Two numbers published by the Energy Information Administration in February frame the US power sector in 2024. Developers and plant owners plan to add 62.8 gigawatts of utility-scale capacity this year, 55% more than the 40.4 GW added in 2023, which was itself the largest addition since 2003. Operators plan to retire only 5.2 GW, a 62% fall from the 13.5 GW retired last year and the smallest figure in any year since 2008.
On its face, that is a picture of abundance. The country is building far more than it is closing. Combined with a year of relatively low wholesale prices in 2023, it would be easy to conclude that the reliability concerns that dominated the last two winters are fading. We think that conclusion would be premature. The 2024 numbers describe a system adding energy quickly, but the capacity that is leaving and the capacity that is arriving do different jobs.
What is being added
Solar dominates. The EIA expects a record 36.4 GW of utility-scale solar in 2024 if developers meet their schedules, nearly double the 18.4 GW added in 2023, which was also a record. Battery storage is the second-largest category, and together solar and batteries account for 81% of planned additions. The agency attributes the acceleration partly to the easing of supply chain problems and trade restrictions that slowed solar projects in 2022 and early 2023.
Solar's share of additions is not a forecast of solar's share of firm capacity. A gigawatt of solar contributes a full gigawatt at midday on a clear summer day and almost nothing after sunset. Batteries contribute for a few hours. The question for planners is how much of the 62.8 GW will be available in the hours when the system is most stressed.
What is leaving, and what is not
The slowdown in retirements is concentrated in coal. After 22.3 GW of coal capacity retired over the past two years, only 2.3 GW is scheduled to close in 2024, about 1.3% of the coal fleet operating at the end of 2023. Coal and gas together account for 91% of planned retirements this year.
The pause is temporary. Operators expect coal retirements to rise again to 10.9 GW in 2025. That bunching is important. A quiet year followed by a heavy one gives planners a window to bring replacement capacity on line, but it also means that 2025 will test whether that replacement has actually been built.
There are several reasons retirements might be slow in a given year. Some reflect plant economics, some reflect state policy timetables, and some reflect requests from grid operators to keep units available while transmission upgrades are completed. The EIA data record the plans as reported, not the motives. But the pattern is consistent with a system in which operators are cautious about closing dispatchable units until they can see what replaces them.
Low wholesale prices in 2023 had specific causes
Wholesale electricity prices at most major hubs fell in 2023 and traded in narrower ranges than in the volatile 2022. The EIA attributes this mainly to lower natural gas prices, a mild start to the year and reduced average loads in many regional markets. Prices at ISO New England, PJM and MISO hubs were lower in every month of 2023 than in the same month of 2022. The exception was the Northwest Mid-Columbia hub, where continuing drought cut hydropower and pushed prices up in seven months of the year.
That Northwest exception is instructive. In a region where the dominant resource is weather-dependent, a bad water year raises prices even when gas is cheap. As solar and wind become dominant in other regions, a similar exposure to weather will apply there, with wind droughts and cloudy winter weeks playing the role that low snowpack plays in the Pacific Northwest.
The low prices of 2023 were driven mainly by fuel costs, not by a structural surplus of firm capacity. If gas prices rise as LNG exports grow, the price relief will narrow regardless of how much solar is built.
The mismatch to watch
The combination of record additions and minimal retirements creates a specific risk: a system that is long on midday energy and not obviously long on winter evening capacity.
In summer, solar and batteries match the demand profile well. Peaks occur in the late afternoon, solar is still generating, and batteries can cover the last hours of the evening ramp. In winter, the peak in many regions comes in the early morning and the evening, in darkness, and can last for days of cold weather. Short-duration batteries are less useful across a multiday event, and solar contributes little.
The retiring fleet, mostly coal and older gas, was built to run through exactly those conditions. Retiring it while building solar and four-hour storage is not wrong in principle, but it requires an honest accounting of winter capability. Several grid operators are moving to accreditation methods that rate each resource according to its contribution in the riskiest hours. That shift will lower the capacity credit of solar and, eventually, of batteries, as each additional unit contributes less to reducing shortfall risk.
A window, not a cushion
We would describe 2024 as a window. Retirements are low, additions are high, and the price environment is benign. That creates room to do the work that the 2025 retirement wave will require: transmission upgrades, completion of gas pipeline and storage projects where they are needed, and procurement of longer-duration resources.
The risk is that a good year is read as evidence that the problem has gone away. The EIA's own figures show the retirement pause ending in 2025, and the additions in 2024 are overwhelmingly in technologies whose winter contribution is limited. A record build year is a necessary condition for a reliable transition. It is not a sufficient one.
What would change our view
If a meaningful share of the 2024 additions turned out to be longer-duration storage or firm resources, the mismatch would narrow. If grid operators' winter reliability assessments for 2024 to 2025 showed comfortable reserve margins after accounting for the new accreditation methods, the concern would be smaller. And if gas prices stayed low into 2025, the price relief of 2023 would persist even as the fleet changes. Each of these is possible. None is yet visible in the data.