Here's the short version: Velo3D's status as a SpaceX and Anduril supplier is real engineering validation, but it tells you almost nothing about whether their systems make financial sense for your shop. The machine cost is basically 30% of the total ownership cost. I've seen companies budget $500K for a printer, then spend another $400K in the first year on powder, post-processing, and failed builds. If you're evaluating a Sapphire system, start with the full workflow cost, not the machine price tag.
Why You Should Listen to Me
I say this as a procurement manager at a 45-person contract manufacturer on the East Coast. I've managed our additive manufacturing budget—about $430,000 annually—for six years. I've negotiated with 14 different AM vendors, tracked every purchase order in our cost system, and built a TCO model that my CFO actually trusts. Honestly, when I hear 'SpaceX uses them,' I think, 'Great, but what's my part cost per cubic inch?'
Velo3D, SpaceX, and the Real Cost of Qualification
Don't get me wrong—the Velo3D-SpaceX relationship is legitimate. Anduril using Velo3D printers for defense parts is serious business. That tells you the technology can handle complex geometries and rigorous qualification. But here's what those announcements don't say: how much qualification costs. In aerospace, you're not just printing a part; you're certifying a process. That means document control, material traceability, and repeatability testing. Those costs are on you, not the machine maker.
For those tracking additive manufacturing use-cases, Velo3D's niche is the hard stuff: fuel manifolds with internal channels, heat exchangers, and structural parts that would otherwise require machining from solid billet or assembly of multiple components. That's why companies like SpaceX and Anduril use them. But those same use-cases come with high validation costs. The part may be cheaper to print than to machine, but the certification process can make it more expensive for low volumes.
About two years ago, I compared two potential setups side by side: a Velo3D Sapphire XC and a comparably priced laser powder-bed fusion system from another vendor. Same part, same material, same build volume. The other machine had a lower sticker price by about 7%—pretty typical. But when I ran the full numbers—including consumables, argon usage, filter changes, and the time to remove supports—the Velo3D came out ahead on total cost per part, mostly because its support-free capability cut post-processing by nearly 40%. Seeing that made me realize: comparing machines is comparing ecosystems, not printers.
Of course, I learned it the hard way first. Three years ago, I approved a purchase for a different AM system without properly accounting for powder loss and sieve time. We budgeted for the lease, the materials, and the technician's salary. We forgot to include the fact that every build requires fresh powder, and the old powder has to be sieved and tested. By the end of the year, we'd overspent by $18,000—17% over our AM budget. That's a regret I still carry. Now every RFP I send out includes a 'consumables per 100 build hours' line.
A Realistic Budget Checklist
- Machine price: $300K–$800K depending on config (Sapphire, XC, 1MZ)
- Installation and site prep: $15K–$30K (foundation, power, argon lines)
- Year 1 materials and consumables: $50K–$100K (powder, wiper blades, filters, argon)
- Post-processing equipment: $40K–$80K (depowdering station, heat treatment, CNC for any machining)
- Qualification and testing: $20K–$50K per material/process if you're serving aerospace
- Training: $5K–$10K per operator
That's why I say the machine is only 30% of the total. I'm not 100% sure on Velo3D's current pricing, but those are ballpark numbers I've seen in quotes from multiple vendors since late 2024. The reason most buyers ignore these line items is that the machine quote looks deceptively simple. The list price includes the system, maybe a powder loading station, and basic installation. The rest comes as 'phased' costs—you only discover them when you start production. That's where the TCO model matters.
According to ISO/ASTM 52920, the additive manufacturing process chain includes powder handling, building, depowdering, and post-processing. The printer is just one step in a workflow that carries its own costs.
Does Costco Sell 3D Printers?
Now, about that 'does Costco sell 3D printers' question that keeps popping up. As of January 2025, Costco's website lists a few consumer desktop models (think Creality and the like), but they do not carry industrial systems like Velo3D. And you wouldn't want them to. A Sapphire system requires dedicated installation, specialized training, and a supply chain for metal powders. That's not a Costco aisle. If someone tells you they bought a 'metal 3D printer' at Costco, they probably bought a toy that kind of prints metal-filled plastic filament.
When You Shouldn't Buy a Velo3D
All that said, Velo3D isn't the right answer for every use case. If you're making simple brackets or low-cost parts in high volume, traditional CNC or even injection molding will beat it on price per part. In fact, recent injection molding industry news has been about companies using 3D-printed inserts for rapid tooling—not replacing molding entirely. Mold inserts with conformal cooling channels are one of the few places where metal AM genuinely beats conventional machining; I've seen a resulting 25% reduction in cycle time. But that's a different cost model than production parts. And if your parts don't have the complex, support-free geometries that Velo3D excels at, you may be better off with a more standard AM system. The point isn't which printer is 'best'—it's which one actually makes money for your specific part portfolio.
Final Caveat
Part of me wishes I had started with a Velo3D system three years ago. Another part knows the expensive mistake I made with the other vendor taught me the framework I now use. I have mixed feelings about that 'cheap' system—on one hand, it hurt our budget; on the other, it forced us to build proper TCO tracking. So don't repeat my mistake. Build your own model, talk to real users, and include every consumable, every hour of post-processing, and every qualification test. That's the only way to make this decision.