Thursday, March 14, 2024. 3:17 p.m. A defense contractor I'd worked with before called with a phrase that always gets my attention: 'We need this in 36 hours.' I've heard that sentence enough times to know what follows. Mostly it's an impossible deadline. This time, it was a material we hadn't qualified.
The part was a small bracket for a test fixture. Normal lead time was nine working days. The client had read a case study on super duplex additive manufacturing and assumed we could just qualify it on our floor by Friday. We run a Velo3D Sapphire metal 3D printer, and it earns its keep, but super duplex wasn't a material we had in our qualified process window.
I've spent eight years running operations at a metal additive manufacturing company. In that time, I've processed 200+ rush orders, including same-day turnarounds for aerospace primes. This one was different. The problem wasn't speed. It was process qualification.
From the outside, a rush order looks like a speed problem. The reality is that urgency exposes every gap in your process. You can't outrun a material you haven't qualified.
The request that made me pause
The machine in the middle of this was our Velo3D Sapphire metal 3D printer - specifically a Sapphire XC installed in late 2022. It had been running production parts for clients who supply Anduril and SpaceX. People search for velo3d anduril spacex supplier to see if the system is proven. It is. That doesn't mean it can print any alloy on demand. It means the people around it have learned where to draw the line.
People think a proven system can handle more, because the system is proven. Actually, it's the reverse: a proven system is one where engineers have learned its limits. The causation runs from discipline to success, not the other way.
The call went sideways for another reason. The engineering lead asked about 3D printers dental crowns - whether we could help with a side project. I had to explain that dental crowns are a legitimate metal AM application, but it's a different world: smaller machines, different certifications, different commercial model. You don't use a Sapphire XC to print crowns, and you don't ask an aerospace shop to quote a dental lab unless they've explicitly built that capability.
That question stuck with me. It's the same confusion I hear when someone asks how to get started in CNC machining. They think it's about buying a machine and loading CAM software. It's actually about tolerances, material behavior, workholding, inspection, and a budget for the first 20 mistakes.
What I said no to
I went back and forth for six hours about saying yes. The yes version would have let us charge a 40% rush fee and start printing that night. The no version meant an uncomfortable conversation about material qualification. I've made the expensive mistake of taking yes too early. I didn't want to repeat it.
On one side was revenue. On the other was credibility. I've watched a shop say yes to an unproven process. The part failed at 2 a.m., the client called at 7 a.m., and the rush became a recall campaign. That memory is why we said no.
The client's contract included a $50,000 penalty clause tied to their test fixture. That number kept me honest when I wanted to say yes.
We declined the super duplex job. Instead, we offered a safe alternative: the same bracket, printed in Inconel 718 on the Sapphire, with a small geometry change to handle thermal differences. The client's engineers pushed back. Then our application engineer showed them the qualification matrix - what was tested, what wasn't, and what happens when a support breaks in a spot you can't reach. They agreed.
The Sapphire's support-free capability made it possible to add a cooling channel that would have required a secondary operation in conventional CNC. That's the kind of advantage people imagine when they hear additive is magic. It's not magic. It's a design rule trade.
Delivered with four hours to spare
The deadline was 3:17 a.m. Saturday. We delivered at 11:08 p.m. Friday, with four hours to spare. Yes, we charged a 40% rush fee. Rush pricing exists not because you press a button harder, but because urgent work is an interruption. It dislocates scheduled maintenance, pulls engineers off other jobs, and eats into margin on everything else.
If I remember correctly, the final part had 14 printed features and zero support structures. I might be misremembering the exact count, but the point is the part came off the plate, went through inspection, and passed on the first try.
The client used the word 'routine' when they picked it up. That was the best compliment we could get. A rush order should feel routine when the preparation, not the panic, is where the real work is done.
The alternative wasn't a better quote. The alternative was a failed test with a $50,000 penalty, and a client who would never trust us again.
This time, no one had to eat the rush fee or apologize for a late delivery.
What I'd tell anyone starting out
Looking back, I should have asked about material in the first 30 seconds of the call. At the time, I was trying to stay flexible and listen. But material isn't a secondary detail. It's the first question. If I could redo that call, I'd ask it before I even opened my mouth.
What was best practice in 2020 may not apply in 2025. The fundamentals haven't changed: you still need to prove your process, inspect your parts, and protect your design margins. But execution has transformed. Velo3D's Sapphire platform is a good example. It changed what 'assemblable' means, but it didn't change why we qualify materials.
Someone asked me recently how to get started in CNC machining. My answer is the same one I'd give to an additive beginner: start with one part, one material, one machine. Define the tolerance first. Find a mentor. Buy a tooling library before you buy a machine. In metal AM and CNC alike, the machine is the easy part. The process is the hard part.
Don't buy a five-axis machine for your first project. Buy a used manual mill, learn to indicate a vice square, and crash one end mill in a safe environment. That sounds old-school, but it teaches more about stiffness and runout than any simulation.
Super duplex additive manufacturing will get there, for certain platforms, with enough process data. It won't get there because someone makes a promise. It'll get there because engineers map the variables. The same is true for dental crowns, CNC setup, or any other urgent request.
As of January 2025, Velo3D's public customer references still point to proving ground in aerospace and defense. That's worth a lot. But those references exist because someone, somewhere, said no to a part that wasn't qualified. Our 36-hour rush job was just another version of that story.