Analysis of Powerwall Battery Retention - Part 2

Introduction

This is Part 2 of the Powerwall battery retention analysis. For an introduction, methodology, and the initial dataset, see Part 1.

Eight months after the initial analysis, we provide an update that includes:

• Significantly expanded dataset for both Powerwall 2 and Powerwall 3. Data come from anonymized statistics collected through Netzero, an app for managing home energy systems that also provides access to Powerwall diagnostic data.
• A longitudinal view that tracks individual units' capacity over time rather than a single snapshot.
• Observations related to the tail of Powerwall 2 production.

Powerwall 2 Update

We generated an updated scatter plot with manufacturing date on the X axis and nominal full-pack energy (capacity) on the Y axis, color-coded by hardware part number. The plot reaffirms the trends from Part 1 of our analysis. For an explanation of the late-2018 to 2019 outlier cluster, see Part 1.

The new data also show that Powerwall 2 manufacturing largely ended around November 2024. After that point:

• No serial numbers beginning with "TG1" appear. TG1 denotes Gigafactory Nevada, where most Powerwalls are built.
• Units observed after that date carry the older "1092170-xx-y" part number (blue points), matching the original Powerwall 2 revision.
• Serial numbers begin with "LTR" or "TAU", suggesting units were likely not built at Gigafactory Nevada.
• Capacity variance increases for these units.

These traits suggest that post-November 2024 units may be remanufactured or warranty-replacement units rather than new builds.

Longitudinal Analysis

The Netzero app allows tracking of capacity over time. In this plot we include only units with at least six months of historical capacity records. This enables us to follow individual battery trajectories and verify whether the trends from snapshot data persist.

Observations:

• Earliest units (1092170-xx-y) show little degradation in the period covered, with some units exhibiting almost no decline over a year.
• Some samples show temporary capacity recoveries. Short-term changes of this kind are expected and can reflect temperature effects, cell balancing, or BMS re-estimation rather than true capacity gain.
• The longitudinal view remains short relative to warranty horizons. We will continue to track these units over time.

Battery Retention by Discharged Energy

This plot shows capacity versus average yearly discharged energy (a measure of battery utilization). Compared to using lifetime discharged energy in Part 1 of the analysis, average discharge highlights the relationship between utilization and degradation earlier in the life of the battery.

This plot uses a sequential color scale, ranging from 0 MWh/year (purple) to 5 MWh/year (bright yellow) to represent average discharged energy. 5 MWh/year is the equivalent of a full Powerwall discharge (13.5 kWh) every day of the year.

The link between utilization and degradation is once again evident. Powerwalls with low utilization (purple) tend to sit at the top of the capacity cluster, while high-utilization units (bright yellow) tend to sit lower. The 2019 outlier cluster should be excluded when interpreting this trend.

Powerwall 3 Update

We now have a substantially larger dataset for Powerwall 3, though the time span is still limited.

Early observations:

• The initial capacity of Powerwall 3 is slightly lower than that of Powerwall 2, but the drop during the first year seems more gradual. Note the different Y-axis scale on this plot compared to Powerwall 2.
• We will continue to monitor these units over time to see whether their long-term behavior changes.

Contact

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