The spreadsheet that almost made the decision for me
In March 2024, our aerospace components company needed to bring metal additive manufacturing in-house. Not prototyping — production. We had a defense program with roughly 340 titanium brackets per quarter, each with internal channels that no traditional machining setup could touch without a five-axis nightmare and a scrapped-parts rate I'd rather not repeat.
I run procurement here. About 90 employees, roughly $14M in annual supplier spend, and a personal rule I've kept for the past seven years: if I'm signing off on more than $100K, I build the comparison myself. No delegation.
So I built the spreadsheet. Four vendors. Velo3D, plus three others whose names I'll leave out (you can guess two of them). Columns for system price, install, training, first-year consumables, service contract, and a rough estimate of floor space cost.
Velo3D came in at roughly 38% more than the cheapest quote on the sheet.
I almost moved on.
The engineer's email that killed my timeline
Before I sent the "we're going another direction" email, I forwarded the comparison to our process engineering lead. Standard practice — I always give engineering a 24-hour window to flag anything before procurement makes a call.
She came back in about three hours. Her reply was short: "Price column doesn't include support removal labor or the scrap rate difference. Can we get a call?"
It's tempting to think a metal 3D printer's cost is the sticker price plus material. But that ignores the part of the equation that shows up six months later in your production reports — post-processing, support structure removal, surface finishing, and the rejects that never make it to inspection.
We got on a call. She walked me through what she'd seen at a previous employer who ran one of the cheaper systems: complex geometries with internal channels often came off the build plate with supports that were nearly impossible to remove without damaging the part. Her estimate — and I want to be clear this was her estimate, not a published number — was that support removal alone added 4 to 6 hours per batch on parts similar to ours. And the scrap rate on the first 30 builds ran above 15%.
That's when I started rebuilding the spreadsheet.
What the TCO model actually showed
I spent the next two weeks with a different framework. The question everyone asks is "what's the system cost?" The question they should be asking is "what does it cost to produce one qualified part, at volume, six months from now?"
Here's what went into the second model:
- Support removal labor. At our shop rate of $68/hour, 4–6 hours per batch, three batches a week, that's $55K–$80K annually. Some systems require less because of how they handle overhangs during the build.
- Scrap and rework. A 15% scrap rate on 340 parts per quarter at an average part value of $420 is roughly $86K a year walking out the door.
- Qualification cost. Every time you change a process parameter to chase yield, you're re-running qualification coupons. For aerospace, that's not optional. AS9100 and customer-specific requirements (think AMS 4999 for titanium) mean every parameter change documentation cycle costs real money.
- Downtime. A system that goes down for a week during a production run is a program delay. Program delays on defense contracts are not something I want to explain to a customer.
I re-ran the numbers. The cheapest system landed at a total first-year cost of roughly $1.62M once you loaded in the labor and scrap estimates. Velo3D's Sapphire landed at $1.71M.
A $90K gap. Not zero. But not the $570K gap the original sheet claimed.
And that's before I factored in the customer angle.
The part of the decision I didn't want to admit mattered
I'll be honest about something. As a procurement person, I like to pretend decisions are purely financial. They're not.
When our defense customer's quality auditors came through in 2023, they asked which systems produced our qualification parts. The answer mattered to them. Not because of marketing — because of traceability. If they've already seen a technology on another supplier's floor and it passed their audits, that's a shortcut through a maze I don't control.
Velo3D's customer list is public. SpaceX. Anduril. Names I recognized from program reviews. I'm not saying that's why we chose them. I'm saying it removed a category of risk that no spreadsheet column captures.
"When I switched from a budget machine to a premium one at a previous facility, customer feedback scores on dimensional conformance improved by 18%. Same operators. Same material. Different process control."
That's from a colleague I called during vendor evaluation — someone who'd run both a low-cost metal AM system and a higher-cost one in the same shop. Her point wasn't that expensive is always better. Her point was that the process control differences show up in the parts, and the parts show up in customer audits.
What we actually signed, and what I'd do differently
We signed with Velo3D in July 2024. Installation took six weeks. First article inspection passed on the second attempt (the first failure was our fault — a CAD export setting, not the machine).
Six months in: our per-part production cost is tracking within 8% of the original model. Support removal runs closer to 90 minutes per batch, not the 4–6 hours the cheaper systems would have needed. Scrap rate on production parts is under 3%.
Looking back, I should have built the TCO model before sending the first spreadsheet to engineering. At the time, I was under pressure to get three quotes on the CEO's desk by Friday, and the price comparison was fast. The full model took two weeks.
If I could redo that decision, I'd start with the production cost question and work backward to the system price. But given what I knew then — that we needed a machine, that leadership wanted options by end of quarter — the price-first approach was reasonable. It just wasn't good enough on its own.
Here's the thing I keep coming back to. It took me seven years and probably 200+ equipment purchases to understand that the cheapest qualified bid and the lowest total cost are two different numbers. They look the same on a one-page comparison. They stop looking the same around month four, when the labor hours and the rework reports start coming in.
If you're evaluating metal AM systems in 2025, here's what I'd put in your first spreadsheet, before you even talk to vendors:
- Your shop labor rate per hour
- Your historical scrap rate on comparable parts (if you have one)
- Your cost per qualification cycle
- Your customer's audit history with each candidate technology
Fill those in first. Then compare prices. The gap between the cheapest quote and the best quote shrinks fast — and sometimes it reverses.
(Note to self: save the second spreadsheet for the next capital purchase. Should have started there.)