“An EV battery loses half its capacity within a few years”

Define the loss before evaluating the claim

Battery capacity describes how much usable energy a pack can hold under specified conditions. A temporary winter reduction in available energy or driving range is not automatically permanent capacity degradation. Nor does a shorter highway range by itself reveal a failed battery. To test the claim, the same vehicle needs comparable health measurements over time, with measurement conditions and the distinction between usable and nominal capacity made clear.

A useful battery report states the health metric and when it was measured. It should distinguish permanent loss from the available energy on one unusually cold day. Those definitions make a reader’s comparison reproducible and prevent unlike dashboard screenshots from becoming an apparent contradiction in the evidence.

The latest study, with its date attached

Geotab’s January 2026 analysis covered more than 22,700 electric vehicles across 21 makes and models. It reported 2.3% average annual degradation. The company’s 2024 analysis had reported 1.8%, but these are different samples. Their difference cannot be read as the same batteries suddenly aging faster. The observed average is inconsistent with losing half the capacity within a few years as a typical result. ^[1]

A worked example is not a forecast

For intuition only, apply a simple hypothetical loss of two percentage points of original capacity per year. After five years, that arithmetic gives 90% remaining capacity, not 50%. This illustrates the scale of the claim; it is not a prediction for a real pack. Aging can be nonlinear, individual results vary, and a manufacturer’s health threshold need not match the point at which a particular driver finds the car unsuitable.

Capacity and range do not have a perfectly fixed relationship across unlike journeys. If energy consumed per kilometre changes, the same usable capacity can provide a different distance. Consequently, compare a health test with its own measurement protocol, and compare range on reasonably matched routes. A screenshot showing a dashboard estimate is an estimate of a trip outcome; it cannot, by itself, quantify permanent pack degradation. This distinction helps explain why anecdotes that look contradictory can both be accurately reported.

Habits and experimental methods can change the result

Geotab reported differing outcomes by charging behavior and climate, but an observational association should not be treated as a controlled experiment. Separately, the December 2024 dynamic-cycling research showed that constant laboratory discharge and realistic variable loads can yield different lifetime estimates. Together, the studies make a useful methodological point: identify how the battery was used and how its health was measured before extrapolating a headline. ^[1][2]

A Chinese fleet study published in Nature Energy in August 2026 analyzed field data from 116 NMC passenger cars and 17 LFP buses, with records spanning more than three years and some vehicles reaching 300,000 km. In a separate end-of-life analysis of 19 passenger cars, the mean weakest-cell constraint corresponded to a 6.2% reduction in usable pack health. This is a cell-imbalance metric, not capacity lost since new across the 116-car dataset or a fleetwide degradation rate. ^[3]

The useful ownership question

The practical question is whether the car still has adequate range, power and reliability for the owner’s needs. A battery-health certificate and a consistent service history are more informative about one used vehicle than an internet-wide average. Be equally skeptical of catastrophic-loss claims and promises that no modern battery ever needs attention. Neither extreme follows from a sample average, and neither replaces model-specific evidence.