Outsourcing CNC machining: benefits and considerations

• 1
  What outsourcing CNC machining actually changes
• 2
  Benefit one: cost and capacity without capital
• 3
  Benefit two: geometry your own machines cannot reach
• 4
  Consideration: tolerance, inspection and who owns the risk
• 5
  Consideration: drawings, IP and the paperwork chain
• 6
  Which parts belong outside and which do not
• 7
  Outsource or keep in-house: a quick read

Outsourcing CNC machining means the spindle that cuts your part is not on your floor. Everything else stays the same: the same G-code logic, the same tool wear, the same thermal drift. What changes is who owns the machine, the setup time and the metrology.

That sounds trivial until you count the cost structure. A machining center only pays for itself when the spindle is turning. A job shop that runs 127 machines across three plants spreads setup, programming and fixture cost over far more hours than a captive department running two mills for one product line.

The gain is not magic. It is the arithmetic of spindle utilization plus access to machine types you would never buy for a single program. A 4,000 mm travel mill or a 16-station 5-axis cell only makes sense if someone else keeps it busy.

The loss is control. Once the part leaves your dock, you cannot walk over and check a first article. Every benefit below has a matching consideration, and the rest of this page pairs them.

The obvious benefit is avoiding capital. One simultaneous 5-axis machining center costs more than most engineering teams spend on tooling in a decade. Buying one for a program that runs four times a year means paying for idle hours, and idle hours still consume floor space, power and maintenance.

The second layer is quoting. An outside shop prices the operation, not the machine. If your part needs 12 minutes of 3-axis milling and 4 minutes of turning, a mill-turn center does it in one setup and the quote reflects one setup. Your internal cost model would have charged two departments.

The third layer is scale. Runs from one prototype to 10,000+ parts sit on the same process, so the fixture built for the prototype is the fixture used for the production run. No second capital request, no requalification of a machine you just bought.

Where it stops working: volumes above roughly 50,000 parts per year with a stable design usually justify a dedicated cell. At that point the outside shop's margin is larger than your amortized machine cost, and you should bring it back in-house.

Some features are not a matter of skill. They are a matter of travel and axis count. A part with undercuts on five faces, or a 4,000 × 400 × 150 mm envelope, needs a machine most companies do not own and cannot justify.

A 5-axis setup also removes the human error that comes from re-fixturing. Each time you move a part to a new vise, you add a datum shift. Three setups means three chances to be 0.02 mm off before the cutter touches metal.

Multi-axis work is where outsourcing pays back fastest, because the alternative is not a cheaper machine. The alternative is a redesign that removes the feature you needed, or a fixture that costs more than the machining.

One boundary: if the part fits in a 500 × 500 × 450 mm envelope with no undercuts, a competent 3-axis shop down the road will match any 5-axis quote. Do not pay for axes you do not use.

Tolerance is the first thing to settle in writing. A shop that holds ±0.005 mm on a 100 mm aluminum bracket may not hold it on a 600 mm titanium housing, because thermal expansion and tool deflection scale with size. Ask for the tolerance per feature, not per drawing.

Inspection is the second. A certificate that says the part was measured is not the same as a report that shows the numbers. For medical and aerospace work, ask which instruments were used, what the reference temperature was, and whether the report travels with the shipment.

Risk transfers when you approve the first article. If the first article passes on a cold Monday morning and the production run happens in August, the shop still owns the result, but you own the schedule. Build a buffer into anything that feeds a customer deadline.

Third-party inspection is available if your quality system requires it. Budget it early. Adding a CMM audit after the parts are anodized usually means stripping and re-machining, and nobody wants that conversation.

Sending a STEP file and a PDF drawing to an outside shop is a disclosure event. Even with a purchase order, most jurisdictions treat the exchange as confidential only if there is a written agreement. A signed NDA before the first upload is the cheap version of that protection.

Ask how the files are stored, who can open them, and how long they are kept after the order closes. A shop that answers this quickly has answered it before. A shop that improvises is a schedule risk as well as a legal one.

Logistics is the quieter cost. Air freight for a 3 kg bracket is fine. Air freight for a 400 kg casting is not. Work out the shipped cost before you compare quotes, and confirm which Incoterm the price uses, because the number on the quote is rarely the number on the invoice.

Customs classification matters for anything with a surface finish or an anodized coating. A part that leaves as raw aluminum and returns as a finished assembly can be reclassified, and the duty changes with it.

Good candidates share a profile: complex geometry, modest annual volume, a material you do not machine every week, and a schedule that can absorb a few days of transit. Prototypes, bridge tooling and low-volume production all fit that pattern.

Poor candidates are simple parts with high volume and a frozen design. If a bracket runs 80,000 pieces a year and has not changed in three years, the outside margin is pure cost. Bring it in or renegotiate hard.

A third category sits in the middle: parts that are simple but need a finish you cannot apply in-house. Outsourcing the machining and the anodizing together usually costs less than shipping semi-finished parts between two vendors.

There is also a capability question. If your team has never programmed a mill-turn part, the learning curve is real and it shows up as scrap. Paying someone who has run that cycle 400 times is cheaper than paying for your first 40.