What is a print farm? What it takes to run one
Updated 31 October 2026
A print farm is a group of 3D printers run together as production, usually to make parts for sale. There is no official minimum number of printers, and most of the work is not printing but everything around it.
We are Well Earned GmbH in Thun, Switzerland. We sell motorcycle parts through our shop Revsticks and print many of them ourselves on seven printers. This page describes what that involves.
How many printers make a farm
No standard sets a number. SimplyPrint, which sells farm software, calls a farm "a group of 3D printers run in parallel as a business" and names no minimum. The same article advises starting with a single printer.
Our rule of thumb, and it is only ours: you have a farm once starting, tracking and clearing prints has become a job in itself. For us that job was about 45 minutes every morning, just to start prints on seven machines. That works out at roughly 6.4 minutes per printer.
The printers in our print room
We run a mixed fleet.
| Printer | Count | Network control we can use |
|---|---|---|
| Bambu Lab P1S | 3 | Yes |
| Bambu Lab X1C | 1 | Yes |
| Anycubic Kobra S1 | 1 | Yes, with Rinkhals firmware |
| Anycubic Kobra 2 Max | 1 | No |
| Anycubic Kobra X | 1 | No |
So five of the seven can be driven over the network with software we have working. The other two have no usable network control. Buying more printers of a kind you can already control is worth more than it looks on a spec sheet.
Network control on the Bambu machines comes with trade-offs, covered in LAN-only and Developer Mode.
What farms make
SimplyPrint's article names desk organisers, cosplay and prop parts, tabletop terrain, phone and gadget accessories, replacement parts and educational models.
We make motorcycle parts: licence plate holders, bar end caps, sprocket covers, brake line holders and more. The plates vary a lot. Our slicer predicts 6.20 g and 26 minutes for a brake line holder, and 205.44 g and 9 hours 16 minutes for one licence plate holder. Those two jobs need different handling, because a 26-minute plate cannot run unattended overnight in any useful way and a nine-hour plate can. The numbers are in cost per plate.
What the work really is
The printer does the printing. These are the parts that take your time.
Starting prints. This is the 45 minutes above. It is the most repetitive job in the room and the first one worth removing.
Keeping files in order. Our production folder holds 210 model files, 58 saved slicer projects and 50 slicer presets across two slicers. The same part often exists once per printer family. Our V3 plate holder has both V3 FULL BAMBU and V3 FULL ANY. Change the design and each of those needs redoing.
Clearing plates. A finished printer earns nothing until someone removes the parts and confirms the plate is empty.
Handling failures. A failed print costs more than its filament, because someone has to notice it, clear it and queue it again. Printers can also make this harder to automate. A Bambu keeps reporting FAILED until the next print starts, even after the error was dismissed on its screen. We hit this in our own driver: the queue waited for the printer, and the printer waited for a job. Details are on the Bambu page.
Checking generated files. We once built test files by script and left out the slicer's include templates. All 76 files ran generic start G-code, with no AMS load and the nozzle fixed at 205 °C. An ASA test print started at 205 °C on a P1S before we caught it. If software makes your files, check the start G-code before it reaches the fleet.
Filament and maintenance. Spools run out, and machines need servicing. We do not have a measured failure rate or maintenance figure to publish, so we will not quote one. Be careful with any figure that does not state the printer and the material.
What you need
| Item | Why it matters |
|---|---|
| Printers you can control the same way | Each extra control method is extra setup and extra things to debug |
| One tested slicer profile per printer and material | Our two slicers needed separate presets |
| Fixed IP addresses for networked printers | A changed address looks like a broken printer, as in no Moonraker connection |
| A separate network for the printers | Bambu's Developer Mode leaves its control channels open, and securing them is your job |
| Ventilation | See the NIOSH findings below |
| A log of plates, hours and failures | Without it you cannot work out your real cost per plate |
The honest downsides
The labour does not go away. More printers means more plates to start, clear and check. Software can remove some of it, but on our machines someone still has to clear every plate.
Filament cost hides the real cost. Our worked example shows labour taking the largest share of every plate, at researched Swiss prices. Replace them with yours before you trust any margin.
You depend on vendors' firmware decisions. In January 2025 Bambu put an authorisation step in front of third-party control, then added Developer Mode as the way around it. Bambu does not support that mode. A rule like this can change after you have bought the machines.
Printers put things in the air. A 2017 NIOSH evaluation of a manufacturing workplace found that printers using PLA and ABS emitted volatile organic compounds and small particles. Measured personal exposures were well below applicable limits, and NIOSH did not recommend controls at that time. It also wrote that 3D printing "has not been thoroughly evaluated to determine its safety". Two findings matter for a farm:
- The manufacturer's loose-fitting doors and covers did not stop emissions reaching the room.
- A custom ventilated enclosure cut particle concentration by over 99% and total organic chemicals by almost 70%.
That is one site with its own printers, so treat it as a reason to plan ventilation and not as a measure of your room.
We also build Rackrew, software for running a print farm.