There is no single figure, because consumption is a flow rate and flow is a setting, not a property of the machine.
Two real cases from the estimator:
- A 60 mm cube at 0.2 mm layers and 60 mm/s takes 3.95 hours for 71.7 g — 18.2 g an hour.
- A kilogram of PLA laid down at 0.2 mm and 150 mm/s takes 21.8 hours — 45.8 g an hour.
Same material, same nozzle, two and a half times the throughput.
The equation behind it
Flow is layer height × line width × head speed. The first case demands 5.4 mm³/s; the second 13.5 mm³/s. Multiply mm³/s by 3,600 and by the density and you have grams per hour.
That is why the answer is a setting: change any one of the three terms and consumption changes proportionally, with no change to the printer at all.
The ceiling nobody mentions until they hit it
A hotend can only melt so much plastic per second. A mainstream 0.4 mm setup lands somewhere in an 8 to 25 mm³/s band; a 0.6 mm one in 15 to 40.
Command a speed above that ceiling and the printer does not obey — it slows down or under-extrudes. So there is a hard cap on grams per hour for any given hotend and nozzle, and pushing the speed slider past it changes nothing but the number on the screen.
What the figure is useful for
Planning spool changes. At 45.8 g an hour, a fresh kilogram lasts a shade under 22 hours of continuous printing. At 18.2 g an hour it lasts more than two days.
Sizing a farm's material budget. Machine-hours multiplied by grams per hour is the only sensible way to forecast filament spend, and it is far more reliable than counting parts.
Sanity-checking a slicer. If your slicer predicts a mass and a time that imply 90 g an hour on a stock 0.4 mm nozzle, one of the two numbers is wrong.
Work out the flow and the time for your own settings in the print time estimator.