Original research · published 2026-07-31

The Capacity Gap Report

Every power station prints a watt-hour number on the box. Almost none of them tell you how much of it reaches your appliance — and after auditing 80 models, we found the industry cannot tell you either.

Nobody measures this the same way, so nobody can tell you the number

We set out to publish a simple figure: across the catalogue, what share of rated capacity actually reaches a load? We could not. Of 80 models, 21 carry any published capacity measurement at all. Of those, 9 turned out to be rated capacity multiplied by an assumed efficiency, or a figure the source itself flagged as informal. That leaves 12 real measurements — 11 of them on AC — from 3 outlets, and only 3 of the 12 published the load they tested at.

Among the AC measurements the delivered share runs from 75.5% (Anker SOLIX C800 X) to 97.7% (Bluetti Elite 100 v2) — a spread of 22 percentage points. That spread is not a ranking of manufacturer honesty. A battery worked harder gives back less, and the lowest figure here was taken at 1200 W — the heaviest load anyone in this set published. The highest did not state a load at all, which is the more common case: the only test loads published anywhere in this dataset are 50 W, 60 W, 1200 W. So we cannot tell you whether the 22-point spread reflects the hardware or the test bench, and neither can anyone else.

Which is where “you get about 85%” comes from. It is repeated across manufacturer blogs as though it were a measurement. It is a plausible midpoint of numbers that were never comparable — and against the measurements that do exist it lands nine points high at one end of the range and thirteen points low at the other.

The 11 AC measurements we could verify

Delivered energy through the inverter, against the manufacturer’s rating. Where a source published a range, the range is shown — narrowing it would be inventing precision.

ModelRatedDeliveredShare of ratedTest loadSource
Jackery Explorer 300 Plus

Discharged at a 50 W load — among the lightest test conditions in this set.

288 Wh253 Wh87.8%50 WTrusted Reviews
Jackery Explorer 2000 v22,042 Wh1,710 Wh – 1,740 Wh83.7% – 85.2%not publishedOutdoorGearLab
EcoFlow DELTA 2

Drawn from a six-month ownership test rather than a single bench run.

1,024 Wh890 Wh – 900 Wh86.9% – 87.9%not publishedPortable Power Lab
EcoFlow DELTA Pro 34,096 Wh3,790 Wh92.5%not publishedOutdoorGearLab
Bluetti AC1801,152 Wh950 Wh82.5%not publishedOutdoorGearLab
Bluetti AC200L

The source reports the spread as load-dependent, which is the whole problem in one row.

2,048 Wh1,770 Wh – 1,900 Wh86.4% – 92.8%not publishedTrusted Reviews
Bluetti Elite 100 v21,024 Wh1,000 Wh97.7%not publishedTrusted Reviews
Bluetti Apex 3002,765 Wh2,550 Wh – 2,580 Wh92.2% – 93.3%not publishedTrusted Reviews
Anker SOLIX C800 X

Extracted over 29 minutes at 1200 W — the heaviest load in this set, and the lowest share.

768 Wh580 Wh75.5%1200 WTrusted Reviews
Anker SOLIX C10001,056 Wh860 Wh81.4%not publishedOutdoorGearLab
Anker SOLIX C2000 Gen 2

Published as a percentage (~86% on AC); the Wh figure here is that share of rated.

2,048 Wh1,761 Wh86.0%not publishedTrusted Reviews

Measured on DC, and why it sits in its own table

A USB or 12 V figure never passes through the inverter, so it skips the single largest loss in the chain. Putting it in the table above would flatter the unit and mislead you.

ModelRatedDeliveredShare of ratedTest loadSource
Anker SOLIX C300

Measured over USB at 60 W. This never passes through the inverter, so it is not comparable to the AC rows.

288 Wh269 Wh93.4%60 WTrusted Reviews

The 9 claims we threw out, and why

These are shown rather than dropped, because the reason each one fails is the substance of this report. Most of them are the 85% rule of thumb wearing a measurement’s clothes.

  • Jackery Explorer 1000 Plus

    ~1074 Wh usable at ~85% (Backup Power Hub)

    Rated capacity multiplied by an assumed 85% efficiency. That is the rule of thumb, not a test of it.

  • Jackery Explorer 2000 Plus

    ~1736 Wh calc at 85% (Backup Power Hub, computed not bench-measured)

    The source states outright that the figure is computed rather than measured.

  • EcoFlow DELTA 3

    ~870 Wh effective at ~85% (Backup Power Hub, calculated)

    Calculated from rated capacity at an assumed efficiency.

  • EcoFlow DELTA 3 Plus

    ~870 Wh effective at ~85% (est., same cells as DELTA 3)

    Estimated by analogy to a different model. No measurement of this unit exists.

  • Bluetti AC240

    ~1,300-1,679 Wh (StorageReview / Trusted Reviews)

    The upper bound exceeds the 1,536 Wh rating once conversion losses are accounted for, so the two sources are not measuring the same thing.

  • Bluetti Elite 200 v2

    ~2,074 Wh full-rated delivered (Trusted Reviews)

    Equal to the 2,073.6 Wh rating. AC conversion cannot be lossless, so this restates the spec.

  • Anker SOLIX F2000

    Delivered <half rated capacity under very light load — full-load figure not published

    No specific figure, and a result that surprising needs a protocol we can read before it means anything.

  • Pecron E1000LFP

    ~600-700 Wh used in an 8-hr laptop+CPAP+router test with >30% left

    A field test that never ran the pack down, so it bounds the capacity rather than measuring it.

  • VTOMAN FlashSpeed 1000

    ~700-750 Wh usable estimated after AC conversion loss (Battery Skills, informal)

    Recorded by its own source as an informal estimate.

What this report cannot tell you

We own no hardware and ran none of these tests. Every figure above was measured by someone else and is credited to them; our contribution is the audit, the exclusions, and the refusal to average numbers that are not comparable. 59 of the 80 models in our catalogue have no published capacity measurement of any kind, and for those we can tell you nothing — which is why no capacity-derived figure appears in the BatteryRank score.

If you test power stations and publish your load conditions, we would rather cite you than estimate. The gap this report documents closes one published protocol at a time.

Where the numbers live