Metal 3D printing has spent a decade being sold as the future of manufacturing, and most buyers we talk to have concluded it is either magic or a scam. Both positions are wrong. It is a process with a narrow economic band, and inside that band it does things no other process can do at any price. Outside that band it is a very expensive way to make a part that a mill and a lathe would have produced in a week.
So let's start with the band, then get to the money. If your part is a solid block of aluminium with three holes in it, close this page and call your machinist. If it has internal passages a drill cannot reach, or it is one of forty sub-parts currently welded and bolted into an assembly, keep reading.
What the four metal processes actually are
Buyers use "metal 3D printing" as one thing. It is at least four, and they are not interchangeable.
Laser powder bed fusion — sold as DMLS or SLM depending on which machine brand the supplier runs — spreads a 20–50 µm layer of metal powder, melts the cross-section with a fibre laser, and repeats. This is the workhorse for dense, functional metal parts: aluminium AlSi10Mg, stainless 316L and 17-4PH, titanium Ti-6Al-4V, Inconel 718, maraging steel and tool steels. Density runs 99%+ after parameter tuning, and the parts take standard machining, heat treatment and finishing.
Electron beam melting (EBM) does the same job with an electron beam in vacuum, at higher temperature. It suits titanium and other reactive alloys where residual stress matters, and it produces a coarser surface that always needs machining back.
Binder jetting prints a binder into the powder and sinters the green part in a furnace afterwards. Cheaper per part at volume, isotropic, and good for stainless and copper — but it shrinks during sintering and needs a supplier who has actually dialled in compensation for your geometry. Not a process to choose for a one-off.
Metal FDM / bound filament extrudes a filament loaded with metal particles, then debinds and sinters it. Think of it as metal injection moulding with a printer in front. Dimensional accuracy and density are the weakest of the four, but the entry cost is a fraction, and it is genuinely useful for prototypes and low-stress parts. Do not let a supplier quote you "metal printing" without telling you which of these you are buying. They differ by 5–10x in price for the same file.
Where the economics actually work
Additive does not compete with machining on cost per part. It competes on cost per function and on cost per assembly. Four situations where we see it genuinely pay off:
Internal geometry a tool cannot reach. Conformal cooling channels in an injection mould insert, a manifold with curved internal passages, a heat exchanger with a lattice core. A machined insert with drilled cooling is straighter, hotter and slower-cycling; a printed one can follow the cavity wall, and the mould cycle time drop frequently pays for the insert inside a few months of production. This is the strongest business case in the industry and it is not close.
Part consolidation. A bracket that was seven welded and bolted pieces becomes one printed body. Less weight, fewer joints, fewer failure points, no welding distortion, and you delete an assembly operation. Even when the printed part costs more than the individual pieces, the total landed cost can drop because you removed the welding, the fixtures and the quality inspections around the joints.
Topology-optimised and lightweight parts. Brackets and arms in aerospace, motorsport, drones and robotics where every gram matters. Organic-looking ribs and hollows that come out of FEA software are printable and are not machinable at any rational cost.
Low volumes of high-value parts. Spares for legacy equipment, one-piece replacement impellers, tooling for a product that will only ever run a few thousand units. When the alternative is a tool with a 20 000-piece minimum and a six-month lead time, printing one hundred parts is cheap by comparison.
The test that settles most arguments: compare the printed part against the finished, inspected, ready-to-fit conventional version — including welding, tooling amortisation, fixtures and the inspection steps you delete. Buyers who compare a printed price against a bare machining quote always conclude additive is too expensive. Buyers who compare against the whole assembly usually find it is not.
What it costs, honestly
Metal printing is priced by build volume and machine time, not by weight, but weight is the number buyers can sanity-check fastest. As a rough guide for laser powder bed fusion in China, in 2026:
| Material | Indicative price per kg (printed, unfinished) | Notes |
|---|---|---|
| Stainless 316L | USD 90–160 | The cheapest functional option; laser-machinable and weldable |
| AlSi10Mg aluminium | USD 130–220 | Lighter, good thermal parts; needs stress relief |
| 17-4PH / maraging steel | USD 150–260 | Strong, hardenable, common for tooling and functional parts |
| Ti-6Al-4V titanium | USD 350–650 | Where titanium makes sense, nothing else comes close |
| Inconel 718 / 625 | USD 400–800 | High-temperature parts; slow to build, expensive powder |
| Copper / CuCrZr | USD 300–550 | Difficult to print, excellent for thermal applications |
Those are the printed, as-built numbers. Add support removal, stress relief or heat treatment, base-plate cut-off, machining stock removal on critical faces, and HIP or surface finishing where the application needs it. On a typical functional part, post-processing is 30–60% of the total. A supplier who quotes you a raw figure and then reveals the heat-treatment and machining line items later is not lying — they are quoting a different thing than you assumed. Ask for a line-by-line breakdown, and ask which faces are as-built and which are machined.
Sizes matter more than people expect. Most Chinese service bureaus run build envelopes of 250 × 250 × 300 mm up to 400 × 400 × 400 mm for laser powder bed. Larger parts need to be built in sections and welded or bolted, or moved to a large-format machine with long lead times. If your part is 600 mm long, ask early — it changes the supplier shortlist immediately.
How to source it without getting burned
A metal printing service is not a commodity. Two suppliers can run the same file and hand back parts that differ in dimensional accuracy, density and surface in ways your incoming inspection may not catch until the part is in a machine. Six questions do most of the filtering:
Which machine, and how many? A real bureau owns its machines; a broker quotes and subcontracts. Subcontracting is not fatal, but it means the person answering your technical questions is not the person setting your parameters, and traceability gets thin.
How do you qualify a new material or geometry? The good answer involves witness coupons, density measurement, and building a test part before the production run. The bad answer is "our parameters are fine".
Can I see a material certificate and a build report? Powder lot, machine ID, build parameters, build orientation, heat treatment record. If a supplier cannot produce these on request, they cannot support you in a failure investigation — which is exactly when you will need them.
What is your as-built tolerance, and on which surfaces? Powder bed gives roughly ±0.1 mm on good geometry and worse on overhangs, thin walls and long spans. Any supplier claiming general ±0.05 mm across a complex part is quoting a machined part or overpromising.
Who writes the print orientation and supports? Orientation decides residual stress, surface quality on critical faces, and whether the part warps. It is an engineering decision, not a slicer default.
What happens with the leftover powder? Sieved and reused powder is normal in the industry, but the reuse ratio affects final density and mechanical properties. Ask whether your parts run on virgin or blended powder, and ask it in writing for anything safety-critical.
What we tell buyers to do first
Before you send a file to three bureaus for price, answer two questions internally. First: what is this part's function, and which of its features are mission-critical? That tells you which faces need machining and which tolerances matter, and it usually cuts the cost of the printed part substantially, because you stop over-specifying surfaces that do not need it. Second: what is the fully-loaded cost of the conventional version — tooling, assembly, inspection and all? That is the number you are comparing against, and until you have it you cannot make this decision.
Then pick your process deliberately. Powder bed for functional density, binder jetting when you are making hundreds of the same stainless part, metal FDM for prototypes where the geometry matters more than the properties. And pick a supplier who will tell you when metal printing is the wrong answer for your part — because the bureaus worth keeping are the ones who talk you out of a project they cannot win, not the ones who quote everything.
If you are sourcing metal parts in volume and wondering whether any of them belong in additive, send us the drawings and the annual quantities. We run the comparison on our side — printed against the finished conventional route — and if additive loses, we will tell you that too.