When a metal prototype needs to exist this week, two routes dominate: CNC machining it from a solid billet, or 3D printing it in metal. Teams often treat the choice as philosophical — "additive is the future" versus "CNC is the safe bet." It is not philosophical. It is a geometry, material, and quantity problem, and the answer is usually obvious once you price the actual drivers. Here is how we help customers decide, including the cases where the honest answer is "neither alone."
What Each Process Is Actually Good At
CNC machining cuts parts from wrought material — 6061-T6, 7075-T651, 316L, titanium. That means the finished part inherits the wrought alloy's full mechanical properties: tensile strength in the 290–310 MPa range for 6061-T6 in the as-delivered state, consistent microstructure, and established design allowables your stress engineer will accept without argument. Tolerances of ±0.05 mm are routine on features within reach of standard tooling, and ±0.01 mm is achievable on datum-controlled setups with stable workholding. Surface finishes from Ra 0.8 µm down to better than Ra 0.4 µm with light finishing passes.
Metal 3D printing — predominantly laser powder bed fusion (LPBF) for prototyping — builds layer by layer from powder. Its strength is geometric freedom: internal cooling channels, conformal lattices, topology-optimized brackets, and consolidated assemblies that would otherwise need welding. The catch is that as-built properties are not wrought properties. As-built LPBF AlSi10Mg typically lands around 300–330 MPa ultimate tensile in XY orientation, which is respectable, but elongation and fatigue behavior vary with build direction and post-processing, and many stress groups still apply knock-down factors to printed metal for critical features.
The Cost Picture: Where the Crossover Sits
For a single part, the comparison is mostly machine-hours versus build-hours plus post-processing. A typical machined aluminum bracket, 100 × 80 × 25 mm with a handful of pockets and tapped holes, often runs 1–3 hours of cut time. The same part on an LPBF machine might build in 6–14 hours depending on nesting and orientation — and then needs stress relief, support removal, and machining of every interface face. Printed parts are almost never truly done when they come off the plate: threads, sealing surfaces, and bearing bores usually still get machined.
Practical rules of thumb we use:
- Simple geometry, standard stock sizes: CNC wins on cost almost every time. A rectangular housing with drilled and tapped features has no reason to be printed.
- Complex internals — serpentine cooling channels, lattice infill, geometry a cutter cannot reach — additive is the only option, and the premium is worth paying.
- 5–20 units: CNC on a fixture that lets you re-chuck quickly usually beats per-part print pricing, because machine time amortizes and programming is a one-time cost.
- Parts needing stress-critical fatigue data in a wrought alloy: machined, or redesign the part around printed-material behavior.
Tolerance and Finish Expectations Compared
This is where mismatched expectations burn budgets. As-built LPBF surface roughness sits around Ra 6–12 µm, and downward-facing surfaces are worse because of support dross. If your drawing calls for Ra 1.6 on a sealing face, that surface is getting machined regardless of how it was made. Plan for it. As-built linear accuracy on a well-tuned LPBF machine is roughly ±0.1–0.3 mm depending on feature size and orientation; critical features get post-machined to drawing anyway.
CNC machining holds ±0.05 mm as a shop standard, with tighter on selected features. So the honest comparison is not "printed versus machined" — it is printed-plus-machining versus fully machined. Once a printed part needs four or five machined interfaces, the additive cost advantage often evaporates for simple shapes.
Material Availability Is a Real Constraint
CNC machining has a deep material bench: 6061 and 7075 aluminum, 1018 and 4140 steel, 304/316 stainless, brass, titanium. If your product will eventually be die cast in A380 or ADC12, prototype in 6061 with slight wall adjustments, or better yet talk to your die casting supplier early — casting alloys have different allowable wall thicknesses and draft requirements than machined parts.
Metal printing's powder menu is narrower: AlSi10Mg, 316L, 17-4PH, Ti6Al4V, Inconel 718 are the common ones. There is no printed 6061 worth prototyping in, no printed 4140, and printed 17-4PH needs heat treatment to hit useful strength. If your production part is a wrought aluminum alloy, a machined prototype predicts its behavior far better than AlSi10Mg will.
Lead Time: Faster Is Not Always the Printer
The "3D printing is faster" assumption deserves scrutiny. Yes, a machine can start building the moment the file is sliced. But consider the full chain: print (6–14 hours typical for prototype-sized parts), thermal post-processing (1–2 hours), support removal and wire-EDM from the build plate (hours to a day depending on fixturing), then any required machining. A simple machined part with stock on the shelf can genuinely ship in 2–3 days. Where additive genuinely shortens schedules is parts that would need three or four CNC setups — one build replaces the multi-op choreography.
A Decision Sequence That Works
- Does the geometry contain features impossible to machine (internal channels, unreachable undercuts)? If yes, print it.
- Are all features cutter-accessible and the material available as wrought stock? Machine it — cost and material fidelity both favor it.
- Mixed case: print the topology-optimized core, then have a CNC machining partner finish the interfaces. This hybrid is common for conformal-cooled tooling inserts and heat exchangers.
- Not sure whether the design is manufacturable either way? A design-for-manufacturing review before ordering usually pays for itself; our services offered page outlines how we run these reviews alongside quotes.
Where Prototypes Should Point Toward Production
Remember the prototype exists to de-risk production, not to win a process beauty contest. If production is CNC machining of 6061, print nothing — prototype in the same alloy, same billet approach. If production is metal injection molding or casting, a machined prototype is still usually the right first article because it verifies geometry and fit faster; process-specific defects (porosity, knit lines) are better studied later with small casting trials. And if you find yourself iterating geometry every week, 3D printing's per-iteration cost predictability can offset its material compromises — just keep the fatigue-critical, sealed, and threaded features machined.
Neither process is "the future." Machined metal is the fidelity benchmark; printed metal is the geometry unlock. Choose per part, per feature, and be suspicious of any supplier who quotes only their own process.
