Machining 3D Printed Metal Parts: The Finishing Step

A laser-fused titanium bracket comes off the build plate looking finished. It isn't. The bore that locates a bearing is half a thou oversize and shaped a little like a polygon. The threaded boss won't take a tap cleanly. The face that's supposed to seal against a gasket reads 300 µin Ra — rough as a sand casting. None of that is a print failure. It's just where additive stops and cutting begins. This is about which features need the cut, why, and the parts of the job that catch people out.

Machining 3D printed metal parts means using subtractive cutting — turning, milling, drilling — to bring selected features of an additively built component into tolerance. The print produces the bulk geometry near net shape. Machining finishes the surfaces, bores, threads, and datums that need accuracy the printer can't hold.

The Best Metal 3D Printers Right Now

Before any finishing happens, something has to print the part. Metal AM spans a few process families: laser powder-bed fusion (LPBF, also sold as DMLS), which fuses powder into fully dense parts straight off the plate; bound-metal extrusion, which prints a metal-and-polymer blend that's later debound and sintered; and wire-fed deposition, which builds big near-net shapes fast. Five systems worth knowing, one per niche.

EOS M 290

The EOS M 290 is the machine most people picture when they hear "metal 3D printing" — a 400-watt fiber laser fusing powder in a 250 x 250 x 325 mm chamber, building fully dense parts in steels, titanium, aluminum, and nickel superalloys.

Its draw is repeatability. A sealed chamber and years of qualified parameters make it the default for regulated aerospace and medical work, and the benchmark every other LPBF system gets measured against.

Approximate cost: $800,000 to $1,000,000, configuration-dependent.

Velo3D Sapphire

Velo3D built the Sapphire to kill supports. Its Intelligent Fusion process prints low overhang angles — down to roughly 10 degrees — and tall internal channels with little or no scaffolding.

That lets engineers print closed impellers, heat exchangers, and rocket parts in one piece, which has made it a fixture in aerospace and defense. The sealing and mating surfaces still come back to a machine, as with any LPBF part.

Approximate cost: around $1 million and up, depending on build size and laser count.

Sciaky EBAM 300

The Sciaky EBAM 300 is the outlier, and the most relevant one here. It feeds metal wire into an electron beam — directed energy deposition — laying down titanium or Inconel at up to 20 pounds an hour, in the largest build envelope of any commercial metal printer.

The cost is resolution: it builds rough, oversize near-net shapes, so every functional face gets turned or milled to size afterward. No machine makes the case for the finishing half of additive manufacturing more plainly.

Approximate cost: typically quoted in the low millions, often $1.5 million and up, custom-configured.

Markforged Metal X

The Markforged Metal X took metal printing off the cleanroom floor. It extrudes metal powder bound in plastic like a desktop FDM printer, then washes and sinters the green part into solid 17-4 PH, 316L, or tool steel — no loose powder, no laser, no inert gas.

Shops run it in-house for tooling, jigs, and low-volume parts. Sintering shrinks the part about 20 percent, so tight features still finish on a mill or lathe, but it's among the most accessible ways into metal AM.

Approximate cost: printer from $99,500; full workflow with wash and sinter stations about $165,000 to $180,000.

Xact Metal XM200C

The Xact Metal XM200C proves LPBF doesn't have to cost a million dollars. A gantry-mounted laser keeps the price down while parts still come off the plate fully dense, across powders from stainless to cobalt chrome at 20-micron layers.

It's built for small shops, universities, and labs that want real powder-bed printing without industrial overhead — modest build volume, but a credible entry point for small precision parts.

Approximate cost: around $80,000 (entry configuration).

Which brings us back to the cutting.

Why As-Printed Metal Misses the Drawing

Powder-bed fusion builds a part by melting metal one layer at a time. Good for geometry. Less good for the two things engineers actually inspect: dimensional accuracy and surface finish.

Tolerance: Close, But Not the Drawing

A typical DMLS process holds about ±0.003 in. (0.076 mm), plus another ±0.001 in. for every inch of length. Fine for a structural web. Nowhere near a press-fit bore or a bearing journal. The printer gets you into the neighborhood; it doesn't get you to the address.

Heat, Stress, and a Sand-Cast Surface

Then there's heat. The laser melts, the metal cools, over and over, and that bakes internal stress into the part. Stress relief brings it back — but relieving stress moves the part. So even a well-built component shifts a little, and that shift eats whatever accuracy you had. The surface tells the same story: vertical walls come out roughly sand-cast quality, and every angled face shows stair-stepping from the layers. Anywhere a feature has to mate, slide, or seal, as-built just doesn't get there.

What Features Actually Need Machining

Here's the thing most first-timers get wrong: you don't machine the whole part.

You Don't Cut the Whole Part

You machine the features that carry a tolerance the printer can't hold, and you leave the rest as-printed. That's the entire economic argument for the hybrid route — print the complex shape, cut only what has to be precise. Machine everything, and you've thrown away the reason you printed it.

The Features That Carry the Fit

So which ones? The short list that almost always needs cutting:

• Mating surfaces that bolt or clamp to another part
• Bores that locate bearings, dowels, or shafts
• Threaded holes — as-printed threads aren't reliable; a tapped or milled thread is
• Sealing faces for O-rings and gaskets
• Any datum the rest of the part is measured from
• Flatness or parallelism callouts tighter than roughly five thou

As-Printed vs. Finished, by the Numbers

The numbers make the case better than I can:

 

One detail worth knowing: powder-bed metal cuts about the same as wrought stock. A printed 17-4 PH or Ti-6Al-4V part machines like a billet of the same alloy — same speeds, same feeds, give or take. The hard part was never the cutting. It's everything around it.

Fixturing and Sequence: The Part That Catches People Out

A printed part doesn't arrive as a tidy block. It arrives fused to a build plate, often with organic curves and no flat face to clamp. So the sequence matters more than the cutting does.

Off the Plate, Then Stress Relief

First, the part comes off the plate — wire EDM or a bandsaw, usually. Then stress relief, before any machining, not after. Machine a stressed part and it'll move the moment you free it from the fixture, and your good dimensions walk off the drawing. Get the order wrong and you'll chase tolerances you already had.

Workholding With No Datum

Fixturing is its own puzzle. With no natural datum, shops design sacrificial tabs right into the print — small machinable pads that give the workholding something to grip, then get cut away at the end. Soft jaws cut to the part's contour do the same job for batch work. This is routine for a precision machining shop that runs secondary operations on additive blanks, but it's real labor, and it's why "just print it" rarely means just print it.

Leave Stock to Cut

One more design rule: leave stock. Print the features that need machining slightly oversize — a half millimeter of extra material gives the cutter something to clean up. Skip that, and there's nothing left to true the surface with.

Where This Lands: Small Precision Parts

The hybrid route bites hardest on small components with a few critical features buried in a complex body. Think a printed manifold with cross-drilled ports. A titanium implant with a threaded stem. An aerospace fitting where one sealing land has to be perfect and the rest can stay rough.

Why the Finish Is Often Turning, Not Milling

On parts like these, the finishing operation is often precision turning — small diameters, slender features, tight concentricity between a printed body and a machined journal. That's a job for a sliding-headstock setup. The guide bushing supports the work right at the cutting edge, so slender features hold size instead of deflecting away from the tool. Features like these get finished on single-setup Swiss turning for small-diameter parts, which holds the body-to-feature relationship in one clamping — no re-fixturing, no datum shift between operations.

A Worked Example

Take a printed 316L flow body for an analytical instrument. The internal channel network is the whole reason it's printed at all — you'd never machine those passages conventionally. But the two end ports each need a sealing taper and an O-ring groove to mate with steel fittings. So the body prints as one piece, gets stress-relieved, then goes to a lathe that cuts just those two ports to size. Everything else ships as-printed. One part, two finishing operations, the rest untouched.

Feature-by-Feature

The pattern holds across all of them: the print buys you geometry, the cut buys you the fit.

When the Hybrid Route Isn't Worth It

Worth saying plainly — sometimes the hybrid route is the wrong call. If most of a part's features carry tight tolerances, you're machining most of the surface anyway, at which point printing a near-net shape first just adds cost and a stress-relief cycle for no real gain. Cut it from bar or billet and skip the printer. Same goes for simple geometry in modest quantities: a part a lathe can finish in one setup doesn't need an additive step in front of it. Printing earns its keep when the shape is genuinely hard to machine — internal channels, lattices, consolidated assemblies — and only a handful of features need precision. Outside that window, plain subtractive work is usually faster and cheaper.

Bottom Line

Metal additive didn't replace the machine shop. It changed what shows up at the door — near-net shapes with a few features that still need cutting, rather than solid blocks. The skill now is reading a printed part for what the laser couldn't hold: the bore that has to be round, the thread that has to gauge, the face that has to seal. Get the sequence right, leave stock where it counts, fixture the awkward geometry, and a printed blank becomes a finished part. Miss any of those, and you've got a good-looking shape that fits nothing. The print is the start. The cut is what makes it a part.