Continuous vs peak output: reading home battery power ratings

Every home battery publishes at least one power number and most publish two. The larger one is the one that ends up in the brochure. It is also the one that is almost never comparable between two brands, because the duration attached to it varies by more than three orders of magnitude across the models we track, from 300 milliseconds to 30 minutes.

This guide separates the two figures, records what each manufacturer publishes and on what basis, and explains why the second number exists at all. Every figure below comes from a manufacturer datasheet, quoted with the basis that datasheet states. None of it is from our own testing.

The two numbers do different jobs

Continuous power is the output the inverter sustains indefinitely. It is what decides how many things can run at once: the refrigerator, the well pump, the furnace blower, the lights, all added up in kilowatts. In a multi-hour outage this is the only power figure that matters, because runtime is capacity divided by draw, and a rating measured in seconds contributes nothing to hours.

Peak power exists because of motors. An induction motor at rest is effectively a short circuit until it starts turning, so it pulls a large current for a fraction of a second before settling to its running draw. The nameplate figure for that moment is locked rotor amps, or LRA. A refrigerator compressor, a well pump, a furnace blower and a central air conditioner all do this, and the size of the pull is not proportional to the running watts. Goodman publishes 8.2 rated load amps and 41.2 locked rotor amps for its 1.5 ton GLXS3B condenser, a factor of about five between running and starting on the same motor. We work that arithmetic through in the air conditioner guide.

So the peak rating is a permission slip for a moment, not a power reserve. The question it answers is whether the inverter will ride through the start instead of tripping.

Continuous output, model by model

Sorted as each datasheet publishes it. Full sourcing sits on each model page.

• Savant Power Storage 20: 12.5 kW. Published as continuous inverter output on the PS-INV-12.5KW-100A spec sheet.
• Tesla Powerwall 3: 11.5 kW. Tesla's datasheet publishes four selectable nominal output ratings, 5.8, 7.6, 10 and 11.5 kW, and a separate configurable off-grid maximum continuous discharge of 15.4 kW that is available only if the on-grid rating is set to 11.5 kW and the unit is installed with an 80 A breaker.
• Generac PWRcell: 10.5 kW at the six-module M6 configuration, stepping down to 8.7, 7.0 and 5.2 kW at M5, M4 and M3. On this system, power and capacity are the same purchase.
• EG4 PowerPro WallMount: 10.2 kW.
• FranklinWH aPower 2: 10 kW, which FranklinWH prints as "10 kW / 11.5 kVA continuous" and pairs with a maximum charge power of 8 kW.
• Fortress Power eVault Max 18.5: 9.2 kW, published as a recommended continuous discharge rate of 180 A at 51.2 V nominal.
• Panasonic EverVolt 2.0 and Qcells Q.HOME CORE: 7.6 kW, in both cases the hybrid inverter rating rather than a battery figure. Panasonic separately publishes continuous backup power off grid without solar of 5.5 kW on the 9 kWh cabinet and 7.6 kW on the 13.5 and 18 kWh cabinets.
• LG Home 8: 7.5 kVA rated AC power discharging, in both grid connection mode and backup mode.
• Enphase IQ Battery 10C: 7.08 kVA.
• SunPower SunVault: 6.8 kW. Discontinued, listed for owners.
• Anker SOLIX X1: 6 kW on grid, and off grid 6 kW on the 10 kWh model or 6.6 kW on the 15 and 20 kWh models.
• SolarEdge Home Battery 400V and Tesla Powerwall 2: 5 kW.
• sonnen sonnenCore+: 4.8 kW.
• Enphase IQ Battery 5P: 3.84 kVA.
• BYD Battery-Box Premium HVS: not published as a system AC figure, because the HVS is a DC battery sold to pair with a third-party inverter that carries its own rating.

Two notes on comparing these. Several makers publish kVA rather than kW, and the two are equal only at unity power factor, which household loads do not have. And on multi-box systems the number belongs to whichever component is smaller: Generac's spec guide footnotes that system power is limited by the battery's maximum continuous power, so the 11.5 kW printed on the PWRcell 2 Inverter requires two battery cabinets.

Peak output, and the duration attached to it

This is where the field stops being comparable. Same column heading, wildly different measurement window.

• Savant Power Storage 20: 30 kW for 300 milliseconds, on grid. Savant's spec sheet publishes four separate surge figures. On grid: 30 kW for 300 ms and 28 kW for 25 s. Off grid: 24 kW for 400 ms, 19.1 kW for 1 s, and 16 kW for 10 s. The 30 kW headline is the shortest of the four and applies while the grid is up, which is the condition under which you least need it. During an outage the ten-second figure is 16 kW.
• FranklinWH aPower 2: 15 kW for 10 seconds, printed as Max Peak Output Power, alongside a Max Transient load capacity of 25 kW at 1 second and a Max Half-Wave Load Capability of 5 kW. Three figures for three timescales on one page.
• Anker SOLIX X1: 12 kW for 10 seconds on the 15 and 20 kWh models, or 9 kW for 10 seconds on the 10 kWh model. Anker footnotes both as valid at an ambient temperature of 50 to 131 degrees Fahrenheit with a state of charge above 35 percent.
• Fortress Power eVault Max 18.5: 12 kW for 30 minutes. Fortress labels this a Peak Discharge Rate of 230 A and gives it a half-hour window, which is a different class of specification entirely from a ten-second overload. Its actual surge line is separate: Maximum Surge Power Rate of 250 A, 12.8 kW, for 5 seconds.
• SunPower SunVault: 10 kW, per the SunVault ESS datasheet, RevC 2021. Discontinued.
• LG Home 8: 9.0 kVA for 10 seconds, published in the backup mode table as Max AC Power (Discharging), with a matching 37.5 A for 10 seconds. Worth knowing before you cite LG's numbers: the stacking table further down the same spec page lists 7.5 kVA as the 10 second figure and 9 kVA as the rated figure for one unit, which reverses the labels used in the backup mode table above it. We record the backup mode table and flag the inconsistency rather than pick a winner.
• Enphase IQ Battery 5P: 7.68 kVA, and the datasheet shows where it comes from. Enphase publishes peak output current of 32 A for three seconds and 25.6 A for ten seconds. At 240 V those work out to 7.68 kVA and about 6.1 kVA respectively, so the advertised peak is the three second figure and the ten second figure is roughly a fifth lower.
• SolarEdge Home Battery 400V: 7,500 W, labelled in the datasheet as Peak Output Power (for 10 seconds). One of the cleanest lines in the set: the duration is in the row label.
• Tesla Powerwall 2: 7 kW real power, 7.2 kVA apparent, for 10 seconds, off grid or backup only. Tesla scopes both the duration and the operating mode on the same line.
• Enphase IQ Battery 10C: no peak kVA published, but the same current figures are there: 56 A for three seconds and 44.8 A for ten seconds at 240 V. Enphase chose not to convert them into a headline watt figure on this model, which is why our comparison table shows no peak for it.
• Tesla Powerwall 3, Panasonic EverVolt 2.0, Qcells Q.HOME CORE, sonnen sonnenCore+, EG4 PowerPro, BYD Battery-Box Premium: no peak power figure on the datasheets we cite. That is an absence of a published number, not evidence of a weak inverter.

Set the extremes side by side and the problem is obvious. Fortress publishes a 12 kW peak that lasts half an hour. Savant publishes a 30 kW peak that lasts three tenths of a second. Both appear in a "peak power" column. Comparing them as if they measured the same thing is the single most common error in home battery spec tables, including several that look authoritative.

The column that actually answers the motor question

A watt figure with a duration is useful. A figure in locked rotor amps is better, because it is the unit your compressor or pump nameplate already uses, so no conversion or assumption sits between the two numbers.

Three makers in our set publish a figure in locked rotor amps, and two more publish a start-current figure in a different unit that is closer to it than watts are:

• Tesla: Load Start Capability of 185 LRA on the Powerwall 3 and 88 to 106 A LRA on the Powerwall 2, the latter footnoted "Load start capability may vary."
• FranklinWH: Max Load Start Capability of 185 A LRA on the aPower 2.
• Enphase: Power Start capability of up to 48 A LRA on the IQ Battery 5P and up to 90 A LRA on the IQ Battery 10C, both footnoted "Power Start capability may vary."
• Anker: 70 A/s on a single SOLIX X1 and 140 A/s on two units in parallel.
• Generac: Max AC Start Capacity in amps rather than watts, 65 A at M3, 87 A at M4, 108 A at M5 and 130 A at M6, footnoted as based on 2 kW of household consumption before the start, with the battery between 12 and 100 percent state of charge, at an ambient below 104 degrees Fahrenheit. The PWRcell 2 Inverter is separately quoted at 130 LRA.

Notice how much conditioning travels with these. Two say the figure may vary. One is stated with an assumed household load, a state of charge window and a temperature ceiling. That conditioning is a feature, not a weasel: a motor start rating genuinely does depend on what else is running and how full the battery is, and the makers who say so are the ones giving you a usable number.

How to use the two numbers

Start from your own loads, not from the spec table.

For the running load, add up nameplate watts and compare against continuous. This is the figure that governs a long outage. If your essential circuits total 6 kW, a 4.8 kW battery does not carry them regardless of what its peak column says.

For the start, find the largest motor you need and read its nameplate LRA. Then compare against the battery's published LRA if it has one, or against the peak figure at the duration closest to a motor start, which is typically the one to two second row rather than the ten second row. A well pump and a central air conditioner are the two loads where this usually decides the answer.

Ask what happens off grid. Savant's off-grid surge figures are meaningfully lower than its on-grid ones, and Tesla's Powerwall 2 peak is published as off-grid or backup only. Those are the conditions under which you will actually be leaning on the number.

Treat any peak figure without a stated duration as unusable for comparison. Not as a bad product, just as a number with no basis. Where a maker publishes nothing, we record not published rather than borrowing a figure from a retailer page.

Every figure on this page is a manufacturer specification recorded with its published basis, verified against the datasheets we cite on each model page in our battery database. Manufacturers revise datasheets, and the same model number can carry different figures across revisions, so confirm the current document for the exact configuration you are quoted. Your own nameplate is the number that governs, and reading it costs five minutes.