← The Signal Report Work with me

Report 252 · Energy Storage

Why battery capacity fade overstates lifespan

Battery cell spec sheets and most research papers track aging in amp-hours. What you actually get from a battery is watt-hours. A Sandia-led study compared the two across 12 lithium-ion datasets. In gentle use the gap is small. In harder use, still within the manufacturer's limits, it takes 15% or more off the cell's useful life.

The paper is Preger, Wittman, Harris and Dubarry, "Are Capacity and Energy Loss Equivalent Metrics for Battery Aging Reporting?", published online 13 March 2026 in the ASME Journal of Electrochemical Energy Conversion and Storage (May 2026 issue). The authors are from Sandia National Laboratories, Lawrence Berkeley National Laboratory and the University of Hawai'i, and the paper is open access. Sandia's August news item summarized it as finding "up to a 15% error in energy storage system lifetime predictions." That is accurate, but the number is easy to misread, and the full paper says more than the summary does.

Amp-hours and watt-hours are not the same thing

Capacity, in amp-hours (Ah), counts how much charge a battery moves. Energy, in watt-hours (Wh), is that charge multiplied by the voltage it was delivered at. The paper puts it simply: energy "is derived from integrating capacity with voltage." So if a battery's voltage sags as it ages, it delivers less energy for the same charge, and its energy fades faster than its capacity.

Capacity became the standard for a historical reason the authors spell out: it "was simpler to measure, requiring only an ammeter to gauge current and a clock for time." Measuring energy meant integrating voltage too, which was harder before computer-controlled testers. That excuse is long gone, but the habit stayed.

The paper's cautionary example is a lithium-manganese-rich cathode that looked excellent on capacity. Pre-commercial cells kept "91% capacity retention after 300 cycles," but "the energy retention was only 60%," because the voltage kept dropping. That chemistry is an extreme case. The authors say they do not expect commercial cells to show anything like it. The study asks the smaller, more practical question: how big is the gap in ordinary, commercial lithium-ion cells?

What they did

The team pulled 12 public cycling datasets covering four cathode families: LCO, NMC, NCA and LFP (the chemistry I covered in Report 012). Cycling temperatures ran from −5 °C to 45 °C, charge rates from 0.2C to 6C and discharge rates from 0.2C to 4C (1C means a full charge or discharge in one hour). The authors stress that "All cells were cycled within their respective manufacturer-defined limits." For each cell they tracked the ratio of energy retention to capacity retention, which works out to the same thing as the change in average discharge voltage.

One methodological choice matters. They used the everyday cycles, not the slow check-up tests labs run periodically, because those slower tests "may mask kinetic effects." In other words, measuring a battery gently can hide part of what it loses in real use.

What they found

Energy almost always fades a little faster than capacity. In the paper's words, "discharge energy almost always fades faster than discharge capacity, irrespective of chemistry and cycling conditions." There are exceptions: some NMC cells went the other way, with the ratio rising to 104%.

In moderate use the gap is small. The conclusion says that for cells cycled under moderate conditions, energy loss at end of life was "less than 1% greater" than capacity loss. For most cells in their main example dataset, the cycle count to 20% loss differed by less than 1%.

In harder use it is not small. Three NMC cells cycled in the cold (−5 °C, still within spec) hit 20% energy loss while showing only about 17% capacity loss, reaching that point 40 to 70 cycles early. The authors' arithmetic: a 3-point gap against a 20-point end-of-life line "translates to a 15% decrease in the cell's lifetime." Deeper discharge and higher rates had the same effect in other datasets.

The number people will misread

"15%" does not mean a battery holds 15% less energy than its capacity rating suggests. In the cold-cycled cells, the battery had 83% of its capacity and 80% of its energy at the same moment. That is a 3-point difference. It becomes 15% only when you measure it against the battery's useful life, which the industry usually ends at 80% remaining. The authors make this point themselves: the gap "seems trivial," until you set it against that 20% window.

"Up to 15%" is also not the ceiling in the paper. Applying the same arithmetic, the authors note that several NCA cells hit 20% energy loss at only 17% capacity loss, and fast-cycled LFP cells at about 15%, which they say indicates "a 15–25% reduction in the useful life of the cells." The 25% end comes from LFP cells charged at 3.6C to 6C, far harder than the 0.5C to 1C LFP cycling in another dataset in the same paper, which stayed much closer to even. The honest summary is a range: under 1% in mild use, 15% to 25% in the harshest cases they looked at.

What this means for a home battery

First, the limits. These are single lab cells, not home systems, and the paper does not test any home battery. Where a real installation falls on that range depends on the drivers the paper identifies: temperature, depth of discharge and rate. A battery in a cold garage that is routinely run close to empty and drawn hard is on the wrong side of all three.

Second, the reassuring part. The paper notes that "For battery spec sheets at the module-level and above, aging is typically presented in terms of energy retention." The system you buy is usually rated in kilowatt-hours. The amp-hour habit mostly lives one level down, in cell data sheets and in the lifetime and cost models built from them. Their conclusion: "Technoeconomic models based on Ah-capacity metrics may systematically overestimate battery lifetime." If a seller's lifetime or payback figure was built up from cell cycle-life data, ask which metric it used.

Third, what to check. When a warranty promises "70% capacity after 10 years," look for the unit. Is it kilowatt-hours delivered, or an unlabeled percentage? And as I showed in Report 150, the health number in your app is itself an estimate. For how fast real home batteries have actually faded in the field, see Report 205, and for what happens when the warranty line is crossed, Report 144.

A note on my angle: I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator, run by fellow veterans. I do not own that company and earn nothing from this link. Full policy here. Tracking energy, not just charge, is exactly the kind of thing that work cares about, but nothing in this report involves them, and the paper's authors report no conflicts of interest.

The bottom line

Capacity fade is a reasonable proxy for a battery used gently, and an optimistic one for a battery worked hard. The fix the authors recommend costs almost nothing: report aging "in terms of energy loss rather than capacity loss (or present both capacity and voltage loss)." Until cell data sheets do that, treat any lifespan figure built on amp-hours as the best case.

Sources

  1. Y. Preger, R. Wittman, S. J. Harris and M. Dubarry, "Are Capacity and Energy Loss Equivalent Metrics for Battery Aging Reporting?" Journal of Electrochemical Energy Conversion and Storage 23(2): 021109, May 2026 (published online 13 March 2026), DOI 10.1115/1.4071224, CC BY 4.0. Full text via OSTI 3030680. (Primary source, full paper read; supplemental figures not read. Source for the 12 datasets and their chemistries, temperatures and rates; the within-limits statement; the LMR-NMC 91% versus 60% example; the use of standard cycles rather than reference tests; the 99.5% to 92% and 104% ratios; the cold-cycled cells at 17% versus 20% and 40 to 70 cycles early; "less than 1%" in moderate use; the 15% and 15–25% useful-life arithmetic; the module-level spec-sheet and technoeconomic-model statements; the recommendation; and "no conflicts of interest." All quotes verbatim.)
  2. Sandia National Laboratories, DOE Office of Electricity Energy Storage Program, "Research: Using energy fade, not capacity fade, can more accurately predict how long energy storage systems will last," 10 August 2026. (Read. Source for the "up to a 15% error" summary quoted verbatim and the DOE Office of Electricity funding.)
  3. M. Dubarry, S. J. Harris and co-authors, "A mechanistic investigation of the relationship between battery energy and capacity," Joule, 2026, DOI 10.1016/j.joule.2026.102540. (Abstract only, found via Consensus; not read in full and not relied on for any claim above. Listed because it is the same group's follow-up, which models why the energy-to-capacity ratio changes.)
  4. Prior reporting in this publication: Report 150, Report 205, Report 144. (Context only.)
Onur Oncer
Onur Oncer

U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.

← All reports