How Much Does CNC Machining Cost?

• 1
  Key takeaways
• 2
  What actually decides how much does CNC machining cost
• 3
  Where geometry quietly adds hours
• 4
  How tolerance and surface finish move the number
• 5
  How quantity changes the answer
• 6
  Material choice and the scrap you never see
• 7
  Step by step: how to build and reduce a CNC cost estimate
• 8
  How design choices shift cost and lead time

Every quote we build sits on four pillars: raw material, machine and programming time including setup, labor and overhead, and secondary operations such as anodizing or plating. A supplier reads your CAD file and STEP model, then estimates each pillar separately. If a quote arrives as one lump number with no breakdown, you cannot tell whether the shop is expensive or simply quoting a different process plan.

Material cost is the easiest to check. Bar stock, plate, and near-net forgings are priced per kg, and the price you pay includes the material removed as chips. A bracket cut from a 100 × 100 × 50 mm block when it could be cut from 100 × 100 × 20 mm plate wastes both metal and spindle time. Our quoting team looks at stock size before anything else, because shrinking the billet often saves more than negotiating the rate.

Machine time splits into programming, setup, and cycle time. Programming is a one-time engineering cost that scales with geometry complexity, not part count. Setup is also one-time per operation, and its cost is mostly the fixture and the operator's attention. Cycle time is the only part of the pillar that repeats for every unit, which is why it dominates high-volume work and setup dominates one-offs.

Labor and overhead cover the operator, the tooling consumed, coolant, power, and the floor space the machine occupies. This pillar is fairly stable across a project, but it becomes visible when a part needs multiple operations on different machines. Every extra machine in the routing adds a new setup, a new queue, and a new chance for a datum error.

• 1
  MaterialAlloy, stock form, and how much of the billet becomes chips.
• 2
  Machine timeProgramming, setup, and repeating cycle time.
• 3
  Labor and overheadOperator hours, tool wear, coolant, power, floor space.
• 4
  Secondary operationsHeat treat, anodize, plating, laser marking, inspection.

A part with five orthogonal faces and open pockets is cheap. A part with undercuts, deep ribs, or a face that must be reached from three directions is not. Those features do not raise the material price, they raise the number of setups, and each setup consumes fixture time, alignment time, and inspection time.

This is where simultaneous 5-axis machining changes the math. On a 5-axis center with a Ø400 mm rotary table, an angled face, a port, and a mounting pad can often be cut in one setup instead of three. One setup means one datum, so the ±0.005 mm tolerance holds across features without stacking errors between operations. It also means the part sits in the queue once, not three times.

Not every part benefits. A flat plate with holes drilled from one side is faster on a 3-axis machine, and putting it on a 5-axis center only adds programming overhead. The rule we use: if the part has 2 or more features that are not reachable from a single spindle direction, 5-axis usually wins. If everything faces one way, it usually does not.

Thin walls are a separate problem. Below roughly 1 mm wall thickness, cutting forces deflect the workpiece, so the programmer reduces radial engagement and feed rate, adds semi-finish passes, and sometimes leaves material for a final spring pass. The toolpath gets longer and the part gets more expensive, even though the geometry looks simple on screen.

• 1
  Undercuts and side portsUsually force extra setups unless a 5-axis can reach them in one.
• 2
  Deep pocketsLong-reach tooling must run slower and deflect more.
• 3
  Walls under 1 mmLight radial cuts, extra finishing passes, higher scrap risk.
• 4
  Tight corner radiiSmall end mills remove less material per pass and break more often.

Tolerance is not a single dial. A drawing that says ±0.005 mm on every dimension is treated differently from one that calls ±0.005 mm only where a bearing seats and leaves ±0.1 mm elsewhere. Our shops hold ±0.005 mm (±0.0002 in) where it is specified, but applying it to non-functional faces adds finishing passes, more probe time, and more parts held for inspection.

Surface finish works the same way. As-machined surfaces sit around Ra 1.6–3.2 μm. A high-finish requirement of Ra 0.8–1.6 μm usually needs a semi-finish and a finish pass with a fresh tool. Fine finishing at Ra 0.2–0.8 μm often means a separate operation, slower feed, and sometimes a polishing step. Each step adds time that appears in the quote as a higher hourly rate or a longer cycle.

Inspection follows the tolerance. A general dims part can be checked with calipers and a height gauge. A ±0.005 mm part with true position callouts needs a CMM, and that inspection is charged as time, not as a flat fee. Every feature you add to the critical list adds measurement time on every unit, so keep the critical list short and defendable.

The practical advice is to tolerance by function. Put the tight tolerance on the mating feature, give the rest of the part room, and note datums clearly. This one change often removes a finishing operation and a CMM program from the routing.

At one piece, the quote is mostly programming and setup. The cycle time of a single part is a small share of the total, and the shop still has to build a fixture, prove the program, and inspect the first article. This is why a prototype can look expensive per unit.

As quantity rises, setup and programming are spread across more parts, so the per-unit price drops toward the cycle time plus material plus overhead. Around 50-200 pieces, the curve flattens for most machined parts. Beyond that, the savings come from better fixturing, tool-life management, and sometimes a dedicated second operation rather than from the original setup.

The mistake we see most often is assuming that a high-volume price applies at low volume. It does not, because the fixed work does not shrink. If you need 10 parts, budget for setup; if you need 10,000, budget for material and cycle time and pay attention to scrap rate.

There is a middle path. We run no minimum order quantity, from one prototype to 10,000+ part runs, so the same process plan can be quoted at both ends. Ask for a price break table at 1, 10, 100, and 1,000 pieces. That table tells you where the setup stops dominating.

Material price per kg is only half the story. The other half is how much of that material leaves as chips. A part machined from a billet 30 mm larger than the finished envelope can waste more than 50% of the purchased metal, and every kilogram of chips also consumed spindle time to remove.

Aluminum 6061 and 6061-T6 are the default for machined parts because they cut fast and hold tolerance well. Stainless 303 and 304 machine reasonably; 316L is tougher and slower. Steel grades such as 4140 and 4340 need more rigid setups and slower parameters. Titanium Ti-6Al-4V and Inconel cut slowly, wear tools, and often need coolant-through tooling.

Plastics behave differently. POM and PA machine cleanly with sharp tooling and air blast, while PEEK and carbon fibre reinforced grades are abrasive and expensive. Carbon fibre also raises a health and safety requirement, so it is quoted with dust control in mind.

When you compare two quotes for the same part, compare the material callout first. A quote using 6061 and another using 7075 are not the same part, even if the drawing looks identical, because 7075 is stronger, more expensive, and slightly harder to finish.