The price of electricity is not set by averaging the costs of all running generators. It is set at the margin: the cost of the last, most expensive generator dispatched to serve one more megawatt of demand. That single principle — marginal pricing — underpins every organized wholesale market in the United States, and understanding it unlocks nearly everything else about power economics.
The merit order: the supply stack
Start with the mental model every market participant uses: the supply stack (or merit order). Rank all available generators by their offer price, cheapest first. At the bottom sit resources with near-zero marginal cost — wind, solar, run-of-river hydro, and nuclear plants, whose fuel costs are negligible or sunk. In the middle: coal plants and efficient combined-cycle gas turbines, with moderate fuel costs. At the top: gas-fired peakers and oil units, expensive to run but quick to start.
To serve a given level of demand, the operator dispatches units from the bottom of the stack upward until demand is met. The most expensive unit called upon — the marginal unit — sets the clearing price, and every dispatched generator receives that same price for its output. This is the uniform clearing price auction, and it is the core price-formation mechanism of the energy market.
A worked illustration
Consider an illustrative system with 40,000 MW of demand in a given hour. The operator dispatches, for illustration: 15,000 MW of nuclear, wind, and solar offered at $0–10/MWh; 15,000 MW of coal and combined-cycle gas at $25–40/MWh; and 10,000 MW of gas peakers at $65/MWh. The marginal unit’s offer is $65, so the clearing price is $65/MWh — paid to every dispatched generator.
Notice what this means for revenue. The nuclear plant with a $10 marginal cost earns a $55/MWh gross margin; the peaker that set the price earns zero margin on its energy. Those margins on low-cost units, earned during hours when expensive units set the price, are called inframarginal rents, and they are the mechanism by which generators recover their fixed costs — the billions spent building the plant. A generator’s business case is not “earn a margin on every MWh” but “earn enough inframarginal rent across the year to cover capital costs.” This is the single most important concept in generation economics.
Why uniform pricing instead of pay-as-bid?
A natural question: why pay every generator the clearing price rather than each its own bid? The answer is incentives. Under pay-as-bid, every generator would try to guess the clearing price and bid just under it — bidding becomes a guessing game, and the operator loses visibility into true costs. Under uniform pricing, a generator’s best strategy is to bid its true marginal cost: bidding higher risks not being dispatched at all, and bidding lower doesn’t change what it gets paid (the clearing price is set by the marginal unit). Economists call this incentive compatibility, and it is why uniform-price auctions dominate electricity market design worldwide.
Locational marginal pricing: one grid, thousands of prices
Electricity is not priced the same everywhere on the grid. Locational marginal pricing (LMP) computes a distinct price at thousands of nodes — every generator bus and load zone — every five minutes. Each LMP decomposes into three components:
1. The energy component — the system-wide marginal cost of energy, identical at every node. This is the price that would prevail if the transmission network had infinite capacity.
2. The congestion component — the cost imposed when transmission lines reach their limits. If cheap generation in one area cannot reach load in another because the lines between them are full, the operator must dispatch more expensive local generation instead. The price difference between the two areas is the congestion cost. Congestion components are positive at nodes where expensive local generation must run (load pockets) and negative where cheap generation is bottled up (export-constrained areas).
3. The losses component — a smaller adjustment for the energy dissipated as heat when power flows over distance. Typically a few percent of the total.
When no transmission constraints bind, congestion components are zero and LMPs converge across the region to the energy component plus losses. When lines bind, prices separate — sometimes dramatically.
Congestion: the signal in the noise
Persistent congestion price differences are among the most informative data in power markets. For illustration: suppose a wind-rich zone consistently prices at $15/MWh while a nearby city prices at $45/MWh, a $30 spread driven by a constrained transmission path. That spread is simultaneously three things: a cost to load in the city, a revenue haircut to the wind farms, and a glowing signal to developers and transmission planners that a new line — or new generation near the city — would create value.
This is why market participants obsess over congestion patterns. They reveal where the grid is inadequate, and they are the price signal that justifies transmission investment. The RTO’s transmission planning process exists largely to identify congestion that is persistent and costly enough to warrant building wires.
Negative prices: when generators pay to run
One of LMP’s more startling outputs is the negative price. When must-run generation (nuclear plants that cannot easily cycle, renewables with production tax credits that earn revenue per MWh generated) exceeds demand in an export-constrained area, the market price can fall below zero — generators effectively pay to keep running. For a wind farm earning a federal production tax credit per MWh, offering at negative prices down to the credit’s value is rational: it still profits overall.
Negative prices are not a market failure; they are accurate information. They say: at this location and hour, there is too much inflexible supply. The efficient responses — storage to absorb midday surplus, transmission to export it, flexible demand to consume it — are exactly the investments negative prices incentivize.
Managing congestion risk: FTRs
Because congestion creates winners and losers unpredictably, markets offer financial transmission rights (FTRs) — instruments that pay the holder the congestion price difference between two points. A load-serving entity buying power in a cheap zone to serve load in an expensive zone can buy FTRs to hedge the congestion spread. FTRs are auctioned by the RTO, and the auction revenue is returned to transmission owners or load. They complete the market’s risk-management toolkit alongside energy forwards.
Day-ahead vs. real-time LMPs: expectation vs. reality
Every node has two LMPs that matter: the day-ahead LMP, set in the daily auction based on forecasts, and the real-time LMP, set every five minutes based on actual conditions. The day-ahead price is the market’s expectation; the real-time price is reality’s verdict. The difference between them — the spread — is closely watched.
Averaged over time, the two should roughly converge: persistent day-ahead premiums suggest physical suppliers are bidding strategically or load is systematically under-scheduling; persistent real-time premiums suggest the opposite. Virtual traders arbitrage these spreads, and their activity is the mechanism that drives convergence. But the spreads at individual moments can be enormous — a plant trip or forecast error can send real-time LMPs hundreds of dollars above day-ahead for a handful of intervals. Those transient spreads are the price of uncertainty, and they compensate the flexible resources that stand ready to balance the system.
For analysts, the day-ahead/real-time spread is a diagnostic tool: wide, persistent spreads in one direction flag design problems, while occasional spikes are the normal cost of balancing a physical system against imperfect forecasts.
Nodal vs. zonal: a design choice
U.S. organized markets price at the nodal level — thousands of individual prices. Some markets elsewhere (and Texas’s earlier design) used zonal pricing, averaging within large zones. Nodal pricing is more economically precise: it reveals exactly where constraints bind and gives sharper investment signals. Its cost is complexity — more prices to forecast, more basis risk to hedge. The U.S. has consistently chosen precision, and nodal LMP is now the global gold standard for wholesale price formation.
The losses component: small but structural
The third LMP component — marginal losses — rarely makes headlines, but it quietly shapes the grid. When power flows from a distant generator to load, a few percent dissipates as heat in the wires. The marginal loss factor at a node reflects how total system losses change if one more MW is injected there: injecting power near load reduces losses (positive for the injector), while injecting far away increases them.
For illustration, a generator 500 miles from load might face a losses component of negative $2–3/MWh relative to the reference bus — a small but persistent haircut that, compounded over thousands of hours, meaningfully affects siting economics. This is economically correct: distant generation really does impose greater losses on the system. But it creates a subtle tension with renewable development, since the best wind and solar resources are often far from load centers. Transmission expansion effectively arbitrages this away — a new line reduces both congestion and losses — which is another reason transmission investment shows up repeatedly as the highest-leverage action in power markets.
Some markets have debated whether the complexity of marginal loss pricing is worth it versus simpler average-loss approaches. The consensus has held for marginal pricing: like congestion, losses are a real physical cost of delivery, and hiding them in averages would subsidize remote generation at load’s expense. It is a small component with an outsized lesson — in electricity markets, even the rounding errors are priced, because the physics gives you no place to hide them.
The critique: the “missing money” problem
Marginal pricing has a well-known vulnerability. In theory, inframarginal rents plus occasional scarcity prices should cover generators’ fixed costs. In practice, price caps, conservative scarcity pricing, and the influx of zero-marginal-cost renewables can suppress prices below the level needed to sustain investment — the missing money problem. A gas plant that the system needs for reliability might not earn enough from energy markets alone to stay open.
This is not an argument against marginal pricing — it is the central design challenge of electricity markets. The responses are the other markets in this series: capacity markets that pay explicitly for availability, and scarcity pricing reforms that let energy prices rise to reflect the true value of reliability. The debate over how to fix missing money, without abandoning the efficiency of marginal pricing, is the defining policy argument in wholesale market design.
Reading LMP data like a practitioner
A few habits for interpreting locational prices: compare a node’s price to the system energy component to isolate congestion; look at the shape of prices across the day (off-peak vs. on-peak spreads reveal the marginal fuel); watch for persistent negative pricing as a siting signal for storage; and remember that day-ahead LMPs embed expectations while real-time LMPs reveal surprises. The price at any node, at any moment, is the market’s summary judgment on the physical state of the grid — learn to read it, and the grid starts talking to you.