ERCOT vs PJM: Energy-Only, Capacity Markets, and Two Revenue Stacks Sitting on Administrative Ceilings

On 14 July 2026, PJM's capacity auction for 2028/2029 cleared at $325/MW-day — the FERC-approved cap, for the entire footprint. It was the third consecutive auction to clear at its ceiling. PJM's own estimate for the prior auction was that without the cap it would have cleared at $193,907/MW-year, roughly 60% above the capped result.

Seven months earlier, on 5 December 2025, ERCOT went live with Real-Time Co-optimization Plus Batteries, replacing a single $5,000/MWh system-wide offer cap with a day-ahead cap of $5,000 and a real-time cap of $2,000.

Two markets with opposite designs, and in both the revenue that project finance depends on is currently being set by an administrative parameter rather than by supply and demand. In PJM the capacity price is pinned at a ceiling. In ERCOT the scarcity price ceiling just fell by 60%.

This post covers what the two designs actually do, how each pays a generator, and what the difference means for a revenue stack you have to bank for fifteen years.

ℹ️ Note: This describes market structure as at September 2026. Market rules change — RTC+B is nine months old and PJM's capacity construct is under active reform. Treat every figure as a point-in-time reading, not a forecast.

What Is the Actual Difference?

PJM pays generators to exist; ERCOT pays them only to generate. PJM runs a forward capacity auction procuring a reliability obligation years ahead at a $/MW-day price. ERCOT has no capacity market at all and relies on energy prices rising steeply when supply is short.

Everything else — the accreditation debates, the offer caps, the scarcity mechanisms — follows from that single choice about whether reliability is bought explicitly or produced as a by-product of price.

How PJM Pays for Capacity

Through the Reliability Pricing Model, a forward auction held for a delivery year three years out.

The Base Residual Auction procures unforced capacity — UCAP — from generation and demand response, and clears at a price in dollars per megawatt-day. A resource that clears receives that price for every accredited megawatt across the delivery year, whether or not it runs.

The scale is substantial. The 2028/2029 auction secured 138,318 MW of UCAP to serve more than 67 million people across thirteen states and the District of Columbia, and the cleared supply times the clearing price totals $16.4bn.

For a project, the arithmetic is straightforward:

Capacity revenue = Accredited MW × $/MW-day × 365

At $325/MW-day that is $118,625 per accredited megawatt-year — and the phrase carrying all the weight is "accredited megawatt," which is the subject of the next post.

The Cap Is the Story

Three auctions, three clearing prices at the ceiling.

Delivery year	Cleared	Result
2026/2027	~$329/MW-day	At cap
2027/2028	$333.44/MW-day	At cap
2028/2029	$325.00/MW-day	At cap

The 2028/2029 price is 2.5% lower than 2027/2028, which reads like softening and is not. Both cleared at their respective caps; the caps themselves differ. The market did not find a lower price — it hit a lower ceiling.

The suppressed price is the number worth carrying. For 2027/2028, PJM estimated that absent the cap and floor the auction would have cleared at $193,907/MW-year against the capped $121,705 — about 60% higher. That gap is a measure of how far the administrative parameter is from the market-clearing outcome, and it has a direct consequence for anyone modelling capacity revenue: the observed price is not an equilibrium, so it carries no information about where an uncapped price would go if the cap moved.

A model extrapolating $325/MW-day forward is extrapolating a policy setting. That may be the right assumption, but it should be labelled as a view on regulation rather than a view on the market.

Why Is the Price at the Cap?

Because demand is rising faster than supply can clear the queue, and the cap exists precisely to stop that translating into an unbounded price.

Three forces are doing the work. Load is growing sharply, driven by data centre development concentrated in particular zones — the same growth the previous post identified as the source of PJM's worsening congestion, now showing up in the capacity construct as well. Thermal capacity is retiring. And new capacity, much of it renewable and storage, is moving through an interconnection process slowly enough that it cannot arrive in time to bid into the auctions where it is needed.

A capacity auction responds to that combination the only way it can: by clearing higher until either demand response or new supply appears. FERC's approved price collar interrupts that escalation, with a floor as well as a ceiling — for the 2027/2028 auction the floor was $65,535/MW-year against a cap of $121,705/MW-year.

Two things follow that matter for modelling.

The collar is symmetric, and that cuts both ways. A floor protects generator revenue in a soft market. A cap removes upside in a tight one. A project banking capacity revenue is therefore short an option it is not paid for — it holds the downside protection and has sold the upside to the construct.

The cap will move. It is a FERC-approved parameter, revisited under pressure from states, load-serving entities and generators with directly opposing interests. The 2.5% fall from $333.44 to $325 illustrates the point: neither figure was discovered, both were set, and the change between them was administrative.

The honest way to model this is as a policy variable with a range, not a price series with a trend.

What Is the Fixed Resource Requirement?

An exit door, and it is being used at a scale worth noticing.

Load-serving entities can elect the Fixed Resource Requirement alternative, self-supplying their capacity obligation rather than procuring it through the auction. For the 2028/2029 delivery year, FRR regions acquired an additional 10,864 MW of UCAP, bringing the total to 149,182 MW against the 138,318 MW cleared in the auction itself.

That is roughly seven percent of the total sitting outside the auction — and the direction of travel matters more than the level. Where a state or a utility concludes that auction prices at the cap are worse than self-supply, it can leave, taking both its load obligation and, potentially, the generation serving it.

For a project, the consequence is a thinner and more concentrated auction. A capacity market with a shrinking participating footprint is a less liquid one, more sensitive to the entry or exit of individual large parties, and harder to forecast from its own history.

It is also a reminder of what a capacity price actually is. It is not the value of reliability. It is the price at which a specific set of buyers, under a specific set of rules, procure a specific administratively-defined product — and any of those three can change.

How ERCOT Pays for Scarcity Instead

By letting energy prices rise, and by pricing reserve shortfalls into the energy price.

An energy-only market compensates generators for energy produced, with no capacity payment. Reliability incentives come from a scarcity pricing mechanism that raises prices when supply is tight, and an offer cap that limits how far they can go.

Historically ERCOT did this through the Operating Reserve Demand Curve, which applied a price adder to real-time energy prices as reserves fell — penalties for ancillary service deficiency were applied indirectly, through the ORDC adder.

That changed on 5 December 2025. Under RTC+B:

• Security-constrained economic dispatch now co-optimises energy and ancillary services together, every five minutes.
• Ancillary Service Demand Curves — price-responsive demand curves — enter the real-time co-optimisation directly, and the ORDC-related price adders are going away.
• The single $5,000/MWh system-wide offer cap became a day-ahead cap of $5,000 and a real-time cap of $2,000, with the system lambda setting LMPs capped at the effective value of lost load, currently the $5,000 day-ahead figure.
• The supplemental ancillary service market is eliminated, and virtual ancillary service participation enters the day-ahead market for the first time.

The expected direction of travel, on Yes Energy's reading, is "reduced RT energy prices on average over time" through more efficient dispatch, with virtual participation adding liquidity and potentially narrowing day-ahead to real-time spreads.

For a merchant revenue model that is a material change in both directions: lower average real-time prices, a halved real-time ceiling, and compressed arbitrage spreads.

The effect is not uniform across technologies, and that is the part most likely to be modelled wrongly. Co-optimisation means a battery no longer has to choose between bidding into energy and holding an ancillary service award — the dispatch engine decides every five minutes which use is worth more, and the resource is paid accordingly. In principle that raises the value of a flexible asset by removing a choice it was previously forced to make badly.

Against that, the same convergence compresses the spreads a battery earns from. Yes Energy's assessment is that near-term volatility at launch creates "opportunities for participants who are ready," while greater price convergence between day-ahead and real-time markets compresses arbitrage over time.

So the storage revenue stack in ERCOT has been rebuilt rather than trimmed, and a model that simply applies a haircut to a pre-December-2025 ancillary services history is doing the wrong operation on the wrong data. The revenue stack for batteries is the subject of a later post in this series, and RTC+B is the reason it needed one.

What Does This Do to the Revenue Stack?

	PJM	ERCOT
Capacity payment	Yes — $/MW-day, 3 years forward	None
Where reliability value sits	A separate, contracted revenue line	Inside the energy price
Real-time ceiling	Energy offer cap	$2,000/MWh (from Dec 2025)
Day-ahead ceiling	—	$5,000/MWh
Scarcity mechanism	Capacity obligation + penalties	ASDCs inside co-optimisation
Forward visibility	3 years, then nothing	None
What a lender can size	Capacity for 3 years; energy on a curve	Energy on a curve only

The practical difference for financing is not that one market pays more. It is where the money sits and how far forward it is visible.

PJM gives a project a discrete revenue line with a known price for a known period — but only three years, against a debt tenor of fifteen or more. ERCOT gives no such line at all, but puts the same value into an energy price that a project may be able to hedge with a PPA.

That is the reason the two markets attract different capital. ERCOT's revenue is volatile but contractible: a PPA can fix it. PJM's capacity revenue is contracted but only briefly, and there is no long-dated instrument that fixes it.

It is worth naming the asymmetry directly, because it inverts the intuition. A capacity market sounds like the safer design for a lender — a known price, paid whether or not the asset runs, procured years in advance. But the horizon is three years, and no counterparty will write a fifteen-year capacity hedge, because nobody can price one against a construct that has been reformed repeatedly. An energy-only market sounds riskier and offers something a capacity market does not: a deep, liquid, long-dated hedging market in the product itself. A fifteen-year PPA in ERCOT is ordinary. A fifteen-year capacity contract in PJM is not a thing.

What Can a Lender Actually Size Against?

Less than the revenue stack contains, in both markets.

In PJM, capacity revenue beyond the cleared auction horizon is typically haircut heavily or excluded. Three years of known price against a fifteen-year tenor leaves twelve years of capacity revenue resting on a forecast of both the auction outcome and the regulatory construct producing it — and the construct has changed materially more than once. A common treatment is to bank cleared years at contract and apply a substantial discount thereafter, with the accreditation itself assumed to decline.

In ERCOT, the equivalent question is whether scarcity revenue is bankable at all. The honest answer for most structures is no: it is the thinnest, most volatile and most rule-dependent part of the stack, and the December 2025 cap change is a live demonstration of why. A project financed on scarcity revenue in November 2025 had its real-time ceiling halved a month later by a market design decision.

The general principle is one worth stating plainly, because it applies well beyond these two markets: revenue created by an administrative parameter can be removed by an administrative decision, and lenders price it accordingly. Contracted revenue is bankable. Merchant revenue is haircut. Revenue that exists because a regulator set a number is haircut hardest of all.

How Do You Model the Two Stacks in Excel?

As different stacks, not as the same stack with different prices. The line items are not the same.

The inputs

Assumptions, labelled as such:

Capacity                                        200 MW AC
Annual generation                             438,000 MWh
Capacity accreditation (PJM)                          8%   ← see next post
Capacity clearing price                    $325/MW-day
Merchant energy price                        $40.00/MWh

The PJM stack

Capacity revenue
   Accredited MW      200 × 8%                =      16 MW
   Annual rate        $325 × 365              = $118,625/MW-yr
   Capacity revenue   16 × 118,625            =  $1,898,000

Energy revenue        438,000 × $40.00        = $17,520,000

Total stack                                    = $19,418,000
Capacity as % of total                        =        9.8%

The ERCOT stack

Energy revenue        438,000 × $40.00        = $17,520,000
Capacity revenue                              =           $0

Total stack                                    = $17,520,000

The comparison that actually matters

PJM total                                      = $19,418,000
ERCOT total                                    = $17,520,000
Difference                                     =  $1,898,000  (10.8%)

Bankable in PJM (yrs 1-3)                      = $19,418,000
Bankable in PJM (yrs 4-15, 50% capacity haircut assumption)
                                               = $18,469,000
Bankable in ERCOT                              = $17,520,000

The capacity line looks like a ten percent revenue uplift and finances like five, because most of the tenor sits beyond the auction horizon.

The sensitivity that dominates everything else

Accreditation    8%  →  capacity revenue  $1,898,000
Accreditation   15%  →  capacity revenue  $3,559,000
Accreditation   30%  →  capacity revenue  $7,118,000
Accreditation   60%  →  capacity revenue $14,235,000

Nothing else in the PJM stack moves like this. A change in accreditation methodology can multiply or eliminate the capacity line without the clearing price moving a cent — which is why the next post is about accreditation rather than about price.

ℹ️ Note: Model the capacity price and the accreditation as separate inputs, never as a single $/MW-of-nameplate figure. Combining them hides the variable that moves most and makes the sensitivity above impossible to run.

To build both stacks, the accreditation sensitivity and the bankability haircut schedule, prompt Dezzmond with your project's market and technology.

How Do You Model a Market That Keeps Being Redesigned?

Not by forecasting the rules, which nobody can do, but by making the model's dependence on them visible.

Three techniques are worth more than a better central case.

Separate rule-dependent revenue from rule-independent revenue. Energy sold under a PPA does not care what the offer cap is. Scarcity revenue is entirely a function of it. Capacity revenue exists only because a construct exists. Reporting those as one number destroys the information a lender most wants, which is how much of the stack survives a rule change. Report the share of revenue that would disappear if the relevant parameter were removed — and be prepared for that share to be uncomfortable.

Re-run history under current rules, not historical rules. An ERCOT model calibrated on 2021–2024 real-time prices is calibrated on a market with a $5,000/MWh real-time cap, ORDC adders and a separate supplemental ancillary service market. None of those exist now. Backcasting the same years under the current construct is laborious and is the only honest way to use the history at all.

Treat a parameter change as a scenario, not a tail. The December 2025 cap change was announced, consulted on and implemented on a known date. It was not a shock; it was a process with a published timetable that a model could have followed. The same is true of PJM's cap and floor. Regulatory scenarios belong alongside the price scenarios, with the same discipline applied to them.

The underlying point is that "merchant risk" is usually described as price risk and is substantially rule risk. A project can be right about supply, demand, load growth and fuel costs and still lose a revenue line because a ceiling moved.

What Do Sponsors and Lenders Actually Check?

• Is capacity revenue modelled on accredited MW, not nameplate?
• How many delivery years are actually cleared, and what is assumed beyond them?
• Does the model treat the clearing price as an equilibrium? It is currently a cap.
• In ERCOT, what share of revenue depends on prices above $2,000/MWh in real time, and has that been re-run since December 2025?
• Are ancillary service revenues modelled post-RTC+B, or on pre-co-optimisation history?
• What happens to the DSCR if capacity revenue is excluded entirely beyond the cleared years?
• Is the merchant tail in either market assumed to look like the recent past? In both, the rules that produced the recent past have changed.

Frequently Asked Questions

What is the difference between an energy-only and a capacity market?

A capacity market pays generators a separate price to be available, procured in advance of the delivery year. An energy-only market pays only for energy produced, and relies on prices rising steeply during scarcity to fund reliability.

What did PJM's latest capacity auction clear at?

The 2028/2029 Base Residual Auction cleared at $325/MW-day UCAP for the entire footprint — the FERC-approved cap — procuring 138,318 MW at a total cost of $16.4bn. It was the third consecutive auction to clear at its ceiling.

What is ERCOT's offer cap now?

Since RTC+B went live on 5 December 2025, there are two: a day-ahead system-wide cap of $5,000/MWh and a real-time cap of $2,000/MWh, replacing the previous single $5,000/MWh cap.

Can capacity revenue be financed?

Only for the cleared delivery years with any confidence. Beyond the three-year auction horizon it is a forecast of both an auction and the regulatory construct behind it, and lenders typically haircut it substantially or exclude it.

Why did PJM's price fall from $333.44 to $325?

Because the cap fell, not because the market softened. Both auctions cleared at their respective FERC-approved ceilings, so the change between them is a change in a parameter rather than a change in supply and demand.

What is the Fixed Resource Requirement?

An alternative under which a load-serving entity self-supplies its capacity obligation instead of buying through the auction. FRR regions held an additional 10,864 MW of UCAP for 2028/2029, which is capacity and load sitting outside the cleared auction entirely.

Which market is better for a renewable project?

Neither, structurally. PJM adds a capacity line that is real but short-dated and heavily dependent on accreditation. ERCOT has no such line but higher energy volatility, which a PPA can convert into a fixed price. The answer depends on whether the project is contracted or merchant.

Closing: Two Designs, One Problem

The energy-only versus capacity-market argument is usually framed as a question of theory — whether scarcity pricing can fund new entry, whether a capacity construct over-procures. That argument is thirty years old and unresolved.

What the last twelve months show is something more immediate and more relevant to anyone financing a project. Both designs are currently producing revenue numbers that are set administratively. PJM's capacity price has hit its ceiling three auctions running, with the unconstrained price estimated sixty percent higher. ERCOT's real-time scarcity ceiling was halved by a market design change that took effect in December 2025.

Neither of those is a market outcome. Both are parameters, and parameters move by decision rather than by supply and demand — which means the usual modelling instinct, of extrapolating the observed series forward, is extrapolating the wrong thing.

The next post takes the variable that moves PJM capacity revenue more than the clearing price ever does: capacity accreditation, ELCC, and why the credit given to solar and storage keeps being cut.

Sources: PJM — 2028/2029 Base Residual Auction Report · PJM — Capacity Auction Procures 138,318 MW (news release, 14 July 2026) · ERCOT — RTC+B Go-Live, 5 December 2025 · Yes Energy — ERCOT RTC+B Market Redesign FAQ Part II