When the Price Goes Negative
Must-run units, subsidised renewables that earn more by generating than by stopping, and transmission that cannot carry the surplus out.
In California in 2024, the wholesale price of electricity was below zero for roughly 1,180 hours. That is about 13% of the year, up from around 530 hours in 2023, and the typical negative price got deeper as well, with a median near minus 17 dollars against minus 10 the year before. In Germany, the day-ahead price was negative for 457 hours in 2024, up from 301 in 2023, and by late August 2025 the count for that year had already reached 453.
A negative price means the seller pays the buyer. Consuming electricity in those hours is a service somebody is willing to fund. Readers arriving from Oil 101 will remember the one famous instance of a negative price in that market, when WTI settled at minus 37.63 dollars on 20 April 2020. The instinct is to file this alongside it. That instinct is wrong, and correcting it is most of this chapter.
The oil episode lasted part of one afternoon, involved one expiring contract, and happened because the physical storage at Cushing was effectively full, so a buyer of a contract requiring delivery had nowhere to put the barrels. It was a storage failure, and it was over the next day. What happens in power markets is structural. It recurs for thousands of hours a year, was entirely predictable from the design, and is becoming more common every year.
Why anyone would pay to generate
Chapter 8 established that the offer stack is sorted by marginal cost and that the last accepted offer sets the price. Nothing in that machinery forbids an offer below zero. It just requires a generator for whom producing at a negative price is better than not producing. There are three such generators, and they are genuinely different from each other.
Table 11-1: Three reasons an offer goes below zero
| Reason | Who | The calculation |
|---|---|---|
| Subsidy | Wind, some solar | Paid per megawatt-hour generated, so stopping forfeits it. Worth losing money on the energy to keep the credit. |
| Inflexibility | Nuclear, coal, combined heat and power | Shutting down and restarting costs more than running through a few cheap hours at a loss. |
| Obligation | Plants running for heat, steam or a process | Electricity is the byproduct. The plant runs for something else and the power has to go somewhere. |
The subsidy case is the one that changed the character of these markets, and the United States has the cleanest example. The federal production tax credit pays wind generators per megawatt-hour produced. For 2025 it is 5.50 dollars per megawatt-hour at the base rate and 28.00 dollars for a project meeting the wage and apprenticeship conditions, which most pursue. Because it is a tax credit rather than a cash payment, its pre-tax equivalent is larger still, which is why a wind farm can rationally offer at roughly minus 30 dollars and be better off running than idle.
Read that against the merit order and the consequence is immediate. The bottom of the stack no longer starts at zero. It starts below zero, and it starts there for a quantity of megawatts that grows every year. In any hour when demand is low and wind and sun are strong, the operator never needs to reach the part of the stack that costs money, and the marginal unit is a subsidised generator offering a negative number.
This is the corollary from Chapter 8 arriving in full. Renewables mostly do not set the price, because they are inframarginal. The exception is precisely these hours, and it is no coincidence that these are the hours when the price is lowest. Their price-setting share is real, and it is concentrated at the bottom.
The third reason is a wire, not a decision
Transmission deserves separating out, because it produces negative prices without anyone choosing to sell below cost. Chapter 10 established that price is locational. If a large amount of cheap generation sits behind a transmission constraint, the surplus physically cannot reach the demand on the other side, so the price collapses on the generation side while remaining perfectly ordinary a few hundred miles away.
West Texas was the standard illustration for years. Wind capacity was built where the wind is, which is a long way from Dallas and Houston, and for a period the wires were not there to move it. Prices at the wind end went deeply negative on windy nights while the cities paid ordinary prices, and the fix was several billion dollars of transmission rather than any change to the market. This is the same phenomenon as a landlocked crude trading at a discount to the coast, with one difference that matters: an oil producer facing a bad differential can store the barrels and wait. A wind farm cannot.
What it does to everyone involved
The first consequence is that a renewable project cannot be valued from its output alone. If a solar farm generates most heavily in the hours when solar has pushed the price toward or below zero, the average price it actually receives is well below the average price on the system. That gap is the capture rate, or value factor, and it falls as more of the same technology is built nearby. This is cannibalisation, and it is why two solar projects with identical output can have very different revenues.
The second is curtailment. At some point the cheaper answer is to stop generating, either because the operator instructs it or because the generator would rather not pay to run. Curtailed energy is genuinely wasted, and rising curtailment alongside rising negative hours is the clearest signal that a system has more of something than it can use at the times it appears.
The third is that negative prices are the business case for storage. A battery paid to charge, which later sells into an expensive evening, earns at both ends of the same day. Every hour below zero is an advertisement for the equipment in Chapter 5, which is why storage buildout follows solar buildout with a lag rather than accompanying it.
Not a malfunction
Negative prices are regularly reported as evidence that something has broken. They are not. A price below zero is the market correctly reporting that at this moment, in this place, electricity is a nuisance rather than a good, and that somebody should be paid to absorb it. That is information, and it gets acted on: batteries charge, flexible industrial load switches on, exporters push power across an interconnector.
What is genuinely a design problem is narrower, and it is the subsidy structure rather than the negative price. Paying a generator per unit produced, regardless of whether the system wants the unit, guarantees that some generation will be produced when it has negative value. Newer support schemes are built around exactly this, which is one reason contracts for difference that suspend the strike price during negative hours have replaced flat per-unit payments in a number of countries.
And this is only one of the two tails. In these hours the marginal unit is something with no fuel cost and a subsidy. In the opposite hours, a handful of times a year, the marginal unit is a machine that exists purely to run on the worst day of the decade, and the question becomes whether anything in the design pays for it to be standing there at all. That is Chapter 12.