When a Battery Actually Pays

A battery is sold as the thing that makes NEM 3 work. On a real PG&E quote, it is often the thing that makes the investment stop working. Here is why, and the one case where it flips.

The samples in the calculator come from one Novato proposal: 4.6 kW of west-facing panels, a Powerwall 3 at $12,800, charger and permits left on the solar line at $14,006. Same roof, same rates. The only question is whether the pack earns the extra check.

Ask the incremental question

Don't ask whether solar-plus-battery beats doing nothing. Ask whether the pack beats the same array without it.

First system, solar only is an 8,000 kWh house with the car charged at midday, on net billing, no pack. It returns 9.0% over ten years and pays back in 6.7 years. With a battery is the same house plus $12,800 of storage. Year-one savings rise from $1,834 to $2,184.

That $350 a year is everything the pack added. Even with rate escalation it takes 22 years to return $12,800. The ten-year equivalent return falls from 9% to zero, and leaving the money invested wins by $7,700.

The evening is the ceiling

A 13.5 kWh pack can hold 13.5 kWh. What it can sell is however much load shows up between 4 and 9pm that the panels aren't already covering. In that 8,000 kWh house the whole evening is about 7 kWh a day, and a west-facing array is still producing into part of it. What's left for the battery is about 2.6 kWh a day. A fifth of a cycle.

The midday car makes it worse, but it isn't the cause. Take the car out entirely and the pack still only moves 1,805 kWh a year, worth $663. Better, and nowhere near $12,800. Buying a bigger battery makes this worse, not better, because the ceiling isn't the pack.

On a legacy NEM 2 roof it's worse again. Those midday kWh were already worth full retail as exports, so storing them spends a retail export to avoid a retail import, and loses about 17% in the round trip. Add a shared pack to the 13% California Gold sample and it drops to 3.1%, with payback going 5.7 → 8.7 years.

Move the levers yourself

This runs the same arithmetic the calculator does, on the pack alone.

Ten years is doing a lot of the work

That last slider is the one to play with. The calculator measures returns over your analysis window and counts the system as worth $0 afterwards. No resale, no credit for the twelve or fifteen years of production still to come. At ten years that is deliberately harsh, and it is harsh in a specific way: anything that pays back in year eleven reads as a loss, however good it is.

Drag it on the good case and watch: −8.4% at five years, 10.2% at ten, 15.1% at fifteen, 17.6% at twenty-five. Identical hardware, identical kilowatt-hours. The hard case moves too (−15.4% at ten years becomes +1.3% at twenty-five), it just never becomes a good buy.

This is on purpose. A ten-year window with no residual value is the test a cautious buyer should apply to a roof full of equipment with a warranty. But it means a battery near the edge can be shown as a failure or a success without a single physical fact changing. Read payback and return together, and treat a long horizon as an assumption you are making rather than a result you found. An installer quoting a six-year payback on solar-plus-battery is usually counting twenty-five years of escalating bills, a rebate, or self-use the evening load cannot support.

The case where the pack is the point

One sample does pay: Go with the battery. A 3 kW array stays on its old retail tariff, and a new zero-export array goes in beside it. That's the usual way to add capacity without a new interconnection that would move the whole site off NEM 2. The house is big (16,000 kWh still delivered by the utility, which works out to about 19,900 kWh actually used), and nothing is charging at noon.

Without a pack, the new panels throw away 6,525 kWh a year: they may not export, and midday load is already partly served by the old roof. The battery rescues 4,039 of those into the evening peak. It pays back in 6.4 years on its own, faster than the array beside it, and lifts the addition from 4.2% to 6.6%.

Three conditions, all required:

1. The stored kWh were otherwise worth about $0. Zero-export is the clean case. Net billing at $0.06 is only slightly better. Retail netting is the opposite case.
2. The evening can take close to a full charge. This house absorbs about 11 kWh between 4 and 9pm. A 7,000 kWh house with a car plugged in at noon absorbs two.
3. Peak power is expensive. The spread is what the pack sells. Flatten it and 7.2 years stretches back out.

Weaken any one and the pack stops paying as an investment. It can still be a generator.

What this doesn't price

This tool values a battery as shifted kilowatt-hours. It doesn't price a week of smoke and a PSPS map. It uses a flat export credit rather than an hourly avoided-cost curve, so it understates summer evenings when exports briefly look like peak retail. It doesn't subtract a rebate from the $12,800, doesn't move you to a storage-friendly tariff, and puts no figure on a furnace that runs when the street is dark.

Those are the cases where a pack is the right purchase even with the return under 5%: medical or well-pump backup, a fire-belt house that would otherwise buy a generator, a rebate that cuts the real check toward $8,000, or an export tariff that genuinely spikes at dinner. Run the investment tiles anyway, then decide whether backup is a separate budget rather than a hidden return.

The rule that survives contact with a real quote is narrow. Buy the pack when the kilowatt-hours it stores were going to die anyway, and dinner is still expensive. Buy the panels when the car can eat noon. Don't use one check to answer both.