How often does a home solar install include a battery?

In 2015, one residential solar permit in 800 mentioned a battery. In 2025 it was closer to one in three and a half. That is the fastest change in how American solar is sold that this permit database has recorded, and most of it happened in two years. The figures below come from 4,343,488 U.S. solar permits, derived August 30, 2026 — and the methodology is load-bearing, because this measurement has a specific way of going wrong.

The national curve

Each row is residential solar permits filed that year, and how many of them record energy storage as part of the job.

Three regimes are visible. Through 2016 storage is a rounding error. From 2017 to 2023 it grows steadily but slowly, from under 1% to about 8% — the Powerwall era, when a battery was a premium add-on bought by people who wanted backup power. Then the line breaks: 8.0% in 2023, 18.8% in 2024, 29.0% in 2025.

The break is a policy artifact with a date on it. California replaced retail net metering with NEM 3.0 on April 15, 2023, cutting the credit for exported solar by roughly three quarters. Under that tariff a solar system without a battery gives away most of what it makes at midday and buys it back at dusk, so storage stopped being a backup purchase and became part of the economics of the array. 2024 was the first full permit year under the new rule.

It is not only California

California is 40% of the residential solar permits in this database, so a Californian policy shift can drag a national average on its own. It is worth checking whether that is all this is. The table below restricts to permits whose descriptions are long enough to name equipment at all (60 characters or more — see methodology), and then shows the same series with California removed entirely.

California roughly doubles the national rate, and the answer survives its removal. The rest of the country ran 9.7% in 2023 and 22.5% in 2025 — a slower start and the same shape a year behind. What happened in California is now happening in the states that watched it.

By state, with the caveat that makes it readable

The state table is the most quoted thing on a page like this and the easiest to misread. It measures how often a state’s solar permits say storage, which is real adoption multiplied by how much detail that state’s permit offices write down. Read the top of it with confidence and the bottom of it with suspicion.

The top of the table is the cleanest external check this page has. Hawaii, California, Arizona and Nevada are the four states that ended or never had retail net metering — Hawaii closed it in 2015, Nevada fought over it from 2015, Arizona replaced it in 2016, California in 2023 — and three of the four are at the top. Hawaii at 85.1% is a genuine outlier that other sources also report: with the highest electricity prices in the country and no export credit for a decade, storage is close to standard there.

The bottom of the table is mostly about paperwork. New Jersey reads 0.6%, and its most common 2025 solar permit description is the single word “Solar” — a record that could not report a battery if one were bolted to the garage wall. New York, DC and Montana have the same problem in milder form. These states plausibly do have low attach rates, since all of them still offer some form of net metering, but this data cannot prove how low.

Nevada at 6.5% is the one figure in this table we would actively discourage quoting. Clark County is 86% of Nevada’s solar permits and writes them 37 characters long; Washoe County, an hour and a half of driving away, writes them at 76 characters and reads 21%. The state number is a description-length artifact of where Nevadans happen to live.

How much of this is measurement?

A fair question, and it has a measurable answer. Sorting 2025 solar permits by how long the description is:

27,255 solar permits filed in 2025 carry no description at all, and read 0.0% by construction. Below 30 characters the rate is 8.2%. It then rises and plateaus: 33.1% at 60–99 characters and 34.1% at 100–149. That plateau is the useful part — once a description is about sixty characters long, making it longer stops finding more batteries, which is the signature of a measurement that has saturated rather than one still climbing.

So the honest national range for 2025 is between 29.0% — every permit counted, silent ones counted as no — and 38.9% on the subset of permits detailed enough to say. The truth is inside that band and closer to the top of it. The 150-plus row overshoots because the most verbose jurisdictions are disproportionately Californian and Hawaiian, which is the same confounding running the other way.

Almost nobody adds a battery later

3,757,036 homes in this database hold a residential solar permit. 398,812 of them — 10.6% — also have a permit recording storage. That stock figure is far below the 2025 flow figure for an unremarkable reason: most solar on American roofs was installed before batteries were normal, and it is still up there.

What is remarkable is when the battery arrived. 372,635 of those 398,812 homes — 93.4% — permitted the battery in the same year as the solar. Only 20,420 homes in the entire country added storage to an existing array in a later year. Against 3.76 million solar homes, that is a retrofit rate of 0.54%.

Storage is sold with the system or it is not sold at all. Roughly 3.3 million American homes have a solar array and no battery, most of them installed under a net-metering regime that is now being withdrawn state by state, and fewer than one in a hundred and eighty has been sold storage since. Whether that is a market failure or a market waiting depends on who you ask, but the number is not close to ambiguous.

Methodology

• Cohort. Building permits on residential property that pass our photovoltaic-solar rule, which requires the permit’s own description to name solar, photovoltaic or PV. The raw permit classifier is not usable for this — it tags driveway replacements as solar installations. The cohort is 4,343,488 permits, of which 4,209,656 are residential.
• What counts as an attachment. The permit record states energy storage — a battery, an ESS, a Powerwall, backup storage. Home-level figures additionally count storage filed on a separate permit for the same property, which is why they run 2–3 points higher than the permit-level figures.
• Every figure is a floor. A permit that says nothing about storage is counted as having none. We cannot distinguish “no battery” from “battery not written down”, so all levels understate. Year-over-year changes are the robust part of this page, because a jurisdiction’s writing habits change far more slowly than its solar market.
• One boilerplate string is excluded. Clark County, Nevada files solar permits under a template reading “Photovoltaic system per kw (enter total kw as whole number; ess/battery installations require plans submitted on dashboard)”. The word “battery” there is an instruction to the applicant, not a description of the work. Permits whose text matches that instructional shape are excluded — 8,034 nationally, of which 7,953 are Clark County. Without the exclusion Nevada’s 2025 rate would read several times higher off a form field.
• The commercial forklift trap does not apply. On commercial and industrial permits, “battery” frequently means a forklift charger rather than storage. Of 395,467 residential solar-plus-battery permits, zero mention forklifts or pallet jacks. We checked because the error would have been invisible.
• The 60-character legibility rule. Where this page reports a “legible” subset it means permits with descriptions of at least 60 characters. That threshold was set from the length-versus-rate curve, where the measurement plateaus, and it is defined without reference to battery content — a long description qualifies whether or not it mentions storage.
• Permit coverage. Not every jurisdiction publishes permits and we do not ingest every one that does, so all counts are floors on true activity. 2026 is a part year and its rate is reported as such.
• What this is not. It is not a count of installed batteries, a market-share estimate, or a claim about any manufacturer or installer. It is how often the permit record for a solar job says the job included storage.

Derived August 30, 2026. Spot something that looks wrong? Tell us — we’d rather correct a number than defend it.

Companion pages: solar permit statistics covers the permit volumes underneath this cohort, what percentage of homes have solar counts arrays from rooftop imagery instead of paperwork, solar permits that never got installed checks the same permits against the roof, and how to get solar leads covers turning a cohort like the un-batteried 3.3 million into a route.

Frequently asked questions