A persistent source of confusion in electricity markets: generators sell two different products, and each has its own market. The energy market pays for megawatt-hours actually produced. The capacity market pays for megawatts committed to be available when needed — including hours when they never run. They solve different problems, they have different designs, and conflating them makes both harder to understand.
The energy market: pay for production
The energy market is the marginal-pricing auction described earlier in this series: generators offer output, the operator dispatches the merit order, and every dispatched unit earns the locational clearing price. For units that run most hours — nuclear plants, efficient combined-cycle gas turbines, renewables — energy revenue is the core of the business case, earned MWh by MWh across thousands of hours a year.
But consider a combustion-turbine peaker: cheap to build (a few hundred dollars per kW), expensive to run, called upon perhaps 50 to 200 hours a year during extreme peaks. Its energy revenue — earned only in those few hours, often at prices it barely sets — cannot plausibly cover its fixed costs of construction, financing, and maintenance. Yet the system needs that peaker: without it, the hottest afternoons of the year would bring blackouts. This gap between what energy markets pay and what reliability requires is the missing money problem, and it is the reason capacity markets exist.
The missing money problem, precisely
In textbook economics, the problem shouldn’t exist. During scarcity hours, energy prices should spike to the value of lost load (VOLL) — what consumers would pay to avoid an outage, often estimated at many thousands of dollars per MWh. A peaker running 100 hours a year at VOLL-based prices would earn plenty. In practice, three things blunt this mechanism: administrative price caps (typically $1,000–2,000/MWh in eastern markets, higher in ERCOT), political intolerance for extreme prices, and market power mitigation rules that limit what dominant suppliers can bid. The result is that energy prices alone systematically underpay the capacity the system needs at the peak.
Policymakers face a choice: let energy prices spike freely and accept the political and financial consequences, or create a separate payment for availability. Most eastern U.S. markets chose the second path.
The capacity market: pay for availability
A capacity market is a forward auction for committed availability. Instead of hoping scarcity prices materialize, the market procures reliability explicitly: load-serving entities collectively buy enough capacity commitments to cover forecasted peak demand plus a reserve margin, typically one to three years ahead.
The mechanics: the RTO determines how much capacity the system needs (the target, based on peak forecasts and reliability standards). Generators offer their available capacity at the price they require. A sloped demand curve sets the price — if offered supply exactly equals the target, the price lands at the administratively determined cost of new entry (CONE, the annualized cost of building a reference peaker); excess supply pushes the price down the curve; a shortfall pushes it up. This avoids the cliff-edge pricing of a vertical demand curve, where a tiny shortfall would send prices to the cap.
Generators that clear the auction receive capacity payments — typically quoted in dollars per MW-day — for the commitment period, whether or not they are ever dispatched. In exchange, they take on obligations: to offer into the energy market, to be available during stressed conditions, and in modern designs, to face penalties if they fail to perform when called.
Making capacity real: performance and accreditation
Early capacity markets had an embarrassing flaw: they sometimes paid for “paper capacity” — resources that collected payments but weren’t there when needed. Reforms have focused on making the product real.
Performance requirements tie payment to delivery. PJM’s Capacity Performance construct, the template for modern design, requires committed resources to perform during emergency conditions and imposes steep penalties for non-performance — while paying bonuses to resources that over-perform. The economic logic: capacity is an option on emergency energy, and the option must be exercisable.
Accreditation addresses a subtler problem: not all megawatts contribute equally to reliability. A gas plant available 95% of peak hours is worth more than a solar farm that produces nothing after sunset. Modern markets assign each resource a capacity credit reflecting its expected contribution during risk hours — using methods like effective load-carrying capability (ELCC) that model a resource’s statistical contribution to meeting peak. As wind, solar, and storage grow, accreditation methodology has become one of the most technically contested areas in market design, because it directly determines how much new renewables and batteries can earn in capacity auctions.
The energy-only alternative: ERCOT’s wager
Texas chose the other path. ERCOT runs no centralized capacity market. Instead, it relies on scarcity pricing: an operating reserve demand curve (ORDC) that administratively adds to energy prices as reserves shrink, with a high system-wide price cap (historically $9,000/MWh). The theory is pure: let prices reflect true scarcity, and investment will follow the money.
ERCOT’s design has real virtues — simplicity, no capacity bureaucracy, and powerful incentives for flexible resources like batteries, which thrive on price volatility. It also concentrates risk: the revenue that sustains the fleet arrives in a small number of extreme hours, making investment depend on investors’ belief that those prices will actually be allowed to materialize. The political economy is unforgiving — prices high enough to attract peakers are also high enough to generate enormous bills during extended events, testing the durability of the design each time scarcity persists for days rather than hours.
State intervention and the boundary problem
Capacity markets have been repeatedly destabilized by a tension with state policy. When states subsidize preferred resources (renewables, nuclear plants) for environmental or economic reasons, those resources can offer into capacity auctions at artificially low prices, suppressing clearing prices for everyone. Market operators responded with minimum offer price rules (MOPR) — floors on what subsidized resources could bid — which states viewed as a tax on their policy choices. The resulting FERC-state conflicts produced years of litigation and rule revisions.
The deeper issue is jurisdictional: FERC governs wholesale markets, states govern generation policy and retail rates, and the two increasingly collide as states pursue clean-energy mandates within market regions. Each round of reform — narrower MOPRs, state policy carve-outs, evolving accreditation — is an attempt to let both systems coexist. None has fully resolved the tension, and it remains the most litigated frontier in wholesale market design.
Seasonal and locational capacity: refining the product
The basic capacity product — a MW available year-round — is a blunt instrument, and markets have been refining it along two axes. Seasonal differentiation recognizes that risk isn’t uniform across the year: a winter-peaking system (like ISO New England, with its gas-constrained cold snaps) needs different resources than a summer-peaking one, and a solar farm’s capacity value differs by season. MISO’s move to seasonal capacity auctions was an acknowledgment that a single annual product was mispricing seasonal risk — procuring too much in mild seasons and too little when it mattered.
Locational requirements recognize that capacity must be deliverable, not just existent. A megawatt in an export-constrained zone can’t help an import-constrained city if the transmission between them is full during the peak. Markets like NYISO and PJM therefore set separate capacity requirements — and produce separate clearing prices — for constrained zones. For illustration, capacity in New York City has historically cleared at multiples of the rest of the state, reflecting the transmission limits into the city. Those locational price differences are investment signals as legible as energy congestion: they say “build generation or transmission here.”
Both refinements make capacity markets more complex to model but more accurate in what they buy — continuing the long arc from paper capacity toward verified, deliverable, correctly accredited reliability. For investors, the practical implication is that capacity revenue is becoming more granular and more forecastable at the zonal level, rewarding those who model locational and seasonal dynamics rather than treating capacity as a single system-wide number.
The investor’s lens: two revenue streams
Strip away the policy debate and the distinction is practical. Energy revenue is volatile and weather-driven — it rewards low-cost, well-located generation and punishes high-cost plants. Capacity revenue is contracted and predictable — typically locked in one to three years forward, it underwrites financing for new builds and keeps marginal existing plants open.
Generation portfolios are valued on the blend. A nuclear plant in PJM might earn the majority of its revenue from energy but rely on capacity payments for the margin that keeps it economic. A new battery project stacks capacity revenue with energy arbitrage and ancillary services. When capacity prices crash due to oversupply — as they periodically do — the pain falls hardest on the marginal thermal units whose energy margins are thinnest, accelerating retirements and eventually tightening the market again. It is a cycle, and capacity auction results are the clearest public readout of where a region sits in it.
How the demand curve works: CONE and the sloped curve
The capacity demand curve deserves a closer look, because its parameters move billions of dollars. The anchor is CONE — the cost of new entry: the annualized cost (capital recovery plus fixed operations) of building and operating a reference peaking plant, for illustration often estimated in the range of $80,000–$120,000 per MW-year depending on the region and technology assumptions. When offered supply exactly equals the target, the auction clears at CONE — the price at which a new peaker would theoretically break even.
The sloped demand curve means the price isn’t all-or-nothing. For illustration: if the system is 5% long on capacity, the price might clear at 60% of CONE; if 5% short, it might clear at 150% of CONE, up to a cap. The slope’s steepness is a policy choice — steeper curves produce more volatile prices but stronger investment signals; flatter curves stabilize prices but weaken the signal. Every few years, RTOs recalibrate CONE and the curve shape, and each recalibration is a high-stakes proceeding because it resets the revenue baseline for the entire existing fleet.
An important subtlety: CONE is a net concept in most markets — Net CONE subtracts the energy and ancillary service revenue the reference plant is expected to earn, so the capacity price only needs to cover the residual. This ties the two markets together mechanically: when energy prices are high, Net CONE falls and capacity prices soften; when energy markets are depressed, capacity prices must rise to compensate. The energy-capacity revenue seesaw is one of the most important dynamics in generation economics.
What to watch
For anyone following a capacity market region, the key indicators are: the clearing price relative to CONE (at or near CONE signals a balanced market; far below signals surplus); the accredited supply mix (how much is thermal vs. intermittent vs. storage); performance during actual stress events (do committed resources deliver?); and the pace of rule reform (accreditation changes can revalue entire fleets overnight). Capacity markets are where reliability policy becomes dollars — and where the dollars reveal whether the policy is working.