Curtailment stopped being an operational footnote a while ago. In 2020 Chile's national grid failed to inject 247 GWh of solar and wind energy, 1.8% of what those plants could have produced. In 2025 it was 6,044 GWh and 15.6%: one in every six available renewable megawatt-hours never reached the network.
Against that backdrop, 2026 looks like the good news everyone had been waiting for. More than three thousand megawatts of batteries came online, the January-to-July curtailment rate fell from 13.0% to 11.7%, and June recorded a 2.5%, the lowest month since mid-2022. The easy reading is that storage has started to solve the problem.
The hourly data says something different, and more interesting.
First: six years and a factor of twenty-four
The monthly series is unambiguous. Between 2020 and 2025 curtailment grew twenty-four fold and the rate went from 1.8% to 15.6%. The worst month in the whole series is November 2024, at 26.6%: more than a quarter of that month's available solar and wind energy went uninjected.
What flattened in 2025 was not the volume but the rate: 6,044 GWh against 5,632, yet 15.6% against 15.8%. The system was spilling more energy in absolute terms and losing the same share of it, because the fleet was growing as fast as the problem.
That chart is worth a second look. The two upper curves —2025 and 2026— are the same curve. With 3,300 MW more batteries in the system.
Second: curtailment is not a system-wide problem, it is a nine-hour one
Averaged over the year, curtailment looks diffuse. Hour by hour it is not: 99% of it happens between 8 a.m. and 7 p.m., and the peak always sits in the same afternoon band.
In 2021 the peak was 59 MW at three in the afternoon. In 2026 it is 1,459 MW at four, twenty-five times as much. And this year's change of shape is plain: the midday gives way while the morning and late-afternoon shoulders rise. The curve got wider instead of lower.
Third: the battery arrived, and arrived fast
In January 2023 the grid had 84 MW of storage in operation across four installations, injecting 0.4 GWh a month. By August 2026 it is 60 plants, 6,295 MW and 554 GWh a month. Not many technologies have scaled like that in three and a half years.
And they do exactly what is asked of them. They charge at midday and discharge into the evening peak: between 9 a.m. and 4 p.m. the fleet absorbs energy at more than a thousand megawatts on average, and between 6 p.m. and 11 p.m. it gives it back.
Fourth: and yet
Here is the result that gives this column its title. Between January and July 2026 the grid's batteries charged 2,610 GWh. Over the same window, solar and wind curtailment was 2,677 GWh against 2,638 the year before: it rose 1.5%.
The explanation is in the denominator. January-to-July solar and wind generation grew 2,538 GWh over 2025. The battery absorbed 2,610. Within rounding error, those are the same number.
The breakdown by technology settles the reading. Solar curtailment fell 5%, and in the central hours of the day —eleven to two— it fell 14%: that is the battery working. But wind curtailment rose 24%, and it rose across the whole day.
These are two different problems sharing one name. Solar curtailment is an energy surplus inside a time band, and a battery moves it to another hour. Chilean wind curtailment is, to a large extent, a network problem: it blows where the line cannot carry it, at hours when the battery is already full or already discharging. Storing energy does not fix a transmission limit.
Fifth: where
Eight of every ten GWh curtailed in 2025 were lost in two regions. Antofagasta and Atacama account for 79%, and the concentration has grown: in 2022 it was 70%.
And within those regions, four busbars —Cumbre, Crucero, Cardones and Paposo— account for four of every ten GWh the system curtailed in 2026, this time counting every technology. That is a short list of places where the problem is identifiable, measurable and, above all, located.
Sixth: zero prices, the same problem seen head on
Curtailment is what the system declines to take. A zero marginal cost is what the energy is worth when the system does take it but does not need it. They are the same condition seen from two sides, and the second is measured without ambiguity: how many hours of the month the marginal cost came out at zero.
In 2019 this did not exist: a fraction of a percent of the time. In 2025 Crucero spent 36% of the hours of the year at zero, with months at 46%. What 2026 adds is divergence between zones: against 2024, the central zone fell from 31% to 22% and Puerto Montt from 23% to 11%, while Crucero barely gave way, from 35% to 32% in the first half.
The north does not give way. It is the same map as the previous chart, stated in prices.
The other half of the problem: what that energy is worth
There is a consequence that shows up in no curtailment statistic, because it is not energy lost but energy delivered and unpaid. The spot revenue per MWh captured by the grid's solar fleet fell to 1.6 USD/MWh in November 2024, and averaged 17.8 across 2025 against a marginal cost that year of around 60 USD/MWh at the reference busbars.
A solar plant with no contract and no battery is selling, in the bad months, at less than a tenth of the system's average price. Curtailment takes away its volume; zero prices take away the value of the volume it does deliver. The second loss is much larger than the first and is almost never discussed.
What to watch from here
There are 5,148 MW of batteries under construction according to the CNE registry, with 22,034 MWh of storage —about 4.3 hours of average duration— and 4,693 of those megawatts estimated to connect in 2026 and 2027. That is close to doubling the current fleet. The question is not whether it will arrive, but what can be expected of it.
- Storage buys time; it does not solve growth. As long as variable generation keeps growing at the rate it is growing, each block of batteries is consumed absorbing what is new. For curtailment to fall in absolute terms, the battery has to arrive faster than the panel, and today it is not.
- Half the problem is network, not energy. Wind curtailment rose 24% in a year when solar fell. No amount of storage fixes a transmission limit if the battery is not on the right side of the line.
- Location matters more than the total. Four busbars account for four of every ten GWh curtailed. A megawatt of battery at Cumbre or Paposo is not worth the same as a megawatt anywhere else, and the economic signal projects receive today does not tell those two apart well.
- The indicator to follow is the rate, not the volume. Volume rises because the fleet rises. The rate over available energy is the only series that says whether the system is improving or merely growing.