Solar curtailment is electricity a solar plant could have produced but didn't, because the grid told it to back off. By our count of its own data, California's grid operator curtailed about 3.5 million MWh of solar in 2025. That's roughly one in every 17 MWh its solar fleet could have delivered, and in March it was almost 18%.
Those megawatt-hours never reached anyone, so they displaced no other generation and avoided no emissions. A meter on the plant's output never counts them. A figure worked out as capacity times capacity factor usually does.
California's grid operator publishes its curtailment data in five-minute intervals, and we added it up for every year from 2019 to 2025. A worked example further down shows how much it can move a reported total.
What solar curtailment means
A grid has to match supply and demand every few seconds. When there's more solar available than the grid can use or carry, the grid operator reduces how much some plants put out. The panels still see the sun; the plant sends less power.
According to the EIA, wind and solar output is reduced "either through price signals or, rarely, through an order to reduce output," during two kinds of periods. Congestion is "when power lines don't have enough capacity to deliver available energy." Oversupply is "when generation exceeds customer electricity demand."
The California Independent System Operator (CAISO) defines curtailment as "the difference between actual production and the forecast when actual production is less than the forecast." So even the official curtailed figure is an estimate. It depends on what the forecast said the plant could have made.
Economic, congestion and reliability curtailment
CAISO's daily curtailment report sorts every curtailed megawatt-hour two ways: why it happened (local congestion or system-wide oversupply) and how (economic bids, self-schedules or exceptional dispatch).
Most curtailment is economic. The plant has bid to reduce its output at a set price, the market price falls that far, and the market dispatches it down. Exceptional dispatch is the operator stepping in directly. In CAISO, 2,740,386 of the 2,741,671 MWh of wind and solar curtailed in 2025 through May 31 were economic, and the report lists no exceptional dispatch at all.
Most of it was local. Through May, about 80% of that curtailment was to relieve congestion on particular lines, with the other 20% for oversupply across the whole system. The EIA dates the rise to 2019: "Congestion-related curtailments have increased significantly since 2019 because solar generation has been outpacing upgrades in transmission capacity."
Reliability rules play a part too. The EIA notes that CAISO "also curtails solar generation to leave room for natural gas generation," because some gas has to stay online to meet NERC reliability standards and to be ready for the evening ramp, when the sun sets and demand climbs.
How negative prices lead to solar curtailment
When midday solar floods a market, wholesale prices can fall to zero and below. The EIA's explanation is that negative prices "usually result when generators with high shut-down or restart costs must compete with other generators to avoid operating below equipment minimum ratings or shutting down completely."
A gas plant that can't easily switch off will pay to keep running. That's when solar plants with economic bids get dispatched down: the price has fallen below what they offered.
Lawrence Berkeley National Laboratory (LBNL) ties the two together in Texas: the "export-constrained Panhandle and other renewable-rich areas experience low or negative local prices." It also found that more batteries in solar-heavy regions "eased negative price events" in 2025. As the EIA puts it, battery storage "allows some renewable energy to be stored and used 4-8 hours later in the day."
Solar curtailment in California, 2019 to 2025
CAISO publishes its production and curtailment data in five-minute intervals, one file per year. We added up the solar columns for each year. CAISO describes these files as raw data that are "not considered operational or settlement quality," so treat the totals as close, not exact.
| Year | Solar delivered (MWh) | Solar curtailed (MWh) | Share of available solar curtailed | Worst month |
|---|---|---|---|---|
| 2019 | 28,616,980 | 921,684 | 3.1% | May, 7.1% |
| 2020 | 29,990,483 | 1,497,220 | 4.8% | April, 10.9% |
| 2021 | 34,057,267 | 1,426,326 | 4.0% | March, 10.3% |
| 2022 | 37,079,956 | 2,320,258 | 5.9% | April, 14.4% |
| 2023 | 39,806,415 | 2,508,923 | 5.9% | March, 19.1% |
| 2024 | 50,882,403 | 3,192,600 | 5.9% | March, 16.2% |
| 2025 | 57,244,767 | 3,482,322 | 5.7% | March, 17.8% |
Source: our sums of CAISO's production and curtailment files, 2019 to 2025. "Share of available solar curtailed" is curtailed divided by delivered plus curtailed. Adding CAISO's wind curtailment, our 2022 and 2024 totals match the EIA's published figures.
Between 2019 and 2025, solar output doubled and the amount curtailed grew almost fourfold. The share has held near 6% since 2022 even as the fleet kept growing, but the spring months tell a different story. In March 2025, 857,177 MWh were curtailed against 3,953,559 MWh delivered. In July and August, the share was about 2%.
Spring is the reason. The EIA explains that CAISO curtails most "in the spring, when solar output was relatively high and electricity demand was relatively low," with mild weather meaning little heating or air conditioning. The EIA's 2025 analysis of California curtailment adds that wind and solar capacity in California grew from 9.7 GW in 2014 to 28.2 GW by the end of 2024.
Other grids, and why the rate isn't just about how much solar there is
Most US grid operators and regions "do not yet report curtailment," according to LBNL. Its 2026 utility-scale solar report gives 2025 rates for three of them, working out ERCOT's from plant-level data. In 2025, the curtailment rate was "highest in ERCOT (6.9%) compared with CAISO (3.8%) and SPP (2.8%), even though solar's penetration rate is far lower in ERCOT (13%) than in CAISO (33%)."
ERCOT, the grid operator for most of Texas, curtails a bigger share of its solar than CAISO does while getting far less of its power from solar. LBNL's explanation points at transmission: West Texas accounted for about half of ERCOT's modeled plant-level curtailed energy in 2025, and LBNL calls the Panhandle export-constrained.
LBNL's CAISO figure (3.8%) is lower than ours (5.7%) mainly because it divides by utility-scale and distributed solar combined, a much bigger denominator. Both are right on their own basis, which is why any curtailment percentage should say what it's divided by.
Texas still curtails the bigger share on our basis. LBNL's ERCOT rate has distributed solar in its denominator too, so ERCOT's grid-scale rate is at least 6.9%, against CAISO's 5.7%.
Modeled MWh versus delivered MWh
There are two common ways to arrive at a solar project's annual megawatt-hours, and only one of them already accounts for curtailment.
Metered output is what the plant sent to the grid. LBNL notes that net generation reported to the EIA excludes "energy that was curtailed (for economic or system stability reasons)." If a provider reports metered MWh, curtailment is already netted out.
Modeled output starts from capacity, multiplies by a capacity factor and by the 8,760 hours in a year. If the capacity factor comes from a solar resource model, it describes what the panels could produce in that sunshine. Unless someone adds a curtailment assumption, it knows nothing about the grid. Our article on how we calculate Dyme's renewable energy impact walks through that arithmetic step by step.
A capacity factor taken from a plant's own metered history is in between. Past curtailment is already in it, but it won't reflect curtailment that has grown since, and in California it has grown almost every year.
Worked example: one 10 MW project
The numbers in this example are illustrative, chosen to show the arithmetic. A 10 MW solar project has a modeled capacity factor of 25%, which gives 21,900 MWh a year. The 4% and 7% rates are round numbers near the low and high ends of the 2025 figures above: LBNL's 3.8% for CAISO and 6.9% for ERCOT, with our 5.7% for CAISO's grid-scale fleet in between. The grid factor of 0.40 metric tons of CO2 per MWh is also illustrative.
| Scenario | Delivered (MWh) | Overstated by modeled figure (MWh) | Avoided CO2 (metric tons) | Overstated CO2 (metric tons) |
|---|---|---|---|---|
| Modeled, no curtailment | 21,900 | 0 | 8,760 | 0 |
| 4% curtailed | 21,024 | 876 | 8,410 | 350 |
| 7% curtailed | 20,367 | 1,533 | 8,147 | 613 |
At a 7% rate, the modeled figure claims 7.5% more electricity than the project delivered, and the avoided-emissions line inherits the same error. A 30-year projection repeats that gap every year.
ISO-wide averages also hide the spread between plants. A single project's rate depends on where it connects, so a regional average tells you little about one plant.
Questions to ask about a provider's solar MWh
These are specific to curtailment. For the broader checks, see our guides to assessing community-solar impact claims and verifying a project is operating.
- Is the annual MWh figure metered or modeled? Our partner impact reporting schema includes a field for exactly this flag.
- If it's modeled, what curtailment did the model assume, and where did that number come from?
- Does the grid operator in that region publish curtailment? If it doesn't, ask where the provider's curtailment figure comes from.
- Can you see generation by month? In California, curtailment is concentrated in spring, and an annual total hides that.
- When meter data replaces a modeled figure, does the provider restate the old number and say what changed?
A provider that reports metered output can answer the first question in one line. One that reports modeled output should be able to answer the second as quickly.
What Dyme's results page reports
Our impact results page reports 5.5 MW of solar built and operating, with 20 MW more in the pipeline, and 7,807 MWh of clean energy produced per year. The page doesn't say whether that annual figure is metered or modeled.
If you're assessing us, ask. You can request project documents from us, including meter data, and compare them with the annual figure. Our impact figures aren't independently audited, so the documents are the way to check them.
FAQ: solar curtailment
What is solar curtailment?+
It's solar power a plant could have generated but didn't send to the grid, because the grid operator reduced its output. It usually happens when power lines are congested or when there's more supply than demand. In California it peaks at midday in spring.
What is curtailment in electricity?+
Any deliberate cut in a generator's output below what it could produce, made to keep supply and demand in balance or to stay within what the transmission lines can carry. The term is used mostly for wind and solar, because their potential output is forecast, so the shortfall can be measured.
Why is curtailment bad?+
Curtailed power is clean electricity that was available and went unused, so it displaced nothing. Some curtailment is normal on a grid with a lot of solar. Rising curtailment usually means transmission or storage hasn't kept up with new capacity.
Do I get money back for excess solar energy?+
For rooftop solar, that depends on your utility's export or net metering terms, which vary by utility and state. It's a separate question from grid-scale curtailment, which is about large plants being dispatched down by a grid operator.
Does metered solar generation include curtailed energy?+
No. Metered net generation counts only what the plant delivered to the grid, so curtailment is already netted out. A figure modeled from capacity and a capacity factor usually isn't, unless the model applies a curtailment assumption.



