Home battery degradation: what datasheets actually guarantee

Every lithium battery loses capacity as it ages. What varies between manufacturers is not the physics but what they are willing to put in writing about it, and the two things they write down are not interchangeable.

Our warranty guide lines up the full terms for every model we track: years, caps, conditions, who administers a claim. This page does something narrower. It takes the two figures that govern degradation specifically, explains what each one promises, and shows where the published bases stop being comparable. Every figure comes from a manufacturer datasheet or warranty document, quoted with the basis that document states.

Two promises, two clocks

A capacity retention floor is a guarantee about the state of the battery. It says: at the end of the term, this unit will still hold at least X percent of what it held new. It is expressed as a percentage and it names an endpoint.

A throughput or cycle cap is a limit on how much work the battery may do while covered. It says: coverage ends after Y megawatt-hours have passed through, or after Z cycles, whichever arrives first alongside the calendar. It is expressed in energy or in cycles and it names a budget.

These answer different questions and a warranty can carry one, both, or neither. A floor with no cap makes no promise about heavy use. A cap with no floor makes no promise about what you own at the end. The models we track split roughly evenly between the two, and several publish neither.

The confusion that follows from mixing them up is not academic. A buyer who reads "15 years" and assumes 15 years of guaranteed capacity has skipped both the floor and the cap. A buyer who reads "8,000 cycles" and assumes the battery will still be near full capacity at cycle 7,999 has invented a floor that the document does not contain.

The retention floors, and what they work out to

Seven models in our set publish a capacity retention figure. The percentage is the manufacturer's; the conversion to kilowatt-hours is our arithmetic on that percentage and the same datasheet's usable capacity.

• Panasonic EverVolt 2.0: 70 percent at the end of 12 years. Panasonic prints the line as "Energy Capacity at the end of 12 years: 70%" in the battery specification table. On the 18 kWh cabinet that floor is about 12.6 kWh.
• LG Home 8: 70 percent at 10 years. LG's spec page states the warranty as "10 years (Battery: 70% @ 10 years)." On 14.4 kWh usable that is about 10.1 kWh. Read the structure alongside it: LG states the standard limited warranty period is five years, with an extended five years available for qualified products registered with LG's ThinQ service and connected to the internet. Half the term is contingent on an administrative step.
• Tesla Powerwall 3 and Powerwall 2: 70 percent at 10 years, expressed against a 13.5 kWh rated capacity, which is 9.45 kWh by our arithmetic. This one lives in the Tesla Powerwall Limited Warranty (USA) rather than on the datasheet, and the full terms including the two operating buckets are in our Powerwall lifespan guide.
• Savant Power Storage 20: an estimated 75 percent at 10 years. The highest floor among the models we track, and the weakest wording. Savant's spec sheet labels the row "Estimated Capacity, 10 Years," and estimated is not warranted. On 18 kWh usable that is about 13.5 kWh.
• Enphase IQ Battery 5P: greater than 60 percent capacity, up to 15 years or 6,000 cycles, whichever occurs first, with Enphase adding that restrictions apply. About 3 kWh on a 5 kWh unit.
• Enphase IQ Battery 10C: 60 percent capacity, up to 15 years or 6,000 cycles, on the same basis. About 6 kWh on a 10 kWh unit.

Notice the trade rather than the ranking. Enphase carries the longest term in our set and the lowest floor. Panasonic carries a middle term and a higher floor. Savant carries the highest number and the softest verb. There is no ordering of those three that is correct for every buyer, because they are priced against different usage patterns.

Not published on the datasheets we cite: FranklinWH, Generac, SolarEdge, sonnen, Anker, EG4, Fortress Power, BYD, Qcells and the discontinued SunPower SunVault. Several of those are strong products. The absence of a published floor is a gap in what the maker commits to in public, and the correct response is to ask the installer to produce the full warranty document rather than to assume a number or to reject the product.

The caps, and the conversion nobody does

Throughput caps and cycle counts are the other clock. Converting a megawatt-hour cap into equivalent full cycles takes one division, and it is the step that makes two warranties comparable.

Published throughput caps among the models we track:

• FranklinWH aPower 2: 60 MWh aggregate warrantied throughput over 15 years, printed as its own row on the datasheet. Across 15 kWh usable that is about 4,000 full-depth equivalent cycles by our arithmetic, or roughly 267 a year across the term.
• Anker SOLIX X1: 32.9 MWh on the 10 kWh model, 49.4 MWh on the 15 kWh, and 65.8 MWh on the 20 kWh. Divide each by its capacity and all three land at about 3,290 equivalent cycles, so Anker's allowance is a constant per kilowatt-hour of capacity rather than a flat number. That is the cleanest scaling in our set and it makes the figure genuinely portable across configurations.
• Generac PWRcell: 7.56 MWh per module, alongside a 10 year term. At the published 3.0 kWh per module, that is 2,520 full module cycles, roughly seven years at one full cycle a day. Read the words "per module" carefully: a six-module M6 cabinet carries a larger total allowance than the single figure suggests.
• Savant Power Storage 20: 73 MWh or 3,650 cycles, published together as warrantied cycles. Those two are the same number: 3,650 multiplied by the 20 kWh rated capacity is exactly 73 MWh. Which means the cap is calculated on rated capacity, not on the 18 kWh usable figure that follows from Savant's own 90 percent recommended depth of discharge. If you cycle to the recommended depth, the calendar and the cycle count run out before the megawatt-hour figure does.
• EG4 PowerPro WallMount: 82.6 MWh of rated lifetime production, printed beside a separate figure of over 8,000 cycles at 80 percent depth of discharge. Those two do not reconcile on a 14.3 kWh battery, which tells you they rest on different assumptions rather than that one is wrong. We work through that in the warranty guide.
• Tesla Powerwall: 37.8 MWh, which applies only to applications outside the warranty's listed set of solar self-consumption, time-based control and backup. Inside that list the cap is unlimited cycles. Across 13.5 kWh, 37.8 MWh is 2,800 full-depth equivalent cycles.

Published cycle counts rather than energy caps: sonnen sonnenCore+ at 10,000 cycles alongside its 10 year term, the highest count in our set; EG4 and Fortress Power at 8,000 cycles, both stated at 80 percent depth of discharge; Enphase at 6,000 cycles on both models; Qcells at 6,000 cycles at 90 percent depth of discharge; Panasonic at 6,000 cycles at 90 percent depth of discharge.

The footnotes that make cycle counts incomparable

A cycle count is meaningless without three conditions, and manufacturers publish different ones.

Depth of discharge. A cycle to 90 percent depth moves more energy than a cycle to 80 percent, so 8,000 shallow cycles and 6,000 deep ones can represent similar total throughput. Qcells and Panasonic both state 90 percent. EG4 and Fortress both state 80 percent. Savant's rated figure is at 90 percent. Four numbers, two different definitions of a cycle. What depth of discharge does to usable capacity is explained in depth of discharge.

Charge and discharge rate. Qcells prints its test conditions explicitly: 90 percent DOD, 0.2 C charge and discharge, at 25 degrees Celsius. Savant footnotes its rated 6,000 cycles at 90 percent DoD and 0.5 C. A 0.5 C test cycles the battery two and a half times faster than a 0.2 C test, which is harder on cells. Two identical-looking 6,000 cycle claims, measured at rates that differ by a factor of two and a half.

Temperature. Qcells names 25 degrees Celsius in the test condition. Most others do not name one in the cycle-life row, though nearly all publish operating temperature bands elsewhere. Enphase publishes an optimum operating temperature range of 0 to 30 degrees Celsius for the IQ Battery 10C and states that keeping the battery in that range maximizes battery life.

There is one more distinction hiding in this section. Several of these figures are engineering ratings, not warranty terms, and Savant is the model that makes the difference visible by publishing both: 6,000 rated cycles and 3,650 warrantied cycles on the same spec sheet. The rated number describes the cells. The warrantied number describes the contract. When only one is printed, check which one you are reading.

Every headline capacity is a new-unit number

The reason a retention floor exists is that the number on the front of the datasheet is not a durable property.

Tesla footnotes the Powerwall 3's nominal battery energy of 13.5 kWh with "Values provided for 25 degrees Celsius (77 degrees Fahrenheit), at beginning of life." Generac footnotes its 9 to 18 kWh usable energy figures with "At beginning of life at 77 degrees Fahrenheit." Fortress publishes its 18.43 kWh as a rated capacity at 0.5 C. Enphase publishes 5.0 and 10.0 kWh usable and explains in a footnote that the usable figure already contains a 2 percent safety critical limit plus a further 3 percent maintained for battery electronics overnight.

So the honest reading of any capacity headline is that it describes a new unit at room temperature, and the retention floor is the manufacturer's statement about where that number is allowed to go. Where no floor is published, the headline is the only capacity figure the maker has committed to, and it applies on day one.

Which clock runs out first for you

This is the only question that turns the two figures into a decision, and it depends entirely on how the battery will be used.

Backup-only duty cycles the battery rarely. A unit that sits at full charge waiting for outages might complete a few dozen full-depth equivalent cycles a year, which will never approach a throughput cap. For this owner the calendar is the binding clock and the retention floor is the number that matters, because the battery will still be sitting there at year ten.

Daily time-of-use cycling is a different machine. One full cycle a day is roughly 365 a year, which against Generac's 2,520 module cycles arrives in about seven years, and against FranklinWH's 4,000 equivalent cycles arrives inside the 15 year term with room to spare. For this owner the energy cap is the binding clock, and a headline year count overstates the coverage. Our load shifting guide works the daily arbitrage arithmetic through, including what the wear costs against the spread.

Grid services and VPP enrollment sit somewhere in between and add a question the other two do not: whether the utility's dispatch counts as an application your warranty covers on uncapped terms. That is a question for the manufacturer in writing before you enroll, not an assumption.

What none of this tells you

It does not tell you the battery will fail at the term. It does not tell you it will not. No manufacturer in our tracked set publishes a service life estimate, a year-by-year degradation curve, or a projection of capacity between year zero and the floor. Anyone showing you a smooth curve for one of these products drew it themselves.

What the documents do tell you is the shape of the promise: a percentage you are guaranteed to still hold, an amount of energy you are allowed to move, and a number of years, with the first of those three to arrive ending the coverage. Read all three, then check which one your own usage pattern reaches first.

For the full terms, the conditions, and who administers a claim on each model, see the warranty comparison. For the specific case of Tesla's two operating buckets and the internet connection requirement, see how long a Powerwall lasts. Every figure here is drawn from the manufacturer documents cited on each model page in our battery database, and manufacturers revise warranty terms between versions, so confirm the document in force for the unit you are quoted.