How many watts is my MacBook using?

Updated 18 August 2026 · 436 completed measurement windows of about 90 seconds each, logged on one MacBook Pro (Mac15,10, M3 Max, 36 GB, macOS 26.5.2) between 16 August 11:12 and 18 August 09:49 · plus a two-machine audit from 14 August 2026

Short version: your MacBook does not have a wattage. It has a distribution. On the machine we logged most closely — a developer's laptop under agent and build load — the power leaving the battery ran from 6.74 W to 59.73 W over two days. One Mac, one owner, both ends of that range inside those same two days. Any page that answers this question with a single figure is answering a different question than the one you asked.

What can be answered is the shape: how wide the spread is, where the middle sits, what the number is made of, and how to read your own. That's what's below.

One machine, two days, 436 measured windows

Each window is about 90 seconds of battery discharge, measured from the battery's own counters and accepted only if the window passed validation. Every window in it came off one owner's working laptop under agent and build load — that is the first thing to know about the figures below, and whose machine that was spells out what it does to them. Here is the whole distribution:

The middle half of the time this Mac sat between 15.76 W and 31.95 W — and even that middle half is a wide band. Laid out in 5 W bins, the 436 windows fall like this:

Note what this is not: a bell curve around a typical value. There is a group down at the bottom of the range, a crowd in the high teens, another in the late twenties and thirties, and a thin tail above 40 W. What the shape shows is a machine moving between modes rather than settling on one value. Where its quiet mode actually sits is a question these logs can't answer: the display was awake for effectively all of this corpus, so even the lowest bin is a machine doing little, not a machine doing nothing.

Whose machine that was — read this before quoting the numbers

This is a developer's laptop under agent and build load, and it would be dishonest to let you assume otherwise. In 239 of the 436 windows the largest named consumer was Python; in another 76 it was a terminal-based client. That is not a typical day's work, and the distribution above is skewed upwards because of it.

Two more limits worth naming out loud: the display was awake for effectively all of this corpus, and it contains no video playback at all. So nothing on this page describes a Mac sitting still, and nothing on it describes watching a film. Those are real questions, and these logs simply cannot answer them.

Two Apple silicon MacBooks, side by side

On 14 August we compared logs from two Macs — a busy one and a quiet one. The averages are far apart, and what the power was spent on is further apart still:

Read the second row twice. On the busy Mac, most of the power went to things the owner started, so a list of apps in watts describes that machine well. On the quiet Mac, most of the power went to being switched on at all — and per-app watts describe a minority of what's happening.

The conclusion we drew in that audit, and still hold: two machines are far too few to call 71% typical — but they are enough to stop calling 32% typical. We don't know the population. We do know that the answer moves further with whose Mac it is than with which apps are open.

What the number is actually made of

Any watt figure for a Mac is two things added together: a floor that exists because the machine is awake — backlight, idle chip, memory, storage, radios — and the work you asked for on top of it. That's why the same app changes the total by different amounts on different machines, and why subtracting an app's watts from the machine's watts overshoots: you cannot land beneath the floor.

The second thing worth knowing is that the chip's own energy counters are not the battery's bill. The counters measure the rails of the chip; the battery additionally pays for power conversion, for memory and fabric outside those counters, for storage, radios and ports — none of which report anything. Regressing battery draw against measured chip energy over 565 completed windows on one machine gives a slope of 2.50 (R² = 0.59): one watt reported by the chip costs the battery about two and a half. That is a rate of change, not a ratio you can apply to a reading — the same fit carries a constant term for the part that doesn't move with the chip at all, so there is no factor to multiply a single chip figure by. What it does mean is that a monitor showing you the chip's number and calling it your power use is reporting a smaller quantity than the one your battery is paying.

One thing we deliberately don't print anywhere: the floor as a watt figure. Our own estimator for it saturates against its own ceiling on most windows, and a saturated estimator is a ceiling wearing the costume of a measurement. So we publish the floor only as a share of the total — which is what the table above does — and we'll publish a number when we can identify one.

That ceiling doesn't disappear when you turn the floor into a share, so read the percentages accordingly. The estimator can be clamped down but never up, and on the busy Mac it sat at that clamp for most windows — which makes 32% a lower bound, not a best guess: the floor there took at least that much, possibly more. So the two machines are closer together than that row makes them look. It leaves the conclusion we drew from them standing, because an under-reported floor is the direction that flatters a list of apps in watts, not the one that inflates the floor at its expense.

The watts on your charger are a different quantity

Many of the pages that rank for this question are about power adapters and power banks, which is worth untangling: the number printed on a brick is what it is able to supply, not what your Mac takes. A machine drawing well under its adapter's rating is normal; a machine that occasionally wants more than the adapter delivers is also normal, and then the battery quietly covers the difference while you're plugged in.

Lab figures from a review are a different quantity again. They describe that reviewer's unit, running that reviewer's test, with that reviewer's screen brightness — usually stated as one number, which the table at the top of this page should make you suspicious of. Ours are one owner's two days. Neither is your Tuesday afternoon.

How to see your own number

macOS itself won't tell you. Activity Monitor's Energy tab has an Energy Impact column, and Apple calls it "a relative measure, lower is better" and never names a unit — so it can't be turned into watts, into hours, or into a comparison between two Macs; we lay that out in how we count. There is no built-in screen that says how many watts are leaving your battery right now.

That is the gap WattMate fills, and the reason it exists: the top line of its menu bar panel is the whole machine in watts, right now, on battery; under it are the apps, each in watts, with the part that has no app behind it kept separate instead of being spread over your app list. Then it does the thing a published figure can't: it names how many minutes quitting the top app is worth, and two minutes after you quit it measures again and prints what actually changed — or says the effect couldn't be isolated, which happens and is worth being told.

If you'd rather do the arithmetic yourself, everything on this page came from the app's own database with Scripts/watts-distribution.py in the repository, which prints the quantiles and the histogram above and nothing else; the app is the same numbers without the SQL.

Questions people ask

So what is a normal wattage for a MacBook?

We can't give you one honestly, and we've now measured enough to explain why. On one machine the range across two days was 6.74 W to 59.73 W. Across two machines the averages were 27.2 W and 8.5 W. A single "normal" figure would have to hide all of that, and the number it hid would be wrong for most readers in the direction they care about.

Why does every site give a different number?

Because they're measuring different machines doing different things, and mostly reporting one moment of it. On the one machine we logged, a single moment could have been any value in the table above — which is why one moment is not an answer, whoever publishes it.

Do these numbers apply to my Mac?

The shape does; the values don't. The energy counters we read belong to the chip, not to the Mac model — on the second Apple silicon chip we examined, some of the channels present on this one don't exist at all, and neither the slope nor the floor was ever measured there. Treat the tables above as a worked example of the method.

Can I just read watts in Activity Monitor?

No. Energy Impact is a dimensionless score, and the best public explanation of its formula describes Intel processors this machine does not have. It's useful for ranking your own apps against each other in the moment, and useless for anything with a unit attached.

My Mac is plugged in. How many watts is it using then?

Plugged in, the battery's counters say nothing about power use — the question turns into an adapter-side one: how much is arriving, how much is running the Mac, how much is going into the battery. That's a separate measurement and a separate screen.

What these numbers don't prove

• One heavily loaded machine and one audit of two. Not a survey, not a population, not a model average.
• No idle stretches and no video in the corpus. Two of the most-asked-about cases are simply absent from these logs, and we're not going to estimate them.
• Nothing here transfers to another chip. The counter set belongs to the silicon; we've checked a second one and found it different.
• The floor is a share here, never a watt figure. Until it's identifiable, publishing it as a number would be inventing precision.

What does generalise is the method: read the battery, not the chip; keep the floor separate from the apps; and distrust any single number for a question whose honest answer is a distribution — including the ones on this page.

Related: does quitting apps actually save battery — mostly no, measurably yes once · how much battery Chrome uses, measured on two Macs · how we count: what Energy Impact means, and what a watt per app is