Most homeowners ask the battery question backwards. They ask "what size battery do I need?" when the real question is "what am I trying to do with it?" The US average home uses 29 kWh per day (U.S. EIA, 2023), but whole-house backup is a very different job than storing solar export for the evening. Battery choice also depends on your solar inverter compatibility with batteries.
What Actually Drives Battery Size?
Three numbers set your baseline: daily use, backup loads, and duration. The average US household uses 10,500 kWh a year, 29 kWh a day (U.S. EIA, 2023). Whole-home backup runs 17-24 kWh a day (HVAC 10-15, water heater 3-5, fridge 1.5, plug loads 2-3); essentials, fridge, lights, router, a CPAP, run just 3-4 kWh. That 4-versus-24 gap is why quotes range from one Powerwall to four (appliance nameplate data; U.S. EIA, 2023).
What Are the Two Real Use Cases?
Storage does two jobs with different sizing logic; get it wrong and you overspend or run dry fast. NREL's 2024 analysis found most US installations land at 10-15 kWh, and systems paired to 6-10 kWh daily exports win on cost-per-cycle (NREL, 2024).
Use Case A: Daily solar self-consumption
Your panels peak from 10 AM to 3 PM, so daytime surplus flows to the grid, often for little where net metering died. A battery shifts it to evening. If your 8 kW system exports 6-10 kWh, a 10 kWh battery catches nearly all of it; 15 kWh adds headroom. Rule of thumb: 1 kWh of battery per 1-1.5 kW of panels, so an 8 kW array wants 8-12 kWh, a 12 kW array 12-18 kWh.
If your tariff is dynamic rather than flat, there is a third job the same hardware can do: buy low and discharge at the evening peak. Our battery sell hours calculator shows the daily price shape for ten markets, and the spread it reports is what decides whether that is worth the extra cycles.
Use Case B: Emergency backup
The question is how many days you need without the grid. For essentials over 2-3 nights (most US weather outages), 10-13 kWh is plenty; a single Tesla Powerwall 3 at 13.5 kWh usable covers them 3-4 days with zero recharge. Whole-home for 24 hours needs 20-30 kWh, or two Powerwalls (27 kWh). Want 48 hours through ice storms? That's 40+ kWh, three to four units. See our off-grid battery systems guide.
Don't ask what size battery you need. Ask what you're trying to do with it.
Should You AC-Couple or DC-Couple It?
Sizing tells you how many kilowatt-hours. This tells you whether the battery can go on your roof's existing system at all, and it's the question that decides most retrofit quotes.
| DC-coupled | AC-coupled | |
|---|---|---|
| Where the battery sits | on the panels' DC side, behind one hybrid inverter | on the household AC side, with its own inverter |
| Round-trip efficiency | ~95-97% | ~90-93% |
| Best for | new installs, or a solar system due an inverter swap | retrofits onto a working system |
| Main cost | you're buying or replacing the hybrid inverter | you're paying for a second inverter inside the battery |
The efficiency gap is real but small, and people overweight it. Three percentage points on a 10 kWh battery cycling daily is roughly 110 kWh a year, call it $20-35 depending on your rate. That is not a reason to rip out a functioning five-year-old string inverter, which is what chasing DC coupling on a retrofit usually means.
The practical rule: installing solar and storage together, DC-couple. Adding storage to an array that already works, AC-couple and don't feel bad about it. The exception is if your existing inverter is already near end of life, because replacing it anyway turns the hybrid upgrade into a marginal cost rather than a new one.
What Backup Actually Requires
Neither coupling choice gives you outage backup on its own. Grid-tied inverters are required to shut down when the grid drops, so keeping the lights on needs hardware that islands your house: a backup gateway or automatic transfer switch, plus a decision about scope.
Whole-home backup sounds better than it usually is. Backing up the whole panel means the battery serves your air conditioning and your oven, and a 10 kWh unit facing a 5-ton AC unit empties in about two hours. A critical loads subpanel, fridge, a few circuits of lighting, internet, well pump if you have one, is the version that runs overnight. Budget $2,000-3,500 for the gateway and the subpanel work, and get it itemised separately in the quote; it is frequently buried in a bundled price where you cannot compare it between installers. One modular route, scaling from 9 to 36 kWh in 3 kWh modules, is covered in the Generac PWRcell review.
Which Batteries Are Available in 2025-2026?
The market settled on four serious options. Compare five numbers: usable kWh, chemistry (LFP vs NMC), depth of discharge, continuous output, and cost per kWh (manufacturer datasheets, 2026).
| Battery | Usable kWh | Chemistry | DoD | Output | Installed cost |
|---|---|---|---|---|---|
| Tesla Powerwall 3 | 13.5 | LFP | 100% | 11.5 kW | ~$16,000 ($1,185/kWh) |
| Enphase IQ Battery 5P (3-unit stack) | 15 | LFP | 100% (5.0 of 5.0 kWh per unit) | 3.84 kVA each | ~$18,000 ($1,200/kWh) |
| Franklin WH10 | 10 | LFP | 95-100% | VPP-ready | $12,000-14,000 |
| LG RESU16H Prime | 16 | NMC (check datasheet) | 80-85% | AC-coupled | Varies by installer |
The Tesla Powerwall 3: 13.5 kWh usable, LFP, 100% DoD, 11.5 kW output (runs central AC plus a fridge). About $16,000 through Tesla ($1,185 per usable kWh), $1,000-2,000 less via third parties.
The Enphase IQ Battery 5P: 5.0 kWh usable per unit, stackable to 4 (20 kWh), 90% AC round-trip efficiency (96% on the DC side), 3.84 kVA continuous each. Three units (15 kWh) run about $18,000 ($1,200 per usable kWh). Self-consumption buyers usually land on two units (10 kWh), backup buyers on the 15 kWh stack.
The Franklin WH10: 10 kWh usable and VPP-ready. Real money: California, Texas, and several northeastern states pay $0.50-1.00 per kWh at peak. Installed cost $12,000-14,000, cheaper per kWh than the rest, though fewer installers stock it.
The LG RESU16H Prime: 16 kWh usable, the largest mainstream unit, AC-coupled to almost any inverter, the default retrofit pick. One catch: the original RESU used NMC capped at 80-85% DoD, so check the datasheet before assuming a full 16 kWh.
The two units homeowners shortlist most often go head to head in Powerwall 3 vs Enphase 5P.
How Long Does a Home Battery Last, and Does Chemistry Matter?
Chemistry matters, and the warranty is where you can read it. LFP (lithium iron phosphate) is the chemistry Enphase's datasheet specifies for the IQ Battery 5P, and it's what most current home batteries use; older NMC units such as the original LG RESU were rated for less of their nameplate capacity, so a "10 kWh" NMC unit might deliver 8 kWh in practice. Always ask for the chemistry and the usable kWh.
The numbers that decide lifespan are in the warranty, not the brochure. Here are the two most-quoted units, read from the manufacturers' own documents:
| Tesla Powerwall 3 | Enphase IQ Battery 5P | |
|---|---|---|
| Warranty term | 10 years | 15 years or 6,000 cycles, whichever comes first |
| Capacity floor | 70% of 13.5 kWh at year 10 | more than 60% of rated capacity |
| Cycle limit | unlimited for solar self-consumption, time-of-use shifting and backup; 37.8 MWh of throughput for anything else | 6,000 |
| AC round-trip efficiency | 89% (solar to battery to home or grid) | 90% (96% DC) |
Two things follow from that table. First, plan on the floor, not the day-one figure: a Powerwall 3 is only guaranteed to hold about 9.5 kWh by year 10 (70% of 13.5), so a household that sized to exactly 10 kWh of usable backup should buy headroom. Second, the cycle arithmetic is gentler than it sounds. One cycle a day is 365 a year, so Enphase's 6,000 cycles last about 16 years and the 15-year clock runs out first. Tesla's throughput cap, 37.8 MWh, equals 2,800 full cycles of 13.5 kWh, roughly 7.7 years of daily cycling, but it only applies if you use the battery for something outside the listed applications, such as a utility program the warranty doesn't name. Check that list before you enrol in anything (Tesla Powerwall Limited Warranty, USA, rev. 2.6; Enphase IQ Battery 5P datasheet).
Heat is the quiet lifespan variable. Enphase publishes an optimum operating range of 0-30 C for the 5P, which is why a garage that cooks in August is a worse home for a battery than a wall that stays in the shade. Ask where the unit goes before you ask the brand.
Does Fast Cycling Wear an LFP Home Battery Faster?
Less than the forum lore says, and the hardware won't let you find out anyway. Battery engineers talk in C-rate: 1C means charging or draining the whole pack in one hour, 0.5C in two hours. Fast cycling is anything above 1C. Now look at what the datasheets allow:
| Unit | Usable kWh | Max continuous discharge | Discharge C-rate | Max charge | Charge C-rate |
|---|---|---|---|---|---|
| Tesla Powerwall 3 | 13.5 | 11.5 kW | 0.85C | 5 kW single unit (8 kW with Expansion units) | 0.37C |
| Enphase IQ Battery 5P | 5.0 | 3.84 kVA | 0.77C | not listed separately; the same six microinverters do both directions | 0.77C at most |
| Generac PWRcell 2 | 9-18 per cabinet | 11.5 kW, but only with two cabinets holding at least 7 modules (21 kWh) | 0.55C | not printed on Generac's page | unknown |
| Anker SOLIX C2000 Gen 2 (portable, for scale) | 2.048 | 2,400 W | 1.17C | 1,800 W AC, 2,600 W AC plus solar | 0.88C to 1.27C |
Read that table as a ceiling, not a habit. Not one fixed home battery here can discharge at 1C, and charging sits lower still, 0.37C on a single Powerwall 3. Solar charging is gentler again: a 10 kWp roof at 52 N makes about 9 kWh on a December day according to the EU's PVGIS model, which trickles into a 13.5 kWh pack at a fraction of its rated charge power. The only product in the table that crosses 1C is a portable station, and it does so on a 2 kWh pack where 2,400 W is a kettle.
So what does fast cycling do to LFP chemistry when it does happen? The best public evidence available is the multi-year Sandia National Laboratories study by Preger and colleagues (Journal of the Electrochemical Society, 2020), which cycled commercial LFP, NCA and NMC cells at 0.5C, 1C, 2C and 3C discharge, at 15, 25 and 35 C, across three depth-of-discharge windows. Every cell was charged at 0.5C, so the study says nothing about fast charging; keep that in mind. Three findings matter for a homeowner:
- LFP reached 2,500 to 9,000 equivalent full cycles before 80% capacity across all conditions, and most LFP cells hadn't got there when the study ended, so those are extrapolations.
- Temperature moved LFP fade more than discharge rate did. Fade rose from 15 C to 35 C, which is the same message Enphase's 0-30 C optimum sends.
- LFP barely cared which state-of-charge window it lived in, unlike NCA and NMC. That's the datasheet reason Enphase and Tesla can rate 100% depth of discharge.
The spread, 2,500 against 9,000, is the honest number. A 3.6x range from cells that all stayed within their makers' limits means a cycle rating on a brochure is a point inside a cloud. No bench data on a Powerwall itself exists in this comparison, so treat the cell study as a range, not a prediction; what holds up is that the warranty, not the cell study, is the figure you can enforce. Tesla's 37.8 MWh throughput cap for non-listed uses works out to 2,800 full cycles, right at the bottom of Sandia's LFP range, which looks like a deliberate choice rather than a coincidence.
Where would fast cycling actually bite? Utility programs that call for repeated full-power discharges on peak afternoons, in a hot garage, in August. Heat plus rate plus a near-empty pack is the corner of the map every chemistry dislikes. If you enrol in a virtual power plant, check two things before the cycle count: the ambient temperature where the unit hangs and whether the program sits inside the warranty's list of covered uses.
What Happens to a Home Battery Below Freezing?
Cold is a temporary tax; heat is the permanent one. Lithium cells of every chemistry charge badly below 0 C because lithium ions slow down entering the anode and can plate out as metal instead, which is a permanent loss and, in the worst case, a safety fault. Every reputable battery management system either cuts charge current back near freezing or refuses to charge at all. Discharging in the cold is allowed much further down, with less available energy and power until the pack warms up.
The published envelopes, read straight from the manufacturers:
| Unit | Charging allowed | Discharging allowed | Full performance | Derating notes |
|---|---|---|---|---|
| Tesla Powerwall 3 | -20 to 50 C | -20 to 50 C | not split out | de-rates above 40 C |
| Enphase IQ Battery 5P | -20 to 50 C | -20 to 55 C | 15 to 45 C charging, 0 to 50 C discharging | currents de-rated outside those bands; optimum 0 to 30 C |
| Anker SOLIX C2000 Gen 2 | 0 to 40 C | not printed on the product page | not printed | a second "recharging" range of -20 to 40 C appears without explanation; check the manual |
Notice the gap in Enphase's numbers. The unit is permitted to charge at -20 C, but only promises full charging current from 15 C upward. Between those two figures the battery is working, just slower, and a homeowner reading the app on a January morning sees a pack that fills less than the spreadsheet predicted. Is that degradation? No. It reverses by lunchtime. The damage case is a cheap pack with a BMS that doesn't derate, which is a reason to care about the certification list (UL 9540A on the Powerwall 3, for example) more than the sticker capacity.
Two practical consequences for anyone north of about 52 degrees latitude. First, winter solar rarely fills the battery whatever the temperature: PVGIS puts a south-facing 35-degree roof in Berlin or Warsaw at 25-33 kWh per kWp for December and January, against 130-135 in June, so a 10 kWp array yields 8-11 kWh on an average winter day for a house that uses more than that. The battery's job in winter is time-of-use shifting from the grid, if your tariff allows charging from it, or backup. Second, the mounting wall matters more than the brand. A garage that stays above freezing beats an exposed north wall, and an exposed north wall still beats a conservatory that hits 45 C in July, because the heat damage is the kind the warranty counts.
The siting questions rank in this order: shade in summer, above freezing in winter, then everything else. Disagree if you've got a Minnesota garage that disagrees with you, but that's the order the datasheets argue for.
What Does Battery Storage Cost in 2026?
Installed storage runs $800-$1,200 per usable kWh in 2026 (NREL, 2024). A $16,000 Powerwall used to net $11,200 after the 30% federal ITC, but that credit ended December 31, 2025 under the One Big Beautiful Bill Act, so a 2026 buyer pays the full $16,000 unless they route through a lease or PPA. California's SGIP has paid $400-500 per kWh for low-income households, dropping net cost to $10,400-$12,800 per system independent of the federal credit. See our solar and battery tax credits guide.
What Decides Whether a Battery Pays Back?
Three inputs: the price you avoid at the evening peak, the price you'd otherwise get for exporting that solar, and round-trip efficiency. A battery earns the gap between the first two, minus the energy it loses on the way through. Storing 10 kWh and delivering 10 kWh at 90% efficiency means charging with about 11.1 kWh, so the daily value is 10 x peak price minus 11.1 x export price.
Here's that formula run at 330 full cycles a year (cloudy and travel days skipped), against a $12,000 installed 10 kWh unit. It's arithmetic on stated prices, not a measured result:
| Tariff situation | Peak price | Export credit | Value per year | Simple payback |
|---|---|---|---|---|
| Time-of-use, export credit near zero | $0.45 | $0.05 | ~$1,300 | ~9 years |
| Time-of-use, moderate export credit | $0.40 | $0.10 | ~$950 | ~13 years |
| Flat rate, low export credit | $0.15 | $0.04 | ~$350 | ~34 years |
| Flat rate, full retail net metering | $0.15 | $0.15 | below zero | never |
Notice what isn't in the table: federal money, which ended for owners on December 31, 2025. And notice the last row. Under full retail net metering the grid already acts as a free battery, so a physical one just loses 10% of the energy it handles.
Capacity fade pulls the first row down further. If the battery slides linearly to the 70% warranty floor over ten years, it earns about 86% of its day-one rate on average, around $11,300 over the decade against a $12,000 price. So in the best tariff a battery roughly breaks even inside its warranty on arbitrage alone. The rest of the case is backup value and any utility program that pays for discharge, which is exactly why backup should be priced as its own line in the quote rather than assumed to be free. For the cost lines calculators omit, read that guide before trusting any payback figure.
How Do You Size It Yourself?
Pull your real daily kWh (12 months of bills divided by 365; a condo runs 12-15 kWh/day, a 3,000 sq ft house with a pool pump 45-55). Add your backup loads: a 1,500W fridge is about 1.5 kWh, a 3-ton, 3,500W AC over 4 hours is 14 kWh, and skip the EV charger. Multiply load by days, add 10-15% for losses, then credit a sunny 8 kW array's 20-30 kWh recharge. Match that to a product; get three quotes in usable kWh, not nameplate. Under-buying is the more common mistake: one Powerwall covers essentials for days but runs out on day two if the whole panel is backed up.
Summary
Two use cases, simple. Self-consumption: match the battery to your export, 10-15 kWh for an 8-12 kW array. Backup: load times window, one Powerwall for 3 days of essentials, two for whole-home. Confirm usable kWh, warranty floor, and chemistry on every quote, and price the payback from your own peak-versus-export spread, not from a headline. Then read the solar payback period and ROI analysis.