Home backup · US 120 V
Best power station for a window air conditioner
Ranked by a calculation, not an opinion. Gate one: this unit held near-continuously, at the compressor’s own 6,900 VA rating — not the label’s.
Our pick
Yes, it runs it — no proven runtime see why below ↓
No proven duration for the 12 h target
4 of EcoFlow’s own published documents, read 2026-08-08, carry no usable capacity for the US 120 V build.
Derived by WattPairverified 2026-08-0812 US records method
EcoFlow DELTA Pro 3 — the cheapest record that clears every constraint for this set. No duration is claimed for it: we read 4 of EcoFlow’s own published documents on 2026-08-08 and found no usable capacity for the US 120 V build in them, and a nameplate divided by a load is not a runtime. $2,799, price checked 2026-08-17.
Said plainly, because the panel above cannot: this record’s own 4,096 Wh nameplate puts an absolute ceiling of 5.4 h on this set — short of the 12 h this page ranks against, and that is before any depth-of-discharge or inverter loss is taken off it. It is the recommendation because it is the only record in this library that clears this load at all, not because it can be shown to run it for 12 h. Nothing on this page reaches that target on this set. A cheaper record with this same property would not be named under this pick at all; the pick is held to the standard it holds them to.
Usable energy · not found in this brand’s published sources
This set, 12 h · draw × duty cycle × hours
The buy link is in the bar at the bottom of the screen — one action per page, and it goes to the merchant we last checked this price at.
Every row opens the calculator, pre-filled, each figure below sourced — re-run any number yourself.
How this list is ordered
Four gates, in this order: does it run the set, can we prove how long, can it be bought today, then price. No score and no weighting.
The four gates in full
Applied in this order to every published US record. There is no score and no weighting — each gate either separates two machines or hands them to the next one.
-
1 Does it run the set?
Every record is put through the compatibility engine against all of the loads at once — running watts against continuous output, the worst start-up instant against the surge ceiling, waveform where a load demands one. Green first, then the tight fits, then anything the engine could not read, then the machines that cannot do it. A machine that cannot start the load in front of it is not a cheaper option.
-
2 Can we prove how long?
A runtime is only printed where the brand publishes what it takes to compute one — a usable-capacity coefficient, whether stated as depth of discharge and inverter efficiency or as one combined figure — and never from the number on the box. Records that reach the target sort first, records with a proven but shorter figure next (longest first: inside that band none of them does the job, so the one that gets closest wins), and records whose duration cannot be proven at all sort last inside their verdict band. That is a statement about published data, not about the hardware, and each of those rows says so.
-
3 Can it be bought today?
A record under a verified out-of-stock policy, or one we may not sell into your state, sorts behind an equivalent one that can be bought. It stays on the list with the reason attached rather than disappearing. Whether WattPair happens to hold a purchase link for a record is not part of this gate and never moves it: a machine we cannot link to is ranked exactly as if we could, and simply carries no buy button.
-
4 Then price.
Cheapest first, inside everything above. One mains standard and one currency per page, so every comparison is like for like. A record with no verified price sorts to the back of its own band — the absence of a price is not a low price.
The loads this ranks against
Why this set, and where its figures come from
A single window or room air conditioner, on its own — the load, not the room it is cooling. This is also the one appliance in this library where the manufacturer nameplate structurally cannot answer the surge question: UL 484, the safety standard plug-in room air conditioners are certified to, does not require a locked-rotor rating on the unit label at all. What this page checks against is the compressor’s own data sheet instead, one level down from the nameplate a reader could otherwise check by hand.
- Running
- 833 W
- Start-up
- 6,900 W
- Duty cycle
- 90%
Derived from official coefficients · verified 2026-08-03 · medium confidence · start-up basis: Nameplate figure
Duty cycle read off this record’s “Commercial setting, outdoor 95°F” profile, which carries the same 90% its default does — the profile is named so the reading is traceable, not because it overrides anything. The figure assumes near-continuous outage duty (90%), a scenario assumption — not a typical-use average or a manufacturer figure.
- Running together
- 833 W
- Worst start-up instant
- 6,900 W
- Average draw
- 750 W
- Energy for 12 h
- 8,996 Wh
Why 12 hours
Twelve hours, to match the site’s home-backup default. Which duty cycle an outage should assume for a window unit was, until this page’s own data was checked, an open question rather than a missing figure — this site now assumes the conservative, near-continuous reading (not the seasonal average an EnergyGuide label uses), and that is why an hour figure for a window air conditioner is computable here at all. It changes less than it sounds: the machines that publish enough to carry that number all fail this load’s electrical check on their own, and this site prints no duration for a machine that will not start the load — so no row below states one. The record this page does recommend carries no duration for the ordinary reason — we have found no depth of discharge or inverter efficiency published for it — not because of anything about this load.
Derived by WattPairverified 2026-08-03appliance records method
All 12 US records, ranked
1 of the 12 US records we publish run this set, 7 publish enough for a runtime to be proven, and 9 can be bought today. The percentage is the set as a share of that machine's continuous output; past 82% a verdict drops out of the green.
- 1 EcoFlow DELTA Pro 3 EcoFlow · US 120 V 4,000 W continuous · 8,000 W surge ceiling 4,096 Wh nameplate · Official spec · verified 2026-08-01 Runs it21%
- 2 Jackery Explorer 2000 Plus Jackery · US 120 V 3,000 W continuous · 6,000 W surge ceiling 2,042.8 Wh nameplate · Official spec · verified 2026-08-01 Too big115% of surge ceiling
- 3 BLUETTI Elite 200 V2 BLUETTI · US 120 V 2,600 W continuous 2,073.6 Wh nameplate · Official spec · verified 2026-08-01 Too big265% of assumed surge ceiling
- 4 Jackery Explorer 1000 Plus Jackery · US 120 V 2,000 W continuous · 4,000 W surge ceiling 1,264.6 Wh nameplate · Official spec · verified 2026-08-01 Too big173% of surge ceiling
- 5 Jackery Explorer 1000 v2 Jackery · US 120 V 1,500 W continuous · 3,000 W surge ceiling 1,070 Wh nameplate · Official spec · verified 2026-08-01 Too big230% of surge ceiling
- 6 BLUETTI Elite 100 V2 BLUETTI · US 120 V 1,800 W continuous · 3,600 W surge ceiling 1,024 Wh nameplate · Official spec · verified 2026-08-01 Too big192% of surge ceiling
- 7 BLUETTI Elite 30 V2 BLUETTI · US 120 V 600 W continuous 288 Wh nameplate · Official spec · verified 2026-08-01 Too big139%
- 8 Jackery Explorer 300 Plus Jackery · US 120 V 300 W continuous · 600 W surge ceiling 288 Wh nameplate · Official spec · verified 2026-08-01 Too big278%
- 9 EcoFlow RIVER 3 Plus EcoFlow · US 120 V 600 W continuous · 1,200 W surge ceiling 286 Wh nameplate · Official spec · verified 2026-08-01 Too big139%
-
Photo shows the EU 230V version — page specs are US 120V. 10 EcoFlow DELTA 3 Plus EcoFlow · US 120 V 1,800 W continuous · 3,600 W surge ceiling 1,024 Wh nameplate · Official spec · verified 2026-08-01 Too big192% of surge ceiling - 11 EcoFlow DELTA 2 Max EcoFlow · US 120 V 2,400 W continuous · 4,800 W surge ceiling 2,048 Wh nameplate · Official spec · verified 2026-08-01 Too big144% of surge ceiling
- 12 EcoFlow STREAM Ultra EcoFlow · US 120 V 800 W continuous 1,920 Wh nameplate · Official spec · verified 2026-08-01 Too big104%
Why 1 record carries no printable runtime
1 of the 12 records here starts and holds this load set and still carries no runtime this page will print, for one reason: we have found no depth of discharge and no inverter efficiency published by EcoFlow for it, so the hours a reader actually wants cannot be computed from anything but the number on the box — and dividing a nameplate by a load is the one thing this site will not do. The other side of that is BLUETTI and Jackery: they are the only brands here that publish both, which is the whole reason every proven runtime above belongs to them. The rule rewarded disclosure, not hardware.
What their rows do carry is the bound: that record’s nameplate over this set’s average draw of 750 W, which is arithmetic anyone can redo against the summary above — floored, never rounded up, so the printed ceiling never claims an hour more than the number supports — and it is printed on every one of those rows. But it is a ceiling and not a duration: it assumes the pack surrenders every watt-hour printed on it and that the inverter loses nothing turning DC into AC, and no battery does either. The real hours sit somewhere below that ceiling, and the two missing figures are exactly what would say how far below. That is the difference between a number this page will print and a number it will not.
The scope of that finding, stated: as of 2026-08-08 we read each manufacturer’s own published documents for these 1 models and found no depth-of-discharge and no inverter-efficiency figure in them. It is a statement about what those documents contained on that date — not a claim that the number exists nowhere. Send us one that is published and the record moves, and the row moves with it. Two counts are in play on this page and both are right: for 5 of the 12 records in this pool we have found neither figure published, which is the number the summary above is the other side of, and the 1 named here is the one that got far enough up the ranking for it to cost it a position. This is a statement about published data, not about the hardware — the verdict column judges that separately, and it comes first.
The load that changes the answer
Add a full-size refrigerator to the same set and the first gate does not move: 1 of the 12 still clears it outright, against 1 on the set above. The reason is one instant: a full-size refrigerator asks for 286 W at start-up on its own — 5.2× its own running draw, against 6,900 W for everything on the list above put together. That start-up figure is a category envelope: we have found no published start-up rating for this unit itself. The fridge does not cost this list anything further, because whatever was already going to fail here failed on the air conditioner alone — the one machine left standing has more continuous output to spare than most of this library has in total. The quiz carries this set as a preset.
start-up figure: Derived from official coefficients · verified 2026-08-01 · low confidence
The other rankings
Where to go next
Questions people ask about this list
How many of the 12 actually run it?
Which size of unit is the 6,900 VA start-up figure actually from?
Pricing, ranking and how the site is funded are not questions about this list — they are the same on every page here, and they are set out in full in the price policy and how the money works.