Last month we quoted the same aluminum housing at four different volumes. Same drawing, same material, same tolerances. The only thing that changed was the quantity.
10 pieces: $58 each.
100 pieces: $23 each.
1,000 pieces: $8.80 each.
5,000 pieces: $6.20 each.
The 10-piece price is nine times the 5,000-piece price. And we were not losing money on either end of that spread. The margins were comparable.
If you are sourcing CNC machined parts, this cost curve matters more than your material choice, more than your tolerance callouts, more than which surface finish you pick. Quantity is the single biggest lever on your unit cost. Understanding what actually changes inside the shop at each volume tier is the difference between getting a fair quote and leaving money on the table.
The cost curve for one real part
Here is how the numbers break down for a specific part. An aluminum 6061-T6 mounting bracket, roughly 100 by 80 by 50 mm. A few pockets, some M6 tapped holes, ±0.05mm on the critical features. Nothing exotic. The kind of part that lands in every third RFQ.
| Cost driver | 10 pcs | 100 pcs | 1,000 pcs | 5,000 pcs |
|---|---|---|---|---|
| Programming and setup (per part) | $18.50 | $2.15 | $0.25 | $0.05 |
| Material | $6.80 | $5.20 | $4.10 | $3.60 |
| Machine time | $22.00 | $10.50 | $3.20 | $1.85 |
| QC and inspection | $5.50 | $3.20 | $0.85 | $0.55 |
| Tooling, packaging, overhead | $5.20 | $1.95 | $0.40 | $0.15 |
| Per-part total | $58.00 | $23.00 | $8.80 | $6.20 |
The exact numbers shift depending on geometry, material, and tolerances. But the shape of this curve repeats across almost every CNC part I have ever quoted. Setup costs dominate at low quantities and fade to nothing. Machine time peaks in the middle. Material becomes the biggest line item at the high end.
From 10 to 100 pieces, the unit price drops about 60 percent. From 100 to 1,000, another 60 percent. Between 1,000 and 5,000, you save maybe 30 percent more. After that, the curve goes flat. That is not a pricing strategy. That is physics.
10 pieces: setup eats everything
At 10 pieces, you are not really paying for the cutting. You are paying for everything that happens before the spindle turns on and after it stops.
Programming takes 2 to 3 hours for a medium complexity part, whether you are making 10 of them or 1,000. CAM programming, toolpath verification, post-processing for the specific machine — all fixed cost. Spread across 10 parts, that is $10 to $18 per part. Spread across 1,000, it is pennies.
Setup is another hour to 90 minutes. Indicating a vise, touching off tools, running the first article, measuring every dimension to confirm the program is right. Fixed cost, same math.
Material at this quantity gets bought at retail. You need a piece of 6061 plate big enough for 10 parts plus clamping stock, and you pay whatever the supplier charges for walk-in quantities. No volume discount on 30 kilos of aluminum.
Inspection is 100 percent. At 10 pieces, checking every single part is practical. But it means every part carries the full inspection burden.
What this means for your RFQ: at 10 pieces, do not negotiate on material cost. It will not move the needle. The setup is 35 to 45 percent of your unit cost. If you can simplify the setup — looser tolerances, fewer flipped operations, no awkward clamping surfaces — that cuts your price more than any line-item negotiation.
Also worth knowing: at this quantity, every shop is a prototype shop. Nobody builds dedicated fixtures for 10 parts. Nobody optimizes toolpaths for cycle time. The job runs the way the programmer thinks will work on the first try. This is fine. It is the right way to make 10 parts. Just understand that the same part at the same shop quoted at 1,000 pieces goes through a completely different process.
100 pieces: the sweet spot for small batch
The steepest drop in unit cost happens between 10 and 100 pieces. Three things pile up at once.
First, fixed costs amortize. That $250 to $350 of programming and setup labor now spreads across 100 parts. It goes from the biggest line item to a modest one — roughly 35 percent of unit cost down to about 8 percent.
Second, it is now worth building a real fixture. For 10 parts, you use soft jaws and move on. For 100, you spend an hour or two making a dedicated fixture. Maybe it holds four parts at a time. Suddenly your load and unload time per part drops from 90 seconds to 15. The cycle time itself might drop because the fixture is more rigid. A four-part fixture on a 100-piece order means the operator loads the machine 25 times instead of 100. That is the difference between a job that finishes before lunch and one that takes two days.
Third, the operator gets fast. This happens on every batch job. By part 15 the sequence is automatic. By part 50 they have found small optimizations the programmer never considered — a faster way to deburr an edge, a technique that gets the surface finish right on the first try. A 22-minute cycle becomes 18. Across 100 parts, saving 4 minutes each is nearly 7 hours of machine time that did not exist at quantity 10.
Material still does not get real volume pricing, but you are past retail rates. Suppliers give a modest break when you are buying enough for 100 parts instead of 10.
If you prototype with a shop and the samples are good, run the first production batch at 100 pieces with that same shop. The learning from the prototype run carries over. The fixture exists. The operator remembers the part. Switching shops to save 5 percent on the unit price usually costs more in re-learning than you save.
1,000 pieces: machine time takes over
At 1,000 pieces, programming and setup are free. Under 2 percent of unit cost. The cost structure has flipped.
Now the two big line items are machine time and material. Together they make up about 75 to 80 percent of the part cost.
Machine time per part drops from the 100-piece level. The fixture is fully optimized. The toolpath is proven. Maybe you invest in a form cutter that machines a profile in one pass instead of three stepovers, or an insert drill that removes material twice as fast as a standard twist drill. At 10 pieces custom tooling is a waste of money. At 1,000 it pays for itself.
Tool life starts to matter. Saving 30 seconds per cycle is worth 8.3 hours across 1,000 parts. That justifies better tooling — coated carbide over uncoated, premium grades that hold an edge longer, more frequent insert changes to maintain surface finish across the entire batch. Tool cost per part might go up while cycle time and scrap rate go down. Net cost per good part still drops.
Material purchasing shifts from buying for this job to buying for inventory. Shops with volume have distributor relationships that make 500 kg the minimum for tier pricing. They stock standard grades and thicknesses, which means material for a 1,000-piece order costs 15 to 25 percent less per kilo than material for a 10-piece order.
QC switches from 100 percent inspection to AQL sampling. Check 50 random parts out of 1,000 instead of every single one. Accept one or two defects within the AQL. Per-part inspection cost drops below a dollar.
The bottleneck changes. At 10 pieces, the constraint was whether the programmer had bandwidth this week. At 1,000, it is whether there are enough spindle hours on the VMCs this month. You are scheduling production, not fitting in a small job.
5,000 pieces and beyond: when to stop CNC machining
From 1,000 to 5,000 pieces, unit cost drops about 30 percent. From 5,000 to 20,000, maybe another 10 to 15 percent. The curve is nearly flat. You have squeezed out the setup cost, optimized the fixturing, negotiated the material, and dialed in the tooling. There is not much left to optimize on the CNC side.
This is when you ask whether CNC is still the right process.
Die casting starts to beat CNC above roughly 2,000 to 5,000 pieces, depending on the part. A mold costs $4,000 to $8,000, but once it is running, each aluminum part might cost $1.50 to $3.00. The crossover point depends on geometry. Thin-walled, simple shapes without internal features cross over early. Complex parts with deep pockets and tight tolerances might never cross over.
Extrusion plus CNC finishing works for long, constant-cross-section parts. Extrude near-net, cut to length, then CNC the features that cannot be extruded. The extrusion die is cheaper than a casting mold and the material properties are better. Good for aluminum enclosures, heat sinks, linear rails.
For small complex parts under about 50 grams, metal injection molding can produce near-net shapes at high volume. The upfront tooling is expensive but the per-part cost at 20,000 pieces and up is hard to beat with CNC.
One thing I see buyers get wrong: they calculate the process-switch crossover assuming the mold will amortize across multiple production runs. Then the design changes at 8,000 pieces. Run your crossover calculation assuming one production run. If the numbers still work, go for it. If they only work across three hypothetical runs, stay with CNC until those runs are confirmed.
Some parts should never leave CNC regardless of quantity. Deep internal pockets. Tight positional tolerances. Threaded features that need to be single-point cut. Internal undercuts. If your part genuinely needs what only CNC can deliver, do not force a process change because a spreadsheet told you to. Bad parts at half the cost are still bad parts.
What to do with your next RFQ
Send quantity tiers. 100, 500, 1,000, 5,000. Not just "please quote."
The relationship between those tiers tells you something useful about the shop. If the drop from 100 to 1,000 pieces is less than about 40 percent, the shop probably does not run much production work. They are pricing volume the same way they price onesie-twosies. That is the right shop for your prototype, not your production run.
If the 1,000-piece price is under half the 100-piece price, the shop has production infrastructure — multiple machines, dedicated fixturing, bulk material relationships. That is what you want for volume.
Also: do not optimize your design for 10-piece pricing and then expect that same design to be economical at 5,000. What reduces cost at low quantity is different from what reduces cost at volume. At 10 pieces, reducing setup operations is everything. At 5,000, adding an extra operation that eliminates a secondary process can be worth it even if it requires more programming. You would never pay for that extra programming at quantity 10. At 5,000, you definitely would.
One more thing. Send your drawing with your annual volume estimate. If you are planning 2,000 pieces this year, say so. The quote you get at known volume is different — and better — than the quote you get when the shop assumes you are ordering 20.
Send your STEP file and annual volume estimate for a free DFM review and tiered quote. We quote in four quantity bands by default — 100, 500, 1,000, 5,000 — so you can see exactly where your part lands on the cost curve. Orders over 500 pieces go through our production optimization path, which uses different fixturing logic and material sourcing than prototype work. Upload your drawing here.