China's prototype CNC processing, explained for engineers

Twenty years ago the reason to machine a prototype in China was labor cost. That is no longer the main reason. The shift that matters to engineers is equipment density. A shop running 16 simultaneous 5-axis machining centers can put a complex housing on one setup, and one setup removes most of the stack-up error that comes from re-fixturing a part four times.

The second change is supplier depth. Aluminum plate, titanium bar, and engineering plastics sit within a few hours of most Dongguan machine shops, so a prototype does not wait two weeks for material. Tooling grinders, anodizers, and heat treaters are on the same industrial block. That compresses the schedule more than any single machine does.

China's prototype CNC processing is not automatically cheaper for every part. On a simple turned bushing with a loose tolerance, a domestic shop may match the price once freight and duty are counted. The gap opens on parts with 3D contoured surfaces, thin walls, or several tight bores that must stay coaxial. Those are the parts where one 5-axis setup beats four 3-axis setups, and where the cost model actually changes.

A quote is only as good as the machine it is based on. GreatLight runs 127 high-precision CNC machines across three wholly-owned plants covering 7,600 m²: 16 simultaneous 5-axis centers, 12 four-axis mills, 27 three-axis machines, and 16 mill-turn centers. The choice between them is a geometry question, not a price question.

Five-axis suits impellers, turbine blades, medical instrument bodies, and any part where the feature normal points in many directions. One rotary table at Ø400 mm handles most of it. The payoff is fewer setups, better positional accuracy between features, and shorter programming-to-chip time.

Three-axis still wins on prismatic parts with features reachable from one direction: plates, brackets, manifolds without cross-drilling. Programming is faster and fixturing is simple. Pushing that work onto a 5-axis machine adds cost without adding accuracy.

Mill-turn centers cover parts that are turned and milled, such as shaft collars with cross-holes or hydraulic fittings. The alternative is two operations on two machines, which means two datums and a second setup error. Whenever a prototype has both a rotational axis and off-axis features, mill-turn is worth asking about.

Size sets the outer limit. Maximum processing size is 4,000 mm, with travel envelopes of 4,000 × 400 × 150 mm, 750 × 1,150 × 550 mm, 600 × 600 × 600 mm, 500 × 500 × 450 mm, and 500 × 310 × 200 mm. A part outside the largest envelope cannot be machined in one piece, and splitting it adds a joint to your design.

Achievable tolerance is ±0.005 mm (±0.0002 in) on critical features. That number is not a blanket specification for the whole part. It applies to the dimensions you actually call out, on features the shop can reach with a stable setup and a sharp tool. Applying it to every dimension on a prototype drives cost up and inspection time up for no functional gain.

Be selective with GD&T. A true position callout on two bearing bores is worth the inspection time. The same callout on a cosmetic edge is not. Datum features should be surfaces the shop can hold in a vise or on a fixture, not a theoretical intersection that cannot be touched.

Surface finish follows the same logic. As-machined is Ra 1.6–3.2 μm. A high-finish pass gets Ra 0.8–1.6 μm, and fine finishing reaches Ra 0.2–0.8 μm on sealing faces and bearing journals. Each step adds a tool change and usually a second operation, so mark only the surfaces that seal, slide, or mate.

Ra numbers on a prototype should match the function. A bracket at Ra 1.6–3.2 μm is fine. A hydraulic spool bore at Ra 0.8–1.6 μm needs the extra pass. Specifying fine finish across the whole part rarely changes how it performs in a fit check.

Material availability is one reason China's prototype CNC processing moves quickly. Common grades sit in stock or within a day: aluminum 6061, 6061-T6, 2024, 5052, 5083, 6063, 6082, 7075, and ADC12; stainless 303, 304, 316, 316L, 420, 430, 431, 440C, and 17-4PH (SUS630); steel 1018, 1045, 4130, 4140, 4340, A36, and tool steel.

Copper and brass grades include C101, C103, C110, beryllium copper, C27400, C28000, and C36000. Titanium and specialty stock covers TA1, TA2, TC4 (Ti-6Al-4V), Inconel, and magnesium AZ31B and AZ91D. Plastics cover ABS, PC, PMMA, POM, PA, PEEK, PP, HDPE, and carbon fibre.

Pick the grade for the test, not for the production part. A prototype in 6061 aluminum proves geometry and fit; it does not prove that a 7075 production part will behave the same under load. If stiffness matters, machine the prototype in the alloy you intend to ship. If it is a fit check, aluminum or POM is faster and cheaper.

17-4PH and Inconel are machinable but slow. Tool wear is high, feeds are low, and the part may need stress relief between roughing and finishing. Budget more time for those, and expect a shorter list of shops willing to quote them.

Inspection is the part of a prototype order that is easiest to skip and hardest to recover from. GreatLight inspects 100% of parts before shipment, with a raw material check, in-process monitoring, and a final inspection. Reports are available on request, including dimensional data for the features you call out.

Certifications matter when the prototype feeds a regulated program. Four are held here: ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022. The first covers general quality systems, the second covers automotive, the third covers medical devices, and the fourth covers information security. Historical qualification rate is 99.99%.

Confidentiality is a real concern for unreleased designs. Uploads are handled as secure and confidential, and an NDA is available on request. For programs with a long life, it is worth signing one before the first model leaves your network, not after the quote.

Timing is part of quality too. A quotation and free DFM analysis come back within 12 hours, production can start within 24 hours of approval, and parts ship in 3–5 days. Historical late-delivery probability is below 2%. Those windows assume the model is manufacturable as drawn; DFM feedback often changes the design before the first cut.

Prototype machining is not the right answer for every part. If the geometry has undercuts the tool cannot reach, or internal channels that must be hollow, machining will not produce it. Additive or casting may be the better process, and a good shop will say so rather than quote a job it cannot hold.

Very small features have limits too. Thin walls below roughly 0.5 mm deflect under cutting force, and deep pockets with a high depth-to-diameter ratio need long tools that chatter. These are design conversations, not machine limitations. Change the wall thickness or the corner radius and the part becomes machinable.

Volume is another boundary. Prototypes run from one piece to small batches with no minimum order quantity, and the same process scales to runs of 10,000 or more. Below a few hundred units, machining is usually competitive. Above that, die casting or injection molding may win on unit cost, though tooling adds lead time and upfront spend.

Finally, consider the feedback loop. The value of an overseas prototype run is not just the part. It is the DFM note that says a fillet will crack, or a bore needs a different tolerance to be reamed. A shop that only returns a part and an invoice has given you half the job.