Solar and Storage for 24/7 Loads:
The Capacity-Factor Problem

Every solar proposal leads with the array's nameplate rating, and no continuous load has ever been served by a nameplate. Loads are served hour by hour, through December storms and fourteen-hour winter nights, and the number that measures the distance between those two facts is the subject of this paper.

Section 01One number against 8,760 hours

Capacity factor is the plainest metric in the power business: the energy a plant actually delivers over a period, divided by what it would have delivered running flat out at its nameplate rating the entire time. For the United States utility-scale solar photovoltaic fleet, the federal Energy Information Administration reports an annual capacity factor of 24.4 percent for 2025, following 23.2 percent in each of the two prior years.¹ For comparison, the wind fleet ran at 23.6 percent in 2025 and the nuclear fleet at 91 percent.¹

That number is not a criticism of the technology, and owners should be suspicious of anyone who wields it as one. Capacity factor is not an efficiency defect to be engineered away. It is day length, sun angle, weather, and geometry. A well-sited tracking array in California's interior valleys runs above the national fleet average; a flat coastal rooftop runs below it; every one of them, without exception, produces nothing at two in the morning. The fleet average is simply the honest center of gravity, and the seasonal shape behind it applies to the best sites as much as the worst.

The first consequence is pure arithmetic. A year contains 8,760 hours, and a round-the-clock load stands in all of them. Each megawatt of continuous demand consumes 8,760 megawatt-hours a year. At the 2025 fleet capacity factor, each megawatt of solar nameplate delivers roughly 2,100. Annual energy parity therefore requires an array of roughly 4.1 times the load's average draw, as an illustrative figure derived from the fleet rate, before a single kilowatt-hour has been stored, shifted, or lost in the shifting.¹ And that is the friendliest number in this paper. Matching annual energy is the easy half of the problem. The hard half is time.

Section 02The annual average is a summer number

Annual capacity factor flatters the system because it lets July vouch for December. The monthly figures take that cover away. In 2025 the utility-scale solar fleet's monthly capacity factor ranged from 32.4 percent in July and 32.3 percent in June down to 16.4 percent in January and 13.7 percent in December.¹ December production ran at barely more than half the annual average rate; midsummer ran a third above it.

A system sized on the annual average is therefore not a year-round system. It is a system in structural surplus from roughly April through September and in structural deficit from roughly November through February, and the deficit lands precisely when nights are longest and the load's dark hours are most numerous. The energy is not merely mistimed within the day. It is mistimed across the calendar, and no amount of midday abundance in June is deliverable to a January night without a storage medium that holds energy for months.

Then comes the winter week, which is where continuous-load designs actually fail. A Pacific storm train can suppress solar output for several consecutive days; an array still produces under overcast, but at a fraction of its clear-sky rate, on top of a December baseline that is already the year's lowest. Through those days the reservoir drains. When the sky clears, the array faces two jobs at once: carry the live load and refill the reservoir, at December production rates, through nine-and-a-half-hour days. Deficits compound faster than winter sun can repay them. This is why serious design for continuous load starts from the worst documented week, not the average day, and why a proposal priced on annual averages has quietly assumed away the exact hours the owner is buying.

Section 03The oversize ladder, written down

What follows is the arithmetic most solar-plus-storage proposals leave implicit. All of it is illustrative and normalized to a single megawatt of flat, continuous demand; every figure scales linearly, and the anchor rates are sourced. The point is not precision to the decimal. The point is the shape of the ladder.

Two rungs deserve a closer look. The land figure comes from the National Renewable Energy Laboratory's survey of operating United States plants, which found a capacity-weighted average total footprint of roughly 8.9 acres per megawatt of alternating-current capacity across surveyed solar projects, and about 3.5 acres per gigawatt-hour of annual generation.² That dataset is from the early 2010s and module efficiency has improved since, so a modern plant does somewhat better; it does not do an order of magnitude better. Tens of acres per megawatt of continuous demand is the honest planning range, and rooftops concede the point faster: loads that are dense per square foot of building, such as compute or refrigeration, draw far more energy under their roof than the same roof can harvest above it.

The second rung worth pausing on is the curtailment paradox. An array sized toward December is drowning in June: at winter-parity sizing, midsummer production runs several times the load, and the surplus must be spilled, stored into a reservoir that is already full by mid-morning, or exported at whatever compensation the interconnection and tariff allow, which varies by utility and program and is a study question in its own right. Oversizing buys winter adequacy with summer waste. Each additional increment of self-sufficiency costs more than the one before it, which is why the last rungs of this ladder are where stand-alone economics quietly fail even when the first rungs looked fine.

Section 04What storage shifts, and what it cannot

The battery fleet is the fastest-growing asset class on the American grid: cumulative utility-scale battery capacity passed 26 gigawatts in 2024, with 10.4 gigawatts added in that year alone, a 66 percent increase.³ In the California balancing authority, active battery capacity reached roughly 11,100 megawatts by mid-2024, and the market monitor's report on that fleet records a simple, telling fact: most of those batteries are four-hour machines.⁴

Four hours is not an accident and not a shortcoming. It is the shape of the job the fleet was built to do. The observed daily pattern is charge through the midday solar peak, discharge into the evening ramp.⁴ A battery earns its capital cost by turning over, and a daily cycle gives it several hundred turns a year over which to spread that cost. This is the economics that makes the evening fleet work, and the same economics is what closes the door on seasonal duty: a reservoir held against January cycles a handful of times a year, so its capital cost per delivered kilowatt-hour multiplies by roughly the ratio of the cycle counts. The market has built an evening fleet and not a winter one because the arithmetic of cycling permits the first and punishes the second.

Set the standard product against the ladder in Section 03. A four-hour battery holds four megawatt-hours per megawatt of load. The nightly winter reservoir is twelve to sixteen. The storm reservoir is forty-eight to ninety-six. None of this is a defect in the machine; it is a mismatch between an instrument priced for hours and a problem measured in days. Batteries move energy across hours superbly. Nothing sold at grid scale today moves it economically across seasons, and an analysis that sizes storage in hours while the deficit accumulates in days has changed the subject rather than solved the problem.

A four-hour battery is an evening instrument; a December night is fourteen hours long, and the storm behind it lasts three days.

Section 05Where solar and storage decide the question

Everything above is one half of the honest treatment. Here is the other half, stated with equal conviction: the same physics that disqualifies stand-alone solar-plus-storage for flat loads makes it the outright winner for a large class of sites, and an advisor who cannot say so plainly is not neutral either.

Loads that rise and fall with the sun. Single-shift manufacturing, school and office campuses that empty at night, water systems that can pump to tanks while the sun is up, irrigation districts, cold-chain sites with room to pre-cool: for these, the capacity-factor problem in this paper largely does not arise, because the hours the load needs are the hours the array produces. Coincidence, not storage, does the firming. The array is sized near the daytime draw rather than at a multiple of it, the battery shrinks to shoulders and short gaps, and the economics can be decisive. The residual honesty: winter output still halves, so the grid or another source still carries dark hours and dim weeks, and the design should say so rather than round it away.

Expensive evening windows. Where time-differentiated rates concentrate cost into a late-afternoon and evening peak, a four-hour battery aimed at a four-to-six-hour window is the right tool used for the exact job it is priced for: charge from the midday array, discharge across the window, repeat daily. This is precisely the duty cycle the California fleet demonstrably runs.⁴ The qualification is the same one the cycling math imposes everywhere: the value lives inside the window, and the machine does not extend to overnight coverage merely because the brochure and the tariff both mention evenings.

The energy layer inside a hybrid. For genuinely continuous loads, the array's best role is usually not sole supplier but fuel displacer alongside dispatchable generation, whether that generation is reciprocating engines, turbines, fuel cells, linear generators, or the grid connection itself. Every daytime megawatt-hour the array produces is fuel the firm asset does not burn or energy not purchased at retail; the battery firms the shoulders and the ramps; the firm resource carries night and winter and is sized to the load rather than to the weather. In that architecture the array is sized to the daytime valley and to what the site can physically host, the multiples of Section 03 are never invoked, and the solar economics are judged on displaced energy cost, where they are strongest.

The merits that need no exaggeration. The array burns nothing, buys no fuel, and carries no fuel-price exposure on the energy it produces. Photovoltaic equipment itself raises none of the combustion permitting questions that engines and turbines face in California's stricter air districts, though battery systems carry their own fire-safety and siting review, which is real work and belongs in the schedule. Procurement is modular and comparatively fast. Under current federal law, the investment tax credit for qualifying clean-energy property is 30 percent; statutory adders exist but must be individually qualified, never assumed, and the position belongs with qualified tax counsel.⁶

Section 06Where the structure gives out

For a flat or winter-peaking continuous load asked to stand alone on solar and batteries, the ladder is the argument, and it is worth restating as three structural facts. Energy parity alone demands an array at roughly four times the load. Winter self-sufficiency demands roughly seven times, or a cross-season storage product that does not exist at market economics. And the reservoir that survives a storm week is measured in days of load while the standard product in the field holds four hours.^1,4

The system-level literature reaches the same conclusion from the other direction. A peer-reviewed study in the journal Joule modeled nearly a thousand deep-decarbonization cases and found that the availability of firm resources, meaning generation whose output does not depend on the weather, lowered modeled electricity costs by 10 to 62 percent in fully decarbonized systems relative to portfolios of variable renewables and storage alone.⁵ The mechanism in those models is the mechanism in this paper: the final hours are the expensive hours, and serving them with a weather-driven resource means buying multiples of everything to cover the worst week.

There is also an asymmetry of failure modes that deserves plain language. When a fuel-supplied plant faces a bad week, the remedies are fuel logistics, redundancy, and maintenance planning, and their costs scale roughly with hours served. When a weather-driven system faces a bad week, the principal remedy is more of the weather-driven system, and the cost scales with the depth of the worst deficit. One curve rises gently with duty; the other steepens toward the tail. An owner buying reliability for 8,760 hours should see both curves priced before choosing which one to stand on.

None of this hands the argument to the alternatives, and the neutral treatment requires saying so in the same breath. Engines and turbines are combustion equipment facing demanding control-technology requirements and, in some districts, offset obligations; they carry fuel-price exposure, noise, and scheduled outages. Fuel cells avoid combustion permitting burdens but carry higher capital cost and stack-replacement economics. Full utility service, where a documented energization date actually meets the need, remains the benchmark every self-supply case must beat. The claim of this paper is narrow and structural: for round-the-clock loads, solar plus batteries alone rides a cost curve that steepens toward the worst week, and an honest study prices that curve instead of averaging it away.

Section 07The portfolio answer

The productive question is never whether solar belongs on the site. It is which of the site's 8,760 hours the sun should serve, which hours the battery should move, and which hours need a resource whose output does not depend on the sky. Load shape decides, and the honest fit reads like this:

Inside a portfolio, each asset then does the job its cost structure favors. The array takes the daytime share the site can host, judged on displaced energy cost. The battery takes the window it can cycle daily, judged on turns. Firm capacity takes the night and the winter, judged on the cost of hours served. The grid takes whatever the tariff and a documented timeline genuinely deliver. Owners who let one asset audition for another's job, in either direction, pay for the miscasting every year the system runs.

The discipline that gets a specific site to that answer is short:

1. Start from measured interval data, never from annual totals; the load's shape decides more than any technology choice does.
2. Decompose by season and construct the worst documented week; that week, not the average day, sizes anything asked to stand alone.
3. Audit land and roof against the multiples honestly, at surveyed densities rather than brochure densities.²
4. Price export and curtailment reality under the actual tariff and interconnection terms, not nameplate output.
5. Set the reliability target in writing, price the tail explicitly, and then compare every credible path, including full utility service and no project, on the same dated basis.

The sun publishes its schedule in advance, every day, for free, which makes solar the most candid resource on any site. The discipline owed in return is to size the system to the load's actual hours rather than to the array's best one. Where the hours line up, solar and storage win on the merits and should be chosen without apology. Where they do not, the honest answer is a portfolio, and the owner deserves to see the arithmetic, on one page, before capital moves.