What a 3D print actually costs you, every input measured

Here is the finding, before the working.

The two inputs people argue about hardest — the electricity price and the infill percentage — are the two that move a print's cost least. The two nobody puts in a spreadsheet — how much the machine costs per hour it runs, and how often a print is scrapped — together move it more than moving house to Germany would. And once you count the twenty minutes you spend at the printer, every one of those terms is a rounding error against your own time.

This post costs one specific part all the way through, with every input stated so you can disagree with any of them and re-run it.

The part: a 120 mm wall bracket in PETG

A wall bracket for a cordless drill. Footprint 120 mm square, 40 mm deep, printed flat on the plate in PETG with four perimeters, 25% infill, 0.2 mm layers, 150 mm/s, a 0.4 mm nozzle. It is a boring part on purpose: nothing about it is unusual, and it is roughly what a functional print looks like when someone actually needs one rather than wanting a nice photo.

The print-time estimator puts it at 5.03 hours across 200 layers, laying down 185,883 mm³ of plastic. Treat that as a planning figure: the model carries a stated ±30% band against a real slicer, because a slicer has the toolpath and this has summary geometry. It is good enough to tell you whether this is an afternoon or an overnight job, which is what a cost estimate needs.

The plastic: 236 grams, and the density that decides it

PETG has a published density around 1.27 g/cm³, so a kilogram of it occupies 787.4 cm³. The bracket's 185,883 mm³ therefore weighs 236 g — about 77.26 m of 1.75 mm filament off the reel. Both of those come from the length and weight converter.

At $24 per kilogram, mid-range for PETG at retail, that is $5.66 of plastic.

Notice what the density did there. Run the same geometry in a filled grade — matte, glow, metal-filled — and the mass moves without the model changing at all, because you are buying volume and paying by mass. That is not a footnote; it is the single most common reason a weight estimate comes out light, and the cost per gram figure people carry in their heads is usually the one for whatever they printed most recently.

The electricity: ten cents, and the argument it does not deserve

The bracket runs for 5.03 hours on a machine that meters around 110 W while printing. That is 0.553 kWh, and at the United States national average of 17.5 cents per kilowatt-hour it costs $0.10.

Two things about that figure.

First, the wattage is not a power-supply rating. It comes from repeatedly reported plug-meter readings for that model, and the printer pages say which basis each machine's figure rests on, because for most models nobody has published a meter reading and the number is inferred from the machine's class instead. An estimate that renders like a measurement is worse than a wide range that admits what it is.

Second — and this is the part worth arguing about — the electricity price is the input people obsess over, and it cannot move this print much. The regional rate table runs from 9 US cents per kilowatt-hour to 41 cents. Costing the bracket at both ends gives $8.01 and $8.21. The whole world, from the cheapest rate on the site to the dearest, is worth about twenty cents on this print. If you want the arithmetic in full, the electricity method has it; the longer argument about why the term gets forgotten is a separate piece.

Those rates were compiled in August 2026 from published national-average residential surveys, and they are averages: a time-of-use tariff or a solar export credit will put your household somewhere else entirely. Your own bill beats every figure here.

The machine: the term with the widest honest spread

A printer wears out. Nozzles, belts, plates, fans, hotends and eventually the machine itself all have finite lives, and every one of them is consumed by printing hours rather than by calendar time.

I have used 30 cents an hour here, which over 5.03 hours is $1.51 — more than fifteen times the electricity. That number deserves less confidence than the others in this post, because it depends on what you paid, how long you expect the machine to last and how much of the maintenance you do yourself. The honest position is that it is the input with the widest legitimate spread, and there is a page that argues for a figure rather than asserting one.

What is not defensible is leaving the line out. Of every figure you could enter on this row, the one you can be sure is incorrect is nought — and it is the figure nearly every hobby estimate uses. If you would rather not commit to 30 cents, commit to something — a nozzle replacement alone is not free, and it is the cheapest thing on the list.

Consumables get their own line: adhesive, isopropyl, a share of a build plate. 15 cents a print here, small and real.

The failure allowance: not the same as the failure rate

If eight prints in a hundred are scrapped, the surcharge is not 8%. You need more than one attempt per finished part, and the expected number of attempts is 1/(1−f), so an 8% failure rate multiplies the exposed cost by 1.087. On this bracket that is $0.65. The derivation and its assumptions live on the failed-print method page; the point here is only that the multiplier is bigger than the rate, always.

Push the rate to 20% and the surcharge becomes $1.85, taking the print to $9.27. Read that against the electricity paragraph above: one attempt in five going wrong costs six times what the entire span of world electricity prices does. Nobody prices their failures and everybody prices their kilowatt-hours. There is more to say about that — chiefly that when a print fails matters as much as how often.

Adding it up, and the figure most people quote instead

Now the comparison that matters. Count only the two terms a slicer hands you — filament and electricity — and the same bracket comes to $5.76. That is the number most people quote, and it is short by nearly a third. Everything missing from it is missing for the same reason: nothing on the printer's screen ever displays it.

You can re-run the whole thing with your own spool price, your own rate and your own failure rate. Every figure above is that page with those inputs.

Then there is you

Slicing, loading the plate, watching the first layer, removing the part, cleaning up a bit of stringing, and putting it away. Call it twelve minutes, which is optimistic for a part you have not printed before.

At $20 an hour — not a wage, just a number that says your evening is worth something — that is $4.00, and the bracket becomes $12.07.

The $5.76 that most people would have quoted is under half of that. Not thirty percent light. Half.

This is the uncomfortable part of costing prints honestly, and it is why the labour line sits outside the headline figure rather than inside it: for a hobby print you are genuinely allowed to value your own time at zero, and lots of people do, happily. But you cannot value it at zero and be surprised that printing for other people does not pay. If any of this is going to be sold, the pricing piece is where the argument goes next, and resale pricing has the margin mechanics.

The inputs are not independent: the 0.6 mm nozzle case

Everything above treats the terms as a list. They are not a list; they are coupled, and this is the bit a per-input page cannot show you.

Print the identical bracket with a 0.6 mm nozzle at 0.3 mm layers instead of the stock 0.4 mm at 0.2 mm. Three things change at once: the lines get wider, the usable layer height goes up, and the hotend's flow ceiling rises. The estimator drops to 2.68 hours — but the extruded volume climbs to 201,122 mm³, which is 255 g rather than 236.

So the machine time falls to $0.80 and the electricity to $0.05, while the filament rises to $6.12. Total: $7.75.

Half the print time, and thirty-two cents saved. The time saving is real and worth having — it is half a day of machine availability — but it does not show up as money, because the money moved sideways from the clock into the plastic. Anyone who tells you a bigger nozzle saves cost is describing the term they happened to look at. The nozzle-size argument takes this further; the nozzle-choice guide covers what you give up in detail.

The same coupling runs through every setting. Raising infill adds plastic and hours. Adding walls adds plastic and hours too, but far less of both — which is why wall count is the cheap way to add strength and infill percentage is the expensive one. Supports add plastic, hours and cleanup minutes, and the cost of supports is the one term that hits all three at once.

Ranked by leverage

For this part, at these settings, in rough order of how much attention each input repays:

• Your own time, if you are counting it at all. Nothing else comes close, and it barely varies with anything you change in the slicer.
• Grams, which means geometry and infill and walls before it means spool price. Orientation and part design move this more than shopping does.
• Print hours, because they buy machine wear and, if you are honest, some of your attention. Layer height and nozzle are the big handles; raw speed is not, because the hotend runs out of melt before the motors run out of headroom.
• Failure rate, which is a process problem wearing a cost problem's clothes. Halving it is worth more than any purchasing decision on this page.
• Spool price. Real, but bounded. Across every material catalogued here, the dearest listing is roughly double the cheapest for the same polymer — so shopping well is worth a few dollars a spool, not a transformation.
• Electricity. Last. Genuinely last, for a single machine on a hobby duty cycle. Run four printers commercially and it re-enters the conversation.

That ranking is specific to a five-hour PETG part on a mid-sized machine. A twenty-hour print in a heated chamber shifts hours up and grams down; a tray of miniatures shifts everything toward labour. What does not change is the shape: the terms that scale with time are the ones people forget, and the terms that scale with mass are the ones they measure.

What this figure does not include

Being clear about the edges is part of the estimate.

• Purge and prime. Single-material, so it is trivial here. On a four-colour print it is not, and a purge tower can out-mass the part.
• Supports. This bracket prints flat and needs none. Reorient it and that changes.
• Post-processing. Sanding, threading inserts, painting. All labour, none of it in the model above.
• The estimator's error band. ±30% on print time propagates into machine cost and into any labour you tie to the clock.
• Wattage provenance. This machine's figure is metered by the community. Many are not, and the failure to say which is how a class estimate ends up quoted as a measurement.
• Whether you should have printed it at all. Sometimes the answer is a stamped steel bracket off a hardware-shop peg, and comparing the two honestly is a legitimate use of this arithmetic.

The same question — how much material, at what price, plus everything the estimate quietly left out — turns up whenever anyone quotes a job from a drawing. GetTheAmount does the home-improvement version of it.

Re-run it with your own numbers

Nothing above is a rule of thumb. Every figure is the cost calculator and the time estimator with the inputs printed next to the answer, which means you can change one and watch what happens rather than trusting the ranking.

Two suggestions for what to change first. Put your real electricity rate in, once, so you never have to wonder about it again. Then put in an honest failure rate — not the one you would tell someone, the one your last twenty prints support. The second number will move your costs more than the first, and it is the one you can actually do something about. If you are not sure where to start on that, the material selector is often the real answer: a great many failure rates are a material choice that was never a choice.