Does a home battery pay for itself?
A battery looks like the obvious next step after solar. Two years of real bills said otherwise, and how they said it is the interesting part.
Updated October 2026. The first version of this page used a generic payback model with made-up numbers. This version follows what my real bills showed. The chart and the example tables still use made-up numbers shaped like mine; the rates and usage described in the text are rounded real figures.
The question
My utility is PG&E, which serves most of Northern and Central California. I have solar on the house, installed under NEM 2.0, California's net-metering plan for systems connected before April 15, 2023. Under NEM 2.0, every kilowatt-hour my panels send to the grid is credited at the same retail rate I pay to buy one, and customers can stay on the plan for 20 years from the day the system was connected. That is what makes me grandfathered: the plan is closed to new customers, who get a net billing tariff that usually credits exports at a lower rate. The California Public Utilities Commission explains both on that page.
A battery looked like the natural next step, and every online calculator said it would pay for itself. I wanted to know what it would do on this house, using the bills instead of a generic model.
What power costs and how much I use
Electricity in California is expensive, and a time-of-use plan makes the price depend on the hour. On my PG&E plan in 2026, a kilowatt-hour costs about 31 to 35 cents in winter, depending on the hour. In summer it runs from about 33 cents overnight and through midday up to about 55 cents in the 4 to 9 pm peak. These are rounded, and they change a few times a year. For scale, the U.S. Energy Information Administration puts the average California household price at about 32 cents in 2024, against about 16 cents for the U.S. (EIA, Table 5.a).
The same table shows what a normal home uses: about 500 kWh a month from the utility in California, roughly 6,000 kWh a year, and about 860 kWh a month, or 10,400 a year, across the U.S. My house bought about 11,000 kWh from the grid in the last full year, close to double the California average, and that is before counting what the solar panels supplied directly or the roughly 5,700 kWh they sent back.
Two things explain it. We have two EVs, and we charge them at home. And the house has a heat pump that does both the heating and the cooling, so most of what the house needs, apart from some gas, runs on the electric meter. The cars charge overnight, so about three quarters of my winter grid purchases land in the cheapest, off-peak window. The heat pump works hardest in winter, which is also when my west-facing array makes the least energy. Those big loads sit in the cheapest hours, and that shapes everything below.
Getting the data out of the bills
Most of the statements existed only as scanned, image-only PDFs. I had Claude read them with the text recognition built into macOS at 200 dpi and parse each one into rows: net kWh by time-of-use period, the rates, the charges. A model built on misread digits is worse than no model, so before any analysis the data had to agree with itself:
- The time-of-use kWh added up to the printed net usage on every statement except one, and that one was a single misread digit outside the year I analysed.
- The twelve monthly charges added up to the year-end true-up statement to within a few cents.
- Energy charges plus fees matched the statement's printed pre-tax total, and the taxes worked out to the printed number.
After that the numbers were the bills themselves, not a model of them.
What the bills said
The instinct is that a battery earns money on volume: store cheap energy, use it when it is expensive. The bills said the problem is price. The chart shows the shape with made-up numbers: bars are the net energy my house buys during the 4 to 9 pm peak window each month (below zero means it sells more than it buys), and the shaded months are when the price gap between peak and off-peak is wide.
Made-up numbers shaped like mine. Bars are net kWh during the peak window; shaded months have a wide peak-to-off-peak price gap.
The two seasons fail in different ways:
- Winter. The house buys energy at peak almost every day, but peak is about 35 cents against about 31 off-peak, a gap of only 4 cents or so. Losing about 10% of a 31 cent kilowatt-hour in the battery costs about 3 cents, so moving energy across that gap is roughly break-even before you count what the battery cost.
- Summer. The gap is large, about 55 cents at peak against about 33 off-peak, which is 22 cents. But the west-facing array is already selling energy during the evening peak, and NEM 2.0 credits those sales at the full retail rate. Storing that energy instead earns nothing extra.
- The one overlap is the early autumn shoulder, where the house still buys at peak while summer prices apply. It is worth a few dollars a year, and nobody sells a battery for two shoulder months.
Most "a battery pays for itself" numbers assume exports earn far less than retail. That is true under California's newer net billing tariff, which credits exports at a value that is usually lower than the retail price (the CPUC's comparison table lays out the difference), and that gap is where the battery savings come from. Under NEM 2.0 the gap is already closed.
The ceiling
I computed the theoretical maximum: take every kWh the house bought at peak and part-peak in the last full year, and pretend a perfect battery with unlimited capacity and no losses moved all of it to off-peak. The answer was tens of dollars a year. Fixed monthly charges are the largest part of the bill that no battery can touch.
| Made-up round numbers | Per year |
|---|---|
| Perfect battery, no losses | about $60 |
| Realistic, after about 10% loss | about $50 |
| Installed cost of a system | about $20,000 |
| Price at which it would break even over 15 years | about $1,000 |
One caveat matters. A paper bill only shows monthly totals, and a real battery works on minute-by-minute flows, so the true opportunity is higher than this calculation. But even five times higher is a few hundred dollars a year against a five-figure price. The 15-minute interval data the utility offers would settle it, and it is the one open item that could change the answer.
My own estimates kept getting corrected
I kept the analysis in an Obsidian note and let the numbers overwrite my earlier guesses, with each correction written down. The first estimate fell by more than an order of magnitude over two revisions.
| Version | Based on | Estimate (made-up scale) |
|---|---|---|
| 1 | An assumed price gap | about $900 a year |
| 2 | The nine statements that were readable | about $200 a year |
| 3 | Every statement, computed per rate period | about $60 a year |
Two other numbers moved the same way. Annual net energy import was 42% worse than I had extrapolated from nine statements. And a projection of the year-end bill was off by roughly $900 because I had stopped the series in July: the late-summer months I left out are reliably expensive. An estimate extrapolated from part of a cycle looks best in exactly the months it leaves out.
What beats a battery
Once the battery was ruled out as a savings project, the same bills pointed to cheaper wins, in order of return:
- Check the rate plan. The utility's own twelve-month comparison, printed on a statement and built from my real usage, showed another plan about three times the battery's entire theoretical ceiling cheaper. Costs a web form.
- Find the load that changed. I know where the big loads are, the two cars and the heat pump, so the question is what changed. One winter month used about 3.6 times the net energy of the same month a year earlier, at a cost several times the battery's whole annual ceiling. A similar burst showed up for about two weeks in summer. Candidates are how the cars were charged, or the heat pump working harder than usual in a cold snap or a heat wave, and the interval data can tell them apart.
- Add a little solar. About 1 kW more solar offsets energy at the full retail rate, with no efficiency loss, for roughly a six-year payback. That is far better return per dollar than a battery that never pays back. Check first that the addition is small enough not to end the grandfathered plan.
The electrical limits come first
Before the money question there is a hardware question: can the house take a battery at all? Three things came up.
Backfeed headroom. A main panel has a busbar with a current rating, and the breakers that can feed power into it are limited by a rule of thumb often called the 120% rule. On a 200 A busbar with a 200 A main breaker, the breakers feeding the bus can total up to 240 A, leaving 40 A for solar and storage together. If existing solar already uses 30 A, a battery that wants a 50 A breaker does not fit. Typical fixes are a meter-collar connection, a smaller main breaker or a bigger panel, and each is a decision for an electrician and the utility.
Microinverter compatibility. Solar with microinverters shuts down when the grid does, unless something provides a reference signal. An AC-coupled battery can do that and run the house in an outage, but only with the microinverter models its maker has tested, and some batteries have a minimum size relative to the array. Check the compatibility list before assuming it will work with what is on the roof.
Keeping the grandfathered plan. Some batteries want to replace the solar inverter. That can count as a new interconnection instead of a modification, and a new one can mean losing NEM 2.0, which is worth far more than the battery's efficiency gain. The question for any installer is whether it will be filed as a modification, in writing. While I was in the paperwork I also found that the documents for my own system disagreed with each other about the panel count, which is worth resolving before filing anything.
Where it ended up
On this house a battery is not a bill-savings investment, and two years of bills make that a number rather than an opinion. It would make sense only if outage ride-through is worth the price on its own. The to-do list is free instead: download the interval data, move to the better rate plan, and find the load that doubled.
I'm not an electrician. The breaker math above is a rule-of-thumb illustration of why panels run out of headroom, not advice for any particular panel. An electrician and your utility decide what is allowed.
Sources: California Public Utilities Commission, Net Energy Metering and Net Billing; U.S. Energy Information Administration, 2024 average monthly bill by state, Table 5.a. My own rates and usage come from my PG&E statements.
This is entry 13 of AI Lab Work.