawal raises costs for the remainder; some of the remainder then cross their own 25% exemption threshold; those that do can leave for free; and their departure raises costs again for whoever is left. The withdrawal penalty pool exists precisely to damp this — the funds are applied to offset net increases in upgrade cost assignments for the same cluster — but it is a partial offset, capped at what was actually collected.

Two consequences for a developer.

Your exemption threshold can be crossed by someone else's decision. That is uncomfortable to describe and important to model. It is the clearest case in this series of a project-level number that is determined entirely at the portfolio level of a process the project does not participate in.

Being the last to leave is the worst position. The exit is sequential, the penalty waiver depends on how much your costs have risen, and costs rise as others depart. A project that is marginal at $60m and watches two peers withdraw is not in a stable position; it is in a queue for an exit that gets cheaper the longer it waits and a project that gets worse at the same rate.

Can You Get Cost Certainty?

Not from the study, and only partially from anywhere else.

A cluster study estimate is an estimate produced by a power-flow model against a defined set of assumptions, at a point in time, before detailed engineering. It is not a quote. The facilities study refines it; construction refines it again; and the 100% threshold in the facilities-study exemption exists precisely because FERC anticipated that refinement could double the number.

Three partial protections exist in practice, and it is worth being clear about what each does.

The exemption thresholds are themselves the main protection. They do not cap cost — they cap commitment, by allowing exit when the number moves enough. That is a meaningfully different thing, and a developer treating the 25% threshold as a cost cap has misread it.

Contingency in the project budget is the obvious answer and the least satisfying, because the distribution is not symmetric. Upgrade estimates revise upward far more often than downward, so a symmetric contingency is systematically short. Size it against the observed revision distribution for the region, not as a flat percentage.

Phased or negotiated scope is available in some circumstances — agreeing what is actually needed for the project's own service, versus what the study assigns on a system-benefit basis. This is where the distinction between a DFAX-allocated regional facility and a local reliability upgrade earns its keep, because the arguments available are different for each.

What does not exist is a fixed-price network upgrade. Any model presenting one number with no range has represented an estimate as a contract.

How Do You Model This in Excel?

As a financing cost with a reimbursement schedule, and a withdrawal decision with a cliff.

The reimbursement, which changes the number by half

Assumptions, labelled as such:

Assigned network upgrades                      $60,000,000
Reimbursement via transmission credits    straight-line, 20 yrs
Discount rate                                            8%
Annuity factor, 20 yrs @ 8%                          9.8181

Annual credit          60,000,000 ÷ 20            =  $3,000,000
PV of reimbursement    3,000,000 × 9.8181         = $29,454,442
Net PV cost            60,000,000 − 29,454,442    = $30,545,558

Effective cost as % of face                       =       50.9%

A $60m upgrade obligation costs about $30.5m in present value. A model charging the full $60m against the project has roughly doubled the burden — and one charging nothing, on the basis that it all comes back, has understated it by thirty million.

The withdrawal decision at the cliff

Base case, upgrades $60m         net PV cost      = $30,545,700

Restudy at $73m  (+21.7%)        net PV cost      = $37,163,762
                                 incremental cost =  $6,618,204
                                 exemption?              NO
                                 cost to withdraw =  $1,050,000

Restudy at $76m  (+26.7%)        net PV cost      = $38,691,040
                                 incremental cost =  $8,145,482
                                 exemption?             YES
                                 cost to withdraw =          $0

In both cases the project is worse off by enough that withdrawal is rational. The cliff determines only whether that costs a million dollars or nothing.

The threshold to publish

Exemption threshold = assigned upgrades × 1.25
                    = $60,000,000 × 1.25          = $75,000,000

One number. It should appear in every board paper on the project from the day the cluster study lands.

The decision rule

Withdraw if:  incremental net PV cost  >  withdrawal penalty
Stay if:      project NPV after upgrades still clears the hurdle

Breakeven upgrade cost where project NPV = 0     (solve)

ℹ️ Note: The reimbursement calculation above assumes credits are actually usable — that the project has transmission service charges to offset, over the period assumed, and that the transmission owner has not self-funded. Where the transmission owner self-funded, there are no credits to discount and the entire calculation above does not apply. Establish the funding arrangement before building the schedule.

To model the reimbursement schedule, the effective cost of upgrades and the withdrawal cliff, prompt Dezzmond with your assigned upgrade cost and study stage.

What Do Developers Actually Check?

• Is the upgrade cost modelled net of reimbursement, and is the reimbursement period realistic?
• Has the transmission owner self-funded? If so there are no credits and the economics change entirely.
• What is the 25% exemption threshold in dollars, and is it on the risk register?
• What deposits are at risk at each decision point, in the applicable regional tariff rather than the pro forma?
• Does the withdrawal have a material impact on the cluster? If not, no penalty is assessed.
• What is the breakeven upgrade cost at which the project stops clearing its hurdle rate?
• How much of the assignment is DFAX-driven and therefore liable to move when other parties move?

Frequently Asked Questions

Do interconnection customers get network upgrade costs back?

Generally yes, through credits against future transmission service. The project bears the time value rather than the full amount — on a $60m upgrade repaid over twenty years at an 8% discount rate, the effective cost is about $30.5m, or 51% of face.

When can a developer withdraw without a penalty?

Where assigned network upgrade costs have increased by 25% or more in the most recent cluster study report, or by 100% or more in the individual facilities study report — or where the withdrawal has no material impact on any other request in the cluster.

Where do withdrawal penalties go?

To fund the cluster's studies, and once those are complete, to offset net increases in network upgrade cost assignments for the remaining customers in the same cluster. They are not revenue for the transmission provider.

What is transmission owner self-funding?

An arrangement under which the transmission owner elects to fund network upgrades itself and earns a regulated return, rather than the interconnection customer advancing the capital and receiving credits. FERC ordered PJM, MISO, SPP and ISO New England to show cause on these provisions in June 2024.

Can another project's withdrawal raise my costs?

Yes. Upgrade costs are allocated across the cluster by power flow, so a departure redistributes the remaining share. A project can cross its own 25% exemption threshold — and gain a free exit — purely because a peer left.

Is a cluster study estimate a fixed price?

No. It is a power-flow allocation against point-in-time assumptions, before detailed engineering. The facilities study refines it, and the fact that FERC set the facilities-study exemption at a 100% increase indicates how far it can move.

How are upgrade costs allocated within a cluster?

By a combination of load-ratio share and power-flow modelling. In PJM's regional plan, regional facilities are split 50% on load-ratio share and 50% by the solution-based distribution factor method, with local reliability upgrades assigned entirely to the host zone.

Closing: A Loan, a Lottery and a Line in the Sand

Network upgrades occupy a strange place in a development model. They are the largest single unknown in most projects, they arrive from a process the developer does not control, and they are almost always represented as a single number that is wrong in both directions at once — too large because the reimbursement is ignored, and too certain because the allocation moves with every other party in the cluster.

Getting the first part right is arithmetic, and it halves the number. Getting the second right is harder, because the allocation genuinely is a function of who else stays and who else leaves, and no amount of project-level diligence resolves it.

What a developer can control is knowing where the line is. The exemption threshold is computable the day the cluster study lands, it is a single number, and it determines whether a bad restudy is an expensive problem or a free exit. It is also, in the experience of most people who have been through a restudy, the thing nobody calculated until the week they needed it.

That closes the interconnection pair. The final post in Series C returns to operations, and to the exposure that sits underneath every revenue number in this series: curtailment — economic, congestion and manual — and who bears each.

Sources: FERC — Explainer on the Interconnection Final Rule (Order No. 2023-A) · Akin — FERC Seeks to Eliminate Utility Self-Funding of Network Upgrades · Berkeley Lab — Transmission Cost Allocation Practices · Foley Hoag — A Complete Guide to Order No. 2023-A