Analysis of Powerwall Battery Retention

Update on Oct 21, 2025: Part 2 of the Powerwall battery retention analysis.

Powerwall is a rechargeable lithium-ion home battery, manufactured by Tesla. It stores energy for backup power, solar self-consumption, and time-of-use load shifting^[1]. Powerwall 2 entered mass production in 2017.

Like most lithium-ion batteries, the capacity of a Powerwall gradually diminishes due to its aging and usage (also known as calendar and cycle degradation^[2]). Although various Powerwall 2 warranty versions exist, one common variant ensures 70% of the 13.5 kWh rated capacity for 10 years following its initial installation.^[3]

Information about the real capacity and degradation of a Powerwall battery is not available from the Tesla app. However, this information can be accessed by connecting directly to the device for diagnostics. This analysis utilizes anonymously gathered statistics from Powerwall owners via Netzero, an app for managing home energy systems that also provides easy access to Powerwall diagnostic data.

Background Information

Powerwall Models

The Powerwall 2 Datasheet^[4] lists the relevant specifications:

Usable energy of 13.5 kWh.
Total energy of 14 kWh.

Powerwall 2 is believed to use Nickel-Manganese-Cobalt (NMC) battery cells.

Note that there are other Powerwall models:

Powerwall 1 was introduced in 2015 with 6.4 kWh capacity. Now discontinued, it is not included in this analysis.
Powerwall+ combines Powerwall 2 with an integrated solar inverter. The part numbers for the underlying battery match Powerwall 2 part numbers, so Powerwall+ is treated the same as standalone batteries for this analysis.
Powerwall 3 is the latest generation of batteries. It became available in 2023, so we don't have sufficient data yet. This analysis focuses on Powerwall 2, but we included a preliminary analysis for Powerwall 3 at the end.

Diagnostic Data

By connecting directly to a Tesla Gateway unit on the local network^[5], diagnostic data can be retrieved from each installed battery. The data collected for this analysis includes:

Manufacturing date: Since the installation date of a Powerwall is not readily available, the manufacturing date is used instead as a proxy. The manufacturing date can be inferred from the battery's serial number. For example, a serial number starting with TG123131 carries the following information:
- TG1: Stands for Tesla Gigafactory 1 (also known as Gigafactory Nevada), where most Powerwalls are manufactured.
- 23: Manufactured in 2023.
- 131: Manufactured on the 131st day of the year: May 11, 2023.
Part number: This indicates the hardware revision of a Powerwall, e.g. 3012170-05-E. There are over 30 distinct Powerwall 2 part numbers. For the purpose of this analysis, part numbers are grouped by the first part (e.g. 3012170) for readability.
Nominal full pack energy: The battery management system's estimate of the current full capacity, reported in Wh.
Lifetime energy charged and discharged: This is a measure of the battery utilization, reported in Wh. Discharged energy is used for this analysis, since this metric is also relevant for the warranty.

Methodology

A random sample of 2,000 Powerwall 2 batteries was selected, along with the manufacturing date, part number, nominal full pack energy, and lifetime energy discharged. Powerwalls were limited to the ones manufactured in Tesla Gigafactory 1 (serial numbers starting with TG1). This excluded around 4% of samples, and was done to avoid outliers (e.g. refurbished batteries).

Some random noise was introduced into the manufacturing date, nominal full pack energy, and lifetime energy, to prevent individual battery identification. This random noise does not impact the observations or conclusions.

Analysis

Battery Retention by Manufacturing Date

The first scatter plot shows samples with manufacturing date on the X axis and nominal full pack energy (capacity) on the Y axis. The X axis is reversed to show the most recently manufactured Powerwalls on the left. The oldest Powerwalls in the sample set are from the end of 2017, making them 7 years old as of 2025.

Initial observations:

Powerwalls start with an initial capacity that is higher than the 14 kWh total energy mentioned in the specifications. There is some variation in the initial capacity based on the manufacturing date.
Degradation is more pronounced in the first 4 years but stabilizes thereafter, with many Powerwalls within 90% of the rated 13.5 kWh capacity, 7 years in. However, a distinct outlier cluster from late 2018 through 2019 exhibits higher degradation.
A confounding factor for estimating degradation is the variation in the initial capacity of Powerwall batteries, which may have changed over the years.

Further analysis is required to understand this outlier cluster by incorporating part number (hardware revision) as an additional dimension.

Battery Retention by Part Number

The scatter plot, now color-coded by part number, reveals that the high-degradation cluster consists of Powerwalls with part numbers starting with 2012170, manufactured between 2018 and 2020.

With a clearer understanding of the time degradation and outliers, the next analysis examined battery utilization.

Battery Retention by Discharged Energy

In this scatter plot, the third dimension is replaced with lifetime discharged energy, a measure of battery utilization. Discharged energy varies based on the number of cycles and the depth of each discharge. For example, a Powerwall owner who keeps their battery at 100% backup reserve—e.g. for grid outage protection—will have very low utilization and minimal discharged energy; on the other hand, keeping a low backup reserve and cycling the battery daily—e.g. for solar self-consumption—can result in high utilization. The highest discharge values in the sample set are at around 35 MWh. This amounts to over 2,590 cycles (assuming 100% depth of discharge at the nominal 13.5 kWh capacity), or a full cycle every day for 7 years.

This plot uses a sequential color scale, ranging from 0 MWh (purple) to 33 MWh (bright yellow) to represent lifetime discharged energy.

As expected, the plot shows that the Powerwalls with higher utilization also have higher degradation. For example, in the cluster of Powerwalls manufactured at the end of 2017, the difference in capacity between a low-utilization Powerwall and a high-utilization Powerwall is around 2 kWh.

There is no evidence of higher utilization for the outlier cluster with high degradation.

Powerwall 3 Battery Retention

Powerwall 3 launched in 2023 with higher power output and a built-in solar inverter. The capacity specifications are similar to Powerwall 2: 13.5 kWh nominal capacity^[6]. Powerwall 3 is believed to use Lithium iron phosphate (LiFePO 4 or LFP) battery cells, which have longer cycle life compared to NMC battery cells^[7].

A comprehensive analysis of Powerwall 3 battery retention requires more historical data. However, Tesla's modifications to Powerwall 3 have made diagnostics access more challenging^[8], potentially limiting data availability compared to Powerwall 2. The scatter plot below shows a random sample of 500 Powerwall 3 batteries, analyzed using the same methodology as Powerwall 2.

Initial observations:

The initial capacity of Powerwall 3 is slightly lower compared to Powerwall 2, but more consistent over time.
Further data collection is required to establish definitive degradation trends.

Conclusions

This analysis provides insights into the long-term performance and degradation trends of Powerwall 2 batteries. It presents evidence of reasonable performance consistency 7 years post-manufacture. However, some manufacturing batches exhibit higher-than-expected degradation. The battery utilization plot confirms the expected link between higher usage and increased degradation.

Future research directions include:

Longitudinal study tracking the capacity of Powerwall batteries over an extended period to validate the trends observed in this analysis.
Exploration of additional dimensions such as ambient temperature.
A future study on Powerwall 3 battery retention once sufficient historical data is available.

References

Contact

For questions or comments contact Ziga Mahkovec at ziga@netzero.energy.