Hub-Settled PPAs and the Gap to Nodal Generation: Why a P50 Survives the Model and Not the Settlement Statement
Norton Rose Fulbright records an hour in which the ERCOT North hub spiked to $9,000 per megawatt-hour while node prices were only $1,000. A 300 MW project settling against that hub took a $2.4m loss in a single hour.
Nothing failed. The turbines turned, the meter recorded, the generation hit forecast. The project simply sold at one price and settled against another, and for sixty minutes the gap was eight thousand dollars a megawatt-hour.
This is the gap the previous post described, seen from inside a contract. And it is why an energy yield report can be exactly right and a revenue model built on it exactly wrong: a P50 is a statement about megawatt-hours, and the things that destroy revenue are not measured in megawatt-hours.
This post takes the hub-settled PPA apart — the settlement arithmetic, who bears what, and the three-part bridge from a P50 generation number to the cash that actually arrives.
ℹ️ Note: This describes how these contracts work in practice. It is not legal or investment advice, and every modelling figure below is a labelled assumption.
What Does a Hub-Settled PPA Actually Do?
It fixes the price the generator receives at the hub, and leaves the generator selling at its node. The contract-for-differences settlement references a hub index; the physical energy settles at the project's own location. The difference between the two is the generator's, every hour, in both directions.
A virtual PPA is, in Norton Rose's phrase, "a purely financial contract that exchanges a fixed-price cash flow for a variable cash flow." The project sells into the spot market and receives a settlement from the offtaker for the difference between the strike and a floating index. Everything turns on which index.
The Settlement Arithmetic
Worth writing out once, because the result is not intuitive and the sign errors are expensive.
The generator has two cash flows: what the market pays it, and what the contract settles.
Market receipt = Node price
CfD settlement = Strike − Index (positive = buyer pays generator)
Total to generator = Node + Strike − Index
Node-settled (Index = Node):
Total = Node + Strike − Node = Strike
The generator receives the strike, every hour, whatever happens locally. It is fully hedged.
Hub-settled (Index = Hub):
Total = Node + Strike − Hub = Strike + (Node − Hub) = Strike + Basis
The generator receives the strike plus basis — and since basis is usually negative for renewables, that means the strike minus something.
That single line is the whole post. A hub-settled PPA is not a fixed price. It is a fixed price plus an unhedged, uncapped, unfloored locational exposure that the contract never mentions by name.
Return to the Norton Rose hour with the formula. Node $1,000, hub $9,000, strike $40. The generator receives $1,000 from the market and owes $8,960 on the settlement — a net of minus $7,960 per megawatt-hour, against a strike it thought was $40. On 300 MWh that is the $2.4m.
Who Bears the Basis, and What Does It Cost to Move It?
The settlement point decides it, and the price moves to compensate.
| Node-settled | Hub-settled | |
|---|---|---|
| Generator receives | The strike, always | Strike + basis |
| Who bears basis | The offtaker | The generator |
| Generator's hedge | Complete | Partial |
| Offtaker's hedge | Against a node it has no relationship with | Against a more liquid, less volatile index |
| Strike price | Lower | Higher |
The trade is real and it is priced. Where the buyer takes on basis risk, it typically secures a lower fixed rate — the generator, de-risked, will accept less. Where the generator takes it, the strike has to carry it.
The offtaker's reason for preferring a hub is not obstinacy. Hub prices are aggregations of nodes across a broader geography and are materially less volatile than any single node. A corporate buyer hedging its own electricity cost wants an index that behaves like the market it buys in, not like one substation in West Texas. And a node-settled hedge can break in the buyer's own direction: if the price at the generator's node crashes while the price at the buyer's load centre stays high, the buyer pays a settlement to the generator and pays high prices to its own utility.
So both parties have a real argument, and neither is trying to hide anything. The failure mode is not bad faith. It is that the exposure is priced into the strike as a single number agreed years before, while the underlying congestion evolves for twenty.
Why Does a P50 Survive the Model?
Because a P50 measures the only thing that did not go wrong.
An energy yield assessment answers a volume question: how many megawatt-hours will this project produce, with what uncertainty? It is a careful, well-governed number, and P50, P90 and P99 are honest statements about it.
Revenue is a different quantity, and it has at least three more axes:
Where the energy is sold, relative to where the contract settles — basis. When the energy is produced, relative to when prices are high — shape, or capture rate. What happens at the extremes — negative prices, price spikes, and the contractual floors that handle them.
A project can hit P50 generation to within a percent and miss its revenue case badly on all three, because none of them are generation. That is why "we came in at P48" and "we are in lock-up" are statements that can both be true in the same year, and why sponsors are sometimes genuinely surprised by it.
Shape Risk and the Capture Rate
The second axis, and it is a compounding problem rather than a static one.
Shape risk is the risk that the hours in which a project produces are not the hours in which power is valuable. A solar plant produces on a midday-weighted profile, and the achieved capture price for that profile sits below the flat baseload index — because the plant is producing hardest at exactly the hour every other solar plant in the market is also producing hardest.
For a corporate buyer, shape risk has a second meaning: the risk that the project's production profile does not match its own consumption profile. Those are related but not identical concerns, and a contract that solves one does not necessarily solve the other.
The compounding matters. Capture rates decline as penetration rises, because each new plant on the same profile depresses the price in the hours that profile occupies. A twenty-year model holding the capture rate flat is making a strong assumption about the build-out of the entire market, usually without saying so.
Note also that shape and basis are correlated, not independent. The hours of worst basis are the hours of heaviest local generation, which are the hours of highest output — which are also the hours of weakest capture. All three move together and all three move the same way. A model treating them as separate independent haircuts will understate the joint effect.
Negative Prices and the Floor
The third axis, and the one with the most contractual machinery.
During low-demand periods wholesale prices go negative, and the settlement formula then does something offtakers dislike. If the hub index is minus $20 and the strike is $40, the settlement is $60 — the buyer pays sixty dollars for power that the market was paying to take away.
The market's answer is a floating price floor. Norton Rose describes the convention: for solar projects the floor is typically set at $0; for wind projects with tax equity the floor is set at the value of the production tax credits, because that is the point below which the project will still rationally generate.
The effect of the floor is to move the exposure back to the generator. Where settlement occurs at a negative price, "the corporate offtaker ends up paying the full fixed price with no offset; the floating price is treated as zero." The generator receives the strike, loses the uplift it would otherwise have had, and still receives whatever the negative node price does to its market leg.
That floor is a material revenue item in any market with substantial negative pricing — and the previous post noted over 1,200 hours of negative prices at ERCOT hubs in 2025 alone.
There is a second-order effect worth naming, because it changes behaviour rather than just cash. Once the floor is in place, the generator's incentive in negative hours changes: it is no longer receiving an uplift for generating into a negative price, so the economic decision to keep running rests on its own market leg and, for a PTC project, on the credit. That is the same calculation the deemed generation post described from the contract side — a PTC project will rationally generate down to minus the credit divided by one minus the tax rate — and it means the floor level and the credit interact. A floor set at zero for a project that in fact holds PTCs will produce more curtailment than the parties expected, because it has removed the payment that was funding the decision to keep generating.
ℹ️ Note: The PTC-based floor for wind is a mechanism, not a fixed number. The credit is inflation-adjusted, so the floor level moves with it. Take the applicable rate for the relevant year rather than carrying forward a figure from an older reference.
Basis Is Not Symmetric
The last thing to understand about a hub-settled contract is that the distribution is not the normal one.
Ordinary basis is a steady few dollars a megawatt-hour, and that is what shows up in a generation-weighted average and in a twenty-year model. Scarcity events are different in kind: a hub can clear at thousands of dollars while a constrained node clears at hundreds, and the gap for that hour is not a few dollars but a few thousand.
A single hour at $8,000/MWh of adverse basis on a 200 MW project is $1.6m — comparable to an entire year of ordinary basis on the same project. And the exposure is uncapped in a hub-settled contract, because there is no ceiling in the formula.
The practical consequences are three. Model basis as a distribution with a tail, not as a single average. Ask whether the contract has any cap on adverse settlement, and price it if it does not. And if the project is in a pocket where congestion binds hardest in the same conditions that produce system scarcity — which is common, because both are driven by the network being stressed — treat the correlation as real rather than assuming the spike hours are randomly distributed.
Does a Physical PPA Solve This?
It changes the question rather than removing it, and which way depends on where title passes.
A physical PPA delivers energy rather than settling a difference, and the critical term is the delivery point. A busbar PPA delivers at the project's point of interconnection — title passes at the fence, and the offtaker takes everything downstream, including basis. A PPA that delivers at a hub or at the buyer's load zone leaves the generator responsible for getting the energy there financially, which reproduces the hub-settled exposure under a different name.
So "physical" and "virtual" is not the axis that matters. The axis is the delivery or settlement point, and a busbar physical PPA and a node-settled virtual PPA put the generator in substantially the same place: paid a fixed amount for what it produces at its own location, with the locational risk on the counterparty.
Two differences are worth knowing. A physical PPA brings scheduling and imbalance obligations that a financial contract does not, and those carry their own costs and penalties. And a physical offtaker — typically a utility — is often better placed to absorb basis than a corporate buyer, because it already holds transmission rights and serves load in the relevant zone. That is a genuine reason utility offtake can price better than corporate offtake on the same project, and it is not usually framed that way.
The general instruction is to read the delivery or settlement point first and the contract label second. The label tells you about accounting treatment and scheduling; the point tells you who pays for congestion.
What Can a Sponsor Actually Do About It?
Five things, in roughly descending order of effectiveness and ascending order of difficulty.
Move the settlement point. The cleanest fix and the one most often left on the table, because the strike discount is visible and the basis it removes is not. The worked comparison below is the whole negotiation.
Cap the adverse settlement. A hub-settled contract has no ceiling on the basis it can produce, which is what turns a scarcity hour into a $2.4m event. A cap on adverse settlement per hour, or an aggregate annual cap, converts an uncapped exposure into a priced one. Offtakers resist it, because the cap is precisely the protection they bought — but a cap set far out of the money costs them little and removes the tail that lenders actually size against.
Pass through congestion rights. Where CRRs are available in the relevant direction, allocating them or their proceeds to the party bearing basis aligns the hedge with the exposure. The limits from the previous post apply — availability, tenor, shape and direction — so this mitigates rather than solves.
Co-locate storage. A battery at the same interconnection point can shift energy out of the worst basis hours and into better ones, which attacks basis, shape and negative pricing simultaneously. It is the only item on this list that improves all three, which is a large part of why it has grown so fast in the most congested pockets.
Reprice the strike honestly. If none of the above is achievable, the remaining option is to carry the basis in the strike and defend it with data. A sponsor with a generation-weighted basis series and a covenant breakeven can argue for a strike that covers it. One without either will be negotiated down to the market's average assumption, which is not its project's.
How Do You Model the Gap in Excel?
As a bridge from P50 generation to realised revenue, with each axis on its own line. Not as a single "basis adjustment" plugged into a price.
The inputs
Assumptions, labelled as such:
Capacity 200 MW
P50 net capacity factor 25%
P50 generation 438,000 MWh
Strike price $40.00/MWh
Generation-weighted basis −$6.00/MWh
Negative-price hours 400 hrs
Average hub price in those hours −$15.00/MWh
Output in those hours (60% of nameplate) 48,000 MWh
The bridge
1 P50 revenue at strike 438,000 × $40.00 = $17,520,000
2 Less basis 438,000 × $6.00 = ($2,628,000)
3 Less negative-hour floor 48,000 × $15.00 = ($720,000)
(settlement floored at 0 instead of paying
strike less a negative index)
Realised revenue = $14,172,000
As a % of the modelled case = 80.9%
Nineteen percent of revenue, with P50 generation delivered exactly. Not one megawatt-hour was missed. The entire gap is location, extremes and the contract.
The tail line the bridge does not capture
Single-hour adverse basis event
200 MWh × $8,000/MWh = $1,600,000
≈ 61% of the entire annual ordinary basis cost
This belongs in a separate scenario, not in the base case. A model that buries a tail event in an annual average has averaged away the thing the lender is actually worried about.
The comparison that should decide the settlement point
Hub-settled strike $40.00/MWh
Node-settled strike (offtaker takes basis) $36.50/MWh (assumption)
Hub-settled effective = 40.00 − 6.00 = $34.00/MWh
Node-settled effective = 36.50 = $36.50/MWh
→ Node-settled is worth $2.50/MWh more to the generator
= $1,095,000 per year on 438,000 MWh
The strike discount for moving basis to the offtaker is a negotiation. The test is simply whether the discount is smaller than the basis — and a sponsor who cannot state its own generation-weighted basis has no way to answer that question and will take the higher headline number every time.
ℹ️ Note: Run this test at the lender's basis assumption as well as the sponsor's. If the two settlement points rank differently under the two assumptions, that disagreement is the negotiation, and it is better had before signature than at first drawdown.
To build the full bridge — generation-weighted basis, capture rate, floor mechanics and the settlement-point comparison — prompt Dezzmond with your generation profile and the draft PPA terms.
What Do Sponsors and Lenders Actually Check?
- Which index does the CfD settle against — node, hub, or zone, named precisely?
- What is the strike discount for moving to node settlement, against the generation-weighted basis it would remove?
- Is there a floor, and at what level — zero, the PTC value, or none?
- Is adverse settlement capped in any way? In a hub-settled contract, usually not.
- What capture rate is assumed, and does it decline as penetration rises?
- Are basis, shape and negative hours modelled as correlated? They are.
- What does a single scarcity hour with adverse basis cost, and is that scenario run at all?
Frequently Asked Questions
What is a hub-settled PPA?
A contract for differences whose floating index is a trading hub rather than the project's own node. The generator sells at its node and settles against the hub, so it receives the strike plus basis — which for most renewable projects means the strike minus something.
Who bears basis risk in a virtual PPA?
Whichever party is not settled at its own location. Node settlement hedges the generator fully and leaves the offtaker exposed; hub settlement does the reverse. The strike price moves to compensate.
Why does a project hit P50 and still miss revenue?
Because P50 is a volume forecast. Revenue also depends on where the energy settles, when it is produced relative to price, and how negative hours are treated — none of which an energy yield assessment addresses.
What is a floating price floor?
A contractual floor on the index used for settlement, typically $0 for solar and the production tax credit value for wind with tax equity. It stops the offtaker paying more than the strike when prices go negative, and moves that exposure to the generator.
Does a physical PPA avoid basis risk?
Only if it delivers at the busbar. A physical PPA delivering at a hub or load zone puts the generator in the same position as a hub-settled virtual PPA. The delivery point decides who pays for congestion, not whether the contract is physical or financial.
How large can a single-hour basis event be?
Very large. Norton Rose records ERCOT North hub at $9,000/MWh against nodes at $1,000 — $8,000/MWh of adverse basis, worth $2.4m in one hour to a 300 MW project.
Closing: Two Right Answers to Different Questions
The energy yield consultant is asked how much this project will generate, and answers it carefully. The price consultant is asked what power will be worth, and answers that carefully too. Both are usually close to right.
The revenue is the product of neither. It is the product of a nodal price at a specific point on a network that is getting more congested, in hours determined by a resource profile that every similar project shares, under a contract whose floors and settlement index were agreed before any of that was known.
The nineteen percent in the worked bridge is not a forecasting failure. It is three separate mechanisms doing exactly what they were always going to do, in a model that had one line for revenue and no line for location.
The rest of Series C is about the market structures that produce these effects. The next post steps back to the design level: ERCOT against PJM, energy-only against capacity, and what that difference does to the revenue stack a project can actually bank.
Sources: Norton Rose Fulbright — Corporate VPPAs: Risks and Sensitivities · Pivotal180 — Corporate PPA Explained: Structure, Benefits, and Key Risks in 2026 · RMI — A Local Government's Guide to Off-Site Renewable PPA Risk Mitigation