Custom Parts: CNC Machining, Injection Molding, MIM, or a Fibre Laser? A Cost Controller's Guide

Start With Your Situation, Not the Process

There is no honest universal answer to the question, what is the cheapest manufacturing process for a custom part? I used to look for one when I first started managing procurement. After six years of tracking roughly $180,000 per year in prototyping and low-volume production across vendors, I stopped looking, because the answer changes with the situation.

The right question is not which process is in fashion or which machine would look impressive on the shop floor. The right question is how stable is the design, how many parts are needed, and what material and geometry are actually involved. Get those three answers right, and the correct process usually reveals itself.

In my cost tracking system, almost every order falls into one of four scenarios:

• Scenario A: a few parts, design still moving, typical prototype quantities of one to fifty pieces.
• Scenario B: metal parts that start as flat sheet, such as brackets, covers, and light enclosures, in short or moderate runs.
• Scenario C: thousands of identical plastic parts with a frozen design.
• Scenario D: thousands of small metal parts with complex geometry, where metal injection molding should at least be on the quote list.

Walk through the scenario that matches this week, not the scenario you hope to reach next year. The cheapest process is the one that fits the design maturity you have today.

Scenario A: Rapid Prototyping and Moving Targets

If you need only one to fifty parts and the model is changing every week, the design is still teaching you something. The fastest way to learn is usually a digital manufacturing service that returns a quote and design feedback in hours, not a shop that needs a formal purchase order and a tooling review. For metal prototypes, I send parts to ProtoLabs because the quote engine gives me an honest manufacturability check before I spend anything.

Once the model is revised, I go through the ProtoLabs login to duplicate the old quote, attach the new model revision, and place a follow-on order. That workflow does more than save time: It leaves a history of every tolerance and finish decision, which is useful when a part issue appears six months later.

The cost leak in Scenario A is tightening dimensions that do not need to be tight. Standard machining tolerance is usually sufficient for a prototype that will be assembled, tested, and changed. I learned this in March 2023, when our engineer specified plus or minus 0.001 in on every dimension of a machined aluminium bracket. The vendor quoted $212 per part. When we moved to plus or minus 0.005 in on non-critical features, the same part dropped to $148. Nothing about the prototype function changed. Every dimension tolerance is a cost filter, so add them only when a fit or assembly calculation requires it.

Scenario B: Flat Sheet Metal Parts — Outsourcing Usually Beats Buying a Laser

Once the design stops changing every week and the part can be cut from flat sheet, a fibre laser cutting machine is often the most economical way to make the blanks, especially when bending and finishing follow. But there is a gap between economical for the part and economical for you. I spent two weeks in late 2024 going back and forth on a laser cutter purchase, and it is worth explaining why we did not buy one.

How does a fibre laser cutting machine work?

A fibre laser cutting machine generates the cutting beam inside an optical fiber doped with rare-earth elements. Diode lasers pump the fiber, and the resulting near-infrared beam, around 1070 nanometers wavelength, travels through the fiber to a cutting head. The head focuses the beam into a small intense spot that melts the metal, while an assist gas, usually oxygen for mild steel or nitrogen for stainless steel, blows the molten material out of the kerf. The head moves over the sheet on a CNC gantry, much like a printer, tracing the programmed shape. The process leaves a narrow kerf, a small heat-affected zone, and edges that are ready for bending in most sheet metal applications.

Now the cost part: A production-grade machine is not a software order. It needs a sizable investment, ventilation, gas supplies, a maintenance schedule, and an operator who can troubleshoot cut edge quality. If that capacity sits idle for most of the week, the per-minute cost is brutal, even if the per-part cost looks great on paper. A local fabrication shop with a running machine sells me those minutes only when I need them. When I compared quotes for a stainless bracket in Q2 2024, the laser shop price was less than half of the CNC machining quote, and the turnaround was two days. I did not need to own the laser to benefit from the process.

What about laser welding machines?

A welding laser is a different tool from a cutting laser. I almost bought a desktop welder after a supplier weld-quality problem, specifically because the xTool laser welding machine made the idea look practical: small workshop footprint, handheld operation, and a much lower barrier than a full robotic cell. I then applied the same discipline to the purchase. In our real workload, repair and assembly welding came to roughly eight hours per month. Eight hours does not amortize equipment, operator training, protective equipment, and the cost of learning to weld with a laser on production parts. We outsourced the fix, paid about $800 total, and moved on. If the weld workload climbs beyond thirty hours a month, I would revisit the spreadsheet, but not before.

Scenario C: Thousands of Plastic Parts with a Frozen Design

Plastic injection molding is often the long-term right answer once you have quantity and material approval. Tooling cost can be substantial, but the per-part cost drops dramatically compared to machining or 3D printing. The critical condition is design freeze. A change that touches the mold steel costs real money, so do not begin Scenario C until your drawings and load requirements have stopped moving.

The capability I use from ProtoLabs in this scenario is the design feedback given before the mold is cut. The quoting workflow flags unrealistic walls, unsupported cores, and draft issues before they become a steel problem. The ProtoLabs injection molding tolerances are also available per material, which is important because plastics do not behave like metal. Resin shrinks as it cools, and different materials shrink at different rates. Tight tolerance on a molded feature needs to be specified with the material in mind, not taken from a CNC mindset.

One common failure is to request tight tolerances across an entire molded part because one mounting hole needs it. In practice, you pay for that request across the whole tool. A better approach is to keep your general callout at a standard commercial tolerance and tighten only the two or three features that interact with mating components. The DFM feedback will tell you which ones are realistic. Also remember that molding is for real production commitment. If your total order is only 300 pieces, molding will usually not beat machining, because the tooling cost does not disappear.

Scenario D: Small Metal Parts in High Volume — Take MIM Seriously

Here is the category that surprises buyers who only know machining and die casting: small metal components, produced in high volumes with complex geometry. Think of stainless levers, latch mechanisms, surgical instrument parts, and other components that weigh only a few grams. The metal injection molding parts market has matured to the point where manufacturers quote MIM with confidence, and the process is no longer treated as an exotic alternative.

MIM begins as powdered metal mixed with a binder, injection molded to shape, then debound and sintered. During sintering the part shrinks and densities to roughly 95 to 98 percent of a wrought material, which is good enough for many structural applications but not always equal to forged or machined material. You need to test the properties against your load case, not simply assume that they match a bar stock specification.

The financial analysis is where most buying mistakes happen. At 2,000 pieces of a 4-gram stainless lever, our machining quote was around $2.10 per part. The MIM route quoted less than a third of that per part, but it priced in tooling and required batch volumes. At 2,000 pieces, machining stays cheaper. At 15,000 pieces, MIM starts to get attention. The trick is the annual volume commitment. If you cannot promise enough parts to amortize the tooling and sintering setup, MIM will be a beautiful unit price that never pays you back. Design freeze matters even more here, because changes after tooling affect the whole sintering behavior.

Which Scenario Is Yours?

Here is how I decide when a trade study lands on my desk:

• Design still changing and the order is under fifty pieces? Use Scenario A: digital quick-turn CNC or additive. No molds, no machine purchases.
• Flat sheet metal, stable design, but your laser utilization is under fifteen to twenty hours per week? Scenario B: order from a cutting and bending supplier. Keep your equipment budget as working capital.
• Thousands of plastic parts and the design is frozen? Scenario C: get a mold quote from ProtoLabs and compare total cost including tooling, not just component price.
• Small metal parts at high annual volumes? Scenario D: ask for a MIM quote, but only if the design freeze and volume commitment can support it.

If a part does not fit neatly into any of these, compare total cost rather than unit cost. Total cost includes the quote, tolerance reviews, tooling, transport, inspection, rework, and the cost of waiting while the design changes. The process that minimizes that sum for your scenario is your cheapest process.

I like digital manufacturing services because they make process costs visible up front, and visibility is what lets a cost controller make decisions instead of guessing. That does not make the traditional job shop obsolete. It is simply another scenario-respecting tool in the same toolbox. The more honest you are about your scenario, the less you will overpay, regardless of which route wins.