Every power station spec sheet says "up to 89% efficient" or "92% inverter efficiency" — and every runtime you actually experience comes in a little short of the spec-sheet math. That gap is this article: where the watts go, what the efficiency number really measures, the part of the curve no one prints, and the idle drain that quietly eats your low-load runtime.

If you've ever wondered why a "1,000 Wh" station delivers ~850 Wh, why a 2,000W inverter running a 30W CPAP feels wasteful, or what "96% efficient" actually means at your load — this is the full explanation, with the numbers.

TL;DR:

● Inverter efficiency =AC watts out ÷ DC watts in. The rest becomes heat — everywhere, always

● A modern pure sine wave inverter runs~87–96%, peaking near 50–75% of its rated load— and falling hard at both ends

● The famous~0.85 factor in every runtime calc is the end-to-end number: battery chemistry, DC-DC, inverter, and cable losses stacked. No-load drain

● (5–15W for a mid-size inverter) matters more than anyone budgets for — at light loads it's the difference between 8 h and 5 h of CPAP

● The practical rule: an inverter at 80–90% load is where it's coolest, most efficient, and longest-lived — the same number the sizing guides keep landing on

What the Number Measures (and Doesn't)

Efficiency is a ratio: Eff = P(AC out) / P(DC in). An inverter rated 92% at 1,000W AC output is pulling ~1,087W from the battery — and dumping ~87W as heat into the room.

Three things the single marketing number hides:

1. It's a point on a curve, not a property of the machine. "92% efficient" at 70% load tells you almost nothing about 10% load or 100% load — both are substantially worse.

2. It's AC-side only. The battery's own DC-DC buck/boost stage, the BMS, and the cables each take another bite before the inverter even starts. The end-to-end "what actually reaches my device" number is lower — which is where the 0.85 comes from.

3. It assumes a temperature. Efficiency is measured at ~25°C. A station in a 40°C garage runs hotter, throttles earlier, and lands a few points lower than the datasheet.

Where the Losses Actually Live

Inside a pure sine wave inverter, the DC→AC conversion happens in a bridge of switching transistors (MOSFETs) driven at tens of kHz. Every stage takes its cut:

The shape that falls out: a U-curve. At very light load, the fixed switching and control costs dominate a small signal → efficiency drops. At very heavy load, conduction losses (∝ I²) climb → efficiency drops again. The valley between is the sweet spot: ~50–75% of rated load, where modern units hit 93–96%.

That U explains two "mysteries" at once: why a big inverter feels wasteful running small things, and why sizing with a 10–20% headroom (our [inverter sizing guide]) isn't conservatism — it's putting the machine on the hilltop of its own curve.

The Efficiency Curve, in Real Numbers

A typical 2,000W pure sine wave unit, measured:

Read it two ways. First, the runtime way: at 1,000W output you're losing ~50W (5%); at 200W you're losing ~40W — 20% of what the device actually receives. Same absolute-ish heat, wildly different tax. Second, the no-load way: even at 0W output, the control electronics + display + BMS draw 5–15W (1–3% of rating) to stay ready. That "idle" line is the sleeper in every light-load calculation — more on it below.

(A modified sine wave unit runs a similar curve, typically 1–3 points lower at the peak, which is part of why the cheaper box also wastes more — one more reason in our [waveform guide] to treat it as the exception, not the option.)

The No-Load Drain: The Runtime Killer Nobody Budgets For

This is the section that changes how you run light loads. A 2,000W inverter at no-load pulls ~10W. Run it to power a 30W CPAP for 8 h:

●The CPAP wants 240 Wh

●The conversion + control overhead  at that light load is ~5–15W of extra draw →~1–4 h of "invisible" drain depending on the unit's light-load curve

●The idle tax: if the inverter stays on through the quiet hours (normal station mode), the standby 5–15W × 8 h =another 40–120 Wh— up to 4 more hours of CPAP, gone, for doing nothing

Now you see why the [CPAP guide] pushes two things so hard: the 12V DC feed (skips the inverter stage entirely — the ~10–30% runtime difference is literally this section), and the low-load/CPAP mode (shuts the inverter down when the load drops below ~50W, so the standby tax stops accruing overnight). Those features aren't convenience; they're the no-load curve, deleted.

The rule for any light-load setup: the smaller and lighter your load relative to the inverter's rating, the more the idle and switching overhead matters — and the more a dedicated low-load mode (or a DC path) is worth.

What "85%" Really Means in Your Runtime Math

Every runtime calculation in this library — the [Wh calculator] and beyond — uses usable = rated Wh × 0.85. Here's the stack that 15% represents:

So "1,000 Wh station runs a 100W lamp for 8.5 h" isn't a rounding choice — it's four named layers of physics. If your inverter sits near its 50–75% hilltop and the ambient is cool, you'll land closer to 0.90; if you're running a 30W load on a 2,000W unit at 35°C, you'll land closer to 0.75. 0.85 is the median honest number.

Five Ways to Cut the Losses (in order of impact)

1. Match the load to the machine. A 500W load on a 2,000W inverter is ~5% overhead; on a 1,000W unit it's ~2%. Right-sizing (the [sizing guide]) is the single biggest efficiency win available — it's free.

2. Use the DC path when the load is DC. 12V fridge, USB-C laptop, 12V fan — every one of these that skips the inverter skips the entire AC column of losses.

3. Kill the standby. Low-load/CPAP mode, or simply power the inverter/station off when the only "load" is a router you can run on its own 12V adapter. The idle 5–15W × hours adds up faster than any other single line.

4. Keep it cool. Efficiency drops and thermal throttling starts as internal temperature climbs; ventilation and shade are worth 2–5 points in a hot garage.

5. Don't chase the 100% point. Running pinned at rating is less efficient than at 75%, and ages the unit faster — the 10–20% headroom in every sizing formula exists to keep you off that right edge of the U.

How to Read the Spec (So "96%" Doesn't Fool You)

●"Peak efficiency 96%" = the hilltop, near 50–75% load. Fine — but it's not your average.

●"Efficiency 89%" with no load stated = assume it's a mid-load point; ask for the curve if the purchase is a big one.

●No-load current (often printed in mA or W) = the standby tax; 1–3% of rating is normal, above 5% is a weak design.

●Temperature rating (e.g., 0–40°C operating) = the envelope where the number is true; outside it, expect degradation.

Frequently Asked Questions

How efficient are modern inverters really? Modern pure sine wave units run ~87–96%, peaking near 50–75% of rated load. End-to-end (battery to device) is lower — plan on ~85% in runtime math, which is the 0.85 factor everywhere.

Why does my station's runtime come in short of capacity ÷ load? Four stacked losses: battery chemistry/BMS (~2–4%), inverter conversion at your actual load (~4–10%, worst at light and heavy loads), cables (~1–2%), and rated-vs-delivered cell capacity (~3–5%). That's the ~15% the 0.85 factor represents.

What is no-load consumption and does it matter? A mid-size inverter draws 5–15W with nothing plugged in, just to stay ready. It doesn't matter for a 1,000W grill run; it defines the economics of an 8-hour 30W CPAP — where the standby and light-load overhead is the difference between hours, and why 12V DC feeds and low-load modes exist.

Does a modified sine wave inverter waste more power? Typically 1–3 efficiency points at the peak, plus the harmonic current that makes downstream devices draw more. The efficiency gap is modest; the compatibility and heat gaps (our [damage guide]) are the bigger story.

Does heat reduce inverter efficiency? Yes — the spec is measured near 25°C, and every 10°C above that costs a few points and ages the unit. Ventilation, shade, and a hard surface (the [safety guide] rules) are also efficiency rules.

Final Thoughts

An inverter isn't a wire — it's a conversion with a toll booth at every stage, and the toll is heat. The single number on the box is a point on a U-curve, not a property; the light-load and full-load ends are where the tax gets steepest; and the no-load line, which almost nobody budgets, is the one that quietly decides your CPAP nights and your router's all-week runtime. Once you see the curve, the design rules stop being folklore: size with headroom to sit at 50–75%, route DC loads around the inverter, kill the standby, and keep the machine cool. Do that and the "85%" in every runtime formula stops feeling like a mystery rounding error — it's just the sum of four named, beatable losses, and you're paying the small end of each one.

[Optional CTA: NEJoye stations publish both the peak-efficiency point and the no-load draw in the spec sheet — the two numbers that predict your real runtime → NEJoye Power Inverters]