Is solar worth it in PG&E territory? We modelled 56 real homes to find out
We ran the SIA calculator (officially the SIA Solar Design Studio) on 56 randomly selected houses across Pacific Gas & Electric's service territory in California. California has some of the best sun and the highest electricity prices in the country. Three years into NEM 3.0, which cut the value of exported power by three-quarters, we wanted to know whether either of those advantages still shows up in the answer.
Companion study: we ran the same method across Ameren territory in Illinois, where the economics rest on a state SREC rather than the rate.
The short answer
Forty-five of fifty-six rate a good investment on the same bar our calculator holds a design to — payback inside ten years and more than $15,000 saved over 25. The other eleven rate marginal. Not one rates poor, shaded or otherwise, which is not true of any other territory we have studied. Best payback 8.6 years, median 9.5, worst 13.6. Median saving $108 a month, $30,520 over 25 years.
It beats the market after tax. The median roof returns 10.6% a year, tax-free, because it is a bill you stop paying rather than income you declare. That is worth 12.4% before tax at the capital-gains rate — ahead of the S&P 500's long-run average of about 10%, and more than double the 4.5% a 25-year municipal bond pays. This is the first territory we have studied where the median roof clears the market rather than matching it.
NEM 3.0 took a lot and left enough. Connect today and exports earn about a quarter of retail. That costs 3.9 years of payback, 8.2 points of return and $16,198 of net present value against a neighbour who connected before April 2023, who is on 5.6 years and 18.7% and keeps those terms for twenty years. It is the largest policy cut we have measured anywhere, and the roofs still work.
What you pay your installer matters more than the policy. These numbers assume the $3.20/W competitive benchmark. Pay the $4.23/W that Californians actually paid on average in 2023–24 and the median payback is 12.1 years, not 9.5 — a 2.6-year swing from the quote alone. More starkly: at the benchmark price 45 of the 56 roofs rate a good investment; at the median price not one does. Get three quotes.
Do not switch to time-of-use for solar alone. It is the obvious move and the wrong one: median payback goes from 9.5 years to 10.8. E-ELEC averages 32.6¢ against E-1's 38.0¢, so every unit your panels offset is worth less, and panels make their power hours before the 4–9pm peak begins.
A battery still does not pay for itself. Not on one of the 56, even under NEM 3.0. The best case shifts 1,513 kWh a year into your own house and saves $458, against $8,500 installed — 18.6 years before discounting, degradation or replacement. Buy one for the outages, not for the arithmetic.
1. Research question
Is solar a good financial investment for homeowners in PG&E territory, three years into NEM 3.0? We ran the SIA calculator over 56 randomly selected rooftops, designing an optimal layout for each and assessing payback period, monthly and lifetime savings, return on investment and net present value — then re-valued every one under the rule that applied before April 2023, to measure what the change actually cost.
2. Method
We sampled 61 detached houses at random from OpenStreetMap across 17 northern and central Californian cities, four per city, using a fixed random seed so anyone can draw the same houses again and check our work. Randomly selecting the homes was necessary to avoid introducing bias into this analysis.
Five were then dropped. Four in Merced are served by Merced Irrigation District, a municipal utility that is not on NEM 3.0 at all, and one roof returned no usable geometry. Municipal utilities matter a great deal here — Sacramento, Palo Alto, Santa Clara, Alameda, Redding, Modesto and Turlock all run their own systems and mostly still offer full retail net metering — so the calculator resolves each address to its actual electric utility rather than assuming. That left 56 homes across 16 cities.
Figure 1. Left: the seventeen cities at their true coordinates, marked by what the utility lookup returned — gold for PG&E, dashed red for Merced, which turned out to be served by its own irrigation district. The state outline is schematic; the dots are not. Right: the sun's track across a full day at six of the sampled homes, on the summer solstice, an equinox and the winter solstice, at each roof's own latitude. The green tick is the direction that roof faces. Concord's faces 289° — west-north-west — and makes 1,170 kWh per kW against Richmond's 1,684 off a nearly identical system.
Under the hood in the SIA calculator: runs for any address you enter.
| Held constant | Value | Source |
|---|---|---|
| Household consumption | 6,036 kWh/yr | California residential average, 503 kWh/month — EIA (2024). Among the lowest of any state |
| Average monthly electric bill | $191 | 503 kWh/month at the E-1 rate below; excludes the $24.12 fixed charge, which solar does not offset |
| Electricity rate | 38.0¢/kWh | PG&E E-1 Residential Service, filed tariff via OpenEI URDB |
| Installed cost | $3.20/W | DOE PVSCB Q1-2025 modelled competitive benchmark + BLS labour, California — the calculator's own default. The $4.23/W market median is swept in section 6 |
| Federal tax credit | 0% | Section 25D expired for expenditures after 31 December 2025 |
| Export credit | 25% of retail | NEM 3.0 Net Billing Tariff: exports earn avoided cost. Systems interconnected before 15 April 2023 keep full retail for 20 years |
| State incentive | none | California has no residential SREC. SGIP is storage-only and means-tested, so it is excluded here |
| Maintenance | $34/kW/yr | NREL Annual Technology Baseline, residential PV fixed O&M |
| Escalation / degradation / discount | 2.5% / 0.5% / 5% | Planning assumptions |
Every roof was modelled individually: Google's 3-D building data for plane geometry, per-plane shading measured from Google's annual flux raster, and one NREL PVWatts® simulation per plane at its own tilt and azimuth. Systems were sized to each household's own consumption. Several states cap a system at your own consumption; California does not. But under NEM 3.0 an exported unit is worth a quarter of a used one, so building past your own load is a decision to sell cheap. Median system: 4.4 kW producing 1,524 kWh per kW a year.
What “modelled individually” means
Three of the 56, chosen to span the shading range rather than to flatter the result. The left column is Google's aerial photograph. The right column is the same roof as Google's annual flux raster measures it — one reading of annual sunlight for every 10 cm of roof — with the panels the model actually placed drawn on top.
What the roof looks like
What the model measures
Napa — one plane, open. A single face carrying all ten panels, with almost nothing overhanging it. Mean shade factor 0.89 · 1,551 kWh per kW per year · 4.0 kW.
Fairfield — the median roof. Slightly better aimed than Napa and slightly more obstructed, which nets out ahead. Mean shade factor 0.86 · 1,635 kWh per kW per year · 4.0 kW.
Chico — the worst in the sample. Two planes under heavy canopy, needing 15 panels and 6.0 kW to cover the same load the others do with 10 and 4.0. Mean shade factor 0.62 · 1,011 kWh per kW per year — a third less sun than Fairfield off the same state.
Figure 2. Each frame is 28 m across; the scale bar is 5 m. Faded roofs are the neighbours'. The shade factor is the measured flux divided by what an unshaded plane at the same tilt and azimuth would make, so 0.62 means this roof loses nearly two-fifths of its sunlight to obstruction. A dark plane is not automatically a shaded one — a north face is dark because of where it points. Homes are identified by city only.
Under the hood in the SIA calculator: measured on every roof it models.
Figure 3. One roof taken apart into the planes Google resolved, each drawn at its own bearing, foreshortened by its own pitch and scaled by its own area. Gold planes carry panels; grey ones were resolved and rejected.
Under the hood in the SIA calculator: shown per roof; panels editable plane by plane.
Figure 4. Google returns every position a panel could occupy, ranked by what it would generate; the calculator walks that list and stops when output covers the household's consumption. Under NEM 3.0 that stopping point is a financial decision rather than a legal one — every panel past it sells at a quarter of retail.
Under the hood in the SIA calculator: re-walked whenever your usage changes.
Figure 5. Every stage between typing an address and a payback number, and what each produced for the median home in this study.
Under the hood in the SIA calculator: this whole chain, every time.
3. What the model runs on
Figure 6. The load shape behind every number here, for California specifically. Californian households use less than almost any state and peak in the evening, hours after the panels have stopped — which is exactly why NEM 3.0 hurts and why the time-of-use rate does not help.
Under the hood in the SIA calculator: used whenever a battery is modelled.
Figure 7. PG&E's two residential tariffs decoded from the filed period grids. E-1 is flat at 38.0¢. E-ELEC has three genuine periods a day and a 55.2¢ peak from 4pm to 9pm — real time-of-use, unlike many records in the national database.
Under the hood in the SIA calculator: your utility's filed tariff, looked up by address.
Figure 8. One house, two days, everything at once: generation, household load, what a battery does about the gap, and its state of charge. The July surplus is enormous and most of it leaves the house at a quarter of retail. That single fact is the whole of NEM 3.0.
Under the hood in the SIA calculator: all 8,760 hours, free with an account.
4. Results
Figure 9. Tightly clustered, and every one inside fourteen years. The spread is narrow because California's rate is high enough that even a mediocre roof saves real money: returns run 6.2% to 11.8%, and the floor sits well clear of zero.
Under the hood in the SIA calculator: runs for any address you enter.
| Metric | Worst | Median | Best |
|---|---|---|---|
| Payback period | 13.6 yrs | 9.5 yrs | 8.6 yrs |
| Monthly saving, year 1 | $100 | $108 | $112 |
| Return (IRR) | 6.2% | 10.6% | 11.8% |
| Net present value | $2,718 | $9,947 | $11,506 |
| 25-year net savings | $22,253 | $30,520 | $33,022 |
Throughout this study a home is judged on the bar the calculator holds a design to, not on whether it merely breaks even: good means payback inside ten years and more than $15,000 saved over 25; marginal means inside fifteen years and more than $5,000; anything else is poor. The worse of the two axes decides.
A 10.6% annual return, tax-free, on an asset bolted to your roof. The S&P 500 has returned roughly 10% a year since 1928 — before tax, and with single years that lost 44%. A 25-year AAA municipal bond, the usual tax-free benchmark, pays about 4.5%.
Figure 10. The median roof, every dollar the model moves over 25 years. There is no rebate and no SREC in California, so nothing softens the outlay — the line climbs on bill savings alone, which is why it is straighter than the Illinois equivalent. The dashed line is the same money discounted at 5%, so where it ends is the net present value.
Under the hood in the SIA calculator: the series behind your payback and NPV.
5. What NEM 3.0 cost
On 15 April 2023 California replaced net metering with the Net Billing Tariff. Exports stopped earning the retail rate and started earning avoided cost, which works out around a quarter of retail. Anyone connected before that date keeps the old terms for twenty years. We valued the same 56 roofs both ways.
Figure 11. Same roofs, same prices, same installer quotes — only the export rule differs. NEM 3.0 costs 3.9 years of payback, 8.2 points of annual return and $16,198 of net present value on the median house. What is striking is that the roofs absorb it. California's rate is high enough that even a quartered export credit leaves a 10.6% return standing, which is not something a weaker market could have taken.
Under the hood in the SIA calculator: export rules resolve by address, so you see your state's rule rather than a national assumption.
6. What you pay matters more than the rule
Every number above assumes the $3.20 per watt that an efficient competitive install costs in California today. That is the calculator's default. It is not what most Californians paid: the LBL Tracking the Sun median for 2023–24 installs is $4.23, a third more.
| Installed price | Median payback | Median return | Median NPV | Verdict |
|---|---|---|---|---|
| $3.20/W — competitive benchmark | 9.5 yrs | 10.6% | $9,947 | 45 good, 11 marginal, none poor |
| $4.23/W — what Californians paid | 12.1 yrs | 7.5% | $5,469 | none good, 55 marginal, 1 poor |
That is a 2.6-year swing from the quote alone — two-thirds of what NEM 3.0 cost, and unlike NEM 3.0 it is entirely within your control. It is also the difference between a market where four roofs in five are a good investment and one where none are. The most valuable thing a Californian homeowner can do before signing is get a second and third quote. The most valuable thing an installer can do is not tell you that.
Under the hood in the SIA calculator: the price per watt is an editable field, and the calculator shows both the benchmark and the state median beside it.
If you do not have $14,080 in cash
A 10% deposit at 7% over fifteen years means a median $1,408 down and a $114 monthly payment, against a median first-year bill saving of $108. That is a net $3 a month out of pocket — effectively nothing — and on 26 of the 56 the savings cover the payment from the first month. Median payback stretches to 10.2 years and the 25-year position to $23,298, the difference being the interest; every one of the 56 still comes out ahead.
That is a consequence of the rate. At 38¢ a kilowatt-hour a Californian roof earns enough every month to very nearly service its own debt, which is not true in most of the country. That is the single most useful fact in this study for someone who wants solar and does not have the cash.
Under the hood in the SIA calculator: the loan panel shows the net monthly figure directly, and withholds a rate of return when the cash flows cannot support a meaningful one.
7. The time-of-use trap
PG&E offers E-ELEC, a time-of-use rate with a 55.2¢ peak between 4pm and 9pm and 28.5¢ off-peak. Moving to it is the standard advice for solar owners, and for a system without a battery it is the wrong move.
Figure 12. The reason is arithmetic rather than subtle. E-ELEC averages 32.6¢ against E-1's 38.0¢, so every unit your panels offset is worth 14% less. The high peak would only help if you were generating into it, and you are not: panels produce at midday and the peak starts at four. Without storage to move the power, a time-of-use rate mostly just lowers the value of what you make.
8. Batteries: still no, even here
This is the result we expected to come out the other way. Under NEM 3.0 a unit kept at home is worth 38.0¢ and the same unit exported is worth about 9.5¢, so storage finally has a real job. Under the full retail netting California left behind there was nothing to arbitrage at all. It is still not enough.
Figure 13. The gap is not close. A battery that shifts 1,513 kWh a year into the house saves $419 on E-1 and $458 on the time-of-use rate — the tariff switch is worth about 9%, not the doubling people assume. Against $8,500 installed that is 18.6 years undiscounted, and the battery will need replacing before then. None of the five batteries in our catalogue paid for itself on any of the 56 homes.
Under the hood in the SIA calculator: batteries are dispatched hour by hour against your state's load shape, on both your current tariff and the time-of-use alternative.
That is a statement about money, not about batteries. Backup through a public safety power shutoff is a real benefit and this study does not price it. If you want one for that reason, buy it for that reason.
So the honest headline is "California works, if you buy it right"
The sunshine is genuinely better and the electricity is genuinely dearer, and between them they absorb a policy cut that would have sunk a weaker market. Every roof we drew came out ahead at the benchmark price. But the gap between a competitive quote and the average one Californians actually signed is worth more than the policy change everyone is talking about, and it is the one variable here the homeowner decides.
9. Limitations
- Fifty-six homes across sixteen cities describes a distribution. It is not enough to compare cities against one another.
- Roof condition is not modelled. A roof needing replacement in eight years adds a removal-and-reinstall cost that can invert any verdict here.
- Consumption is held at the state average. Californian households vary enormously — a home with air conditioning in Fresno and one in Berkeley are not the same problem.
- E-1 is PG&E's baseline residential rate. Households far above baseline allowance pay more, which would improve these numbers; the study does not model tiering.
- SGIP could change the battery answer for a household that qualifies. It is means-tested and equipment-specific, so we have left it out rather than assume it.
- These numbers are for a system energised now. Anything interconnected before 15 April 2023 is on NEM 2.0 and keeps it for twenty years.
- An earlier draft of this post used the $4.23/W market median as its baseline rather than the $3.20/W competitive benchmark, and reached materially worse conclusions as a result. The benchmark is used throughout, with the median as the sensitivity in section 5.
- Modelled production carries the usual PVWatts uncertainty, and shading is measured from imagery rather than surveyed on the roof.
10. Check it yourself
Every number here came from the free calculator on this site, running public data: NREL PVWatts® for production, Google Solar for roof geometry and shading, OpenEI URDB for the PG&E tariffs, and EIA for consumption. The full methodology documents every model, default and data source, with citations. The municipal bond yield quoted in the summary is the approximate 25-year AAA tax-free yield to maturity, FMSbonds, 7 September 2026.
The dataset is archived alongside the model: the 61 homes as sampled, all 56 valued under three regimes and two price bases, every held-constant input with its source, and the scripts that produced them — with the commit the model was at when it ran.
The point of publishing the method is that you should not have to trust the conclusion. Put your own address in and see whether your roof looks like the median one. No phone number, no sales call, and if the numbers do not work at your address we will say so.
Same method, different market: Is solar worth it in Illinois?
Study run 8 September 2026. 61 homes sampled with seed 20260908; 56 in PG&E territory after utility resolution. Informational estimates, not installation quotes.
Read next
The same method across Ameren territory in Illinois, where the economics rest on a state SREC rather than the electricity rate — and leases and PPAs now lose money.
Is solar worth it in Illinois? →