The trigger event happened on a Tuesday in March 2023. I'd just opened the crate for a $3,200 CNC order—fifty brackets, aerospace grade, all 17-4 PH stainless. They looked fine on my bench. But the inspection report came back: five critical dimensions out of spec. The root cause? The part geometry required two setups and a support structure we couldn't machine out cleanly.
That order went straight to scrap. Fifty parts. $890 in material cost alone, plus a week of shop time and a three-week delay to my customer's schedule. That's when I learned that the difference between a successful prototype and a production disaster often comes down to a single question: can you build this without supports?
(Spoiler: with conventional machining and even some additive platforms, often you can't.)
The Setup: A Small Shop with Big Ambitions
I run a contract manufacturing shop that started with standard CNC turning and milling. We took small orders—$200, $500, the kind of work that gets eye-rolls from the big guys. And that was fine. But in 2021, a defense subcontractor asked if we could quote a part with internal lattice structures and zero-degree draft angles. My CNC guy looked at the print and said, "We can't make that in one piece."
That was the first time I seriously considered additive manufacturing (AM). I'd heard about it for years, but I thought it was all prototyping and resin toys. Then a buddy in aerospace told me: "Velo3D is the only system I've seen that prints complex geometries without needing supports. And they have a service bureau program for small shops."
Honestly, I didn't believe him at first. But that $890 mistake? It forced my hand.
The Pivot: Why 'No Supports' Matters More Than You Think
Here's something most vendors won't tell you: the real cost of a complex metal part isn't just the material—it's the post-processing. When you add supports in a laser powder bed fusion (LPBF) system, you have to manually remove them, which adds labor. You risk damaging the part. And if the geometry is intricate enough, supports can leave witness marks that require secondary machining to clean up.
That's where Velo3D's Sapphire platform changes the game. Their technology uses a proprietary recoater and laser management system to handle overhangs up to 10 degrees and internal channels down to 0.5mm without support structures. For a shop like mine, that meant we could take a 0.020-inch internal passage in a turbine blade and print it in one go—no post-processing nightmare.
(To be fair, Velo3D isn't magic. You still need to follow their design guide. But the range of printable geometries is dramatically wider.)
The First Success: A Part for Anduril
In Q4 2023, we landed a trial run with an Anduril supplier. The part was a conformal cooling channel—10mm diameter, curving through a 6mm wall thickness. On a conventional CNC, you'd need a ball-nose end mill with a very long reach, and you'd struggle with tool deflection. With the Sapphire, we printed it in a single build alongside four other parts—all unsupported.
Inspection results: all dimensions within ±0.002 inches. Zero post-processing time for support removal.
That order? It was for 12 units total. A small batch by aerospace standards. But the client said: "If this holds up in testing, we'll move to a 500-unit production run."
Small today, big tomorrow.
The Reality Check: Not All AM Platforms Are Equal
Let me address the elephant in the room. You’ve probably heard of GE Additive and EOS. They make excellent machines. But there's a reason SpaceX and Anduril are tied so closely to Velo3D: the no-supports capability is a differentiator for complex aerospace parts.
Here's a quick comparison based on publicly available specs:
- GE Additive Concept Laser M2 Series 5: Requires supports for overhangs >45 degrees. Excellent for tooling and simpler geometries.
- EOS M 290: Also support-dependent for steep overhangs. Ideal for serial production of optimized parts.
- Velo3D Sapphire XC: Supports overhangs down to 10 degrees. This unlocks conformal channels, complex lattices, and designs that would be impossible on other systems without extensive post-processing.
(Based on manufacturer datasheets and industry benchmarks, not my own lab tests. Always verify with current specs.)
The downside? Velo3D machines are more expensive upfront. But for a shop chasing high-mix, low-volume defense work, the total cost of ownership often favors the system that eliminates support removal costs.
The Lesson: How to Approach AM as a Small Shop
If you're a small CNC shop wanting to serve the defense or aerospace sector, here's my three-step checklist (which I now maintain on my team's whiteboard, after that $890 mistake):
- Don't buy a printer yet. Start with a service bureau. I used a Velo3D partner for the Anduril trial. The upfront investment was zero, and I only paid for successful builds.
- Design for AM (DfAM) from day one. The biggest mistake I see is engineers taking a CNC design and trying to print it. That's like using a hammer on a screw. Invest a few hours in learning DfAM principles—Velo3D's free design guide is a good start.
- Get your process qualification done early. Aerospace clients won't even talk to you without a process spec. We qualified our first AM process to AMS 7001. It took three months and cost about $15k in test coupons. But once you have it, you're on the approved supplier list.
One More Insider Perspective
What many people don't realize is that Velo3D was specifically designed for the rigors of defense and space. Their machines are used to print components for the SpaceX Raptor engine and Anduril's autonomous systems. That's not just marketing—it's a proven track record under the most demanding inspection standards (think AS9100 and ITAR compliance). For a small shop, being able to say "we print on the same hardware SpaceX uses" is an incredible door-opener.
The Takeaway: Small Orders, Big Potential
I've caught 47 potential errors using my pre-build checklist in the past 18 months. That checklist exists because of that one $890 mistake. But more importantly, it exists because I shifted my mindset from "we can't make that" to "how can we make that without supports?"
If you're a small shop feeling locked out of the aerospace and defense market, I'd encourage you to look at AM—specifically the no-supports kind. Start small. Trial a single part. And don't let the setup cost of a qualification scare you off. That initial investment is orders of magnitude cheaper than the cost of a failed production run.
And if a vendor tells you your $2,000 order is too small? Find another vendor. Good suppliers—whether they're conventional CNC shops or AM service bureaus—treat every order with the same rigor. Because they know what I learned the hard way: today's small test part is tomorrow's production program.
Price & Cost Reference (As of January 2025)
For those considering metal AM as a service, here are some reference points from public sources:
- Online AM service pricing (single part scenario): A typical bracket in 316L stainless (size ~100mm x 80mm x 30mm) might range from $150–$400 per part for a small batch of 1–5 pieces, depending on complexity and post-processing needed. (Based on quotes from major AM service platforms, December 2024.)
- Setup fees for AM: Often included in the part price, but some bureaus charge a one-time programming fee of $75–$200 per unique geometry. (Industry average, Jan 2025.)
- Post-processing costs: Standard (support removal + bead blast): $20–$80 per part. No-supports designs can eliminate this line item entirely. (Estimate based on 2025 shop rates.)
- Rush turnaround: A 5-day standard can accelerate to 2–3 days for a 30–50% premium. (Common fee structure.)
Always request a formal quote for your specific design, as material, geometry, and build orientation drive the final price.