The word fleet

Three hundred machines do not add up to a manufacturing network

Project DIAMOnD has been placing printers with small and mid-sized manufacturers across Michigan since 2020, three hundred of them to date; Markforged supplied the first wave. On headcount alone, the hardware for a distributed manufacturing network was in place six years ago.

But what lets those three hundred sites take on a defense order together is not their number. It is the answer to a narrow question: send one set of manufacturing instructions to any machine in the network, and do you get the same part out. Not a part that looks the same — one whose dimensions and mechanical properties land inside the same acceptable band despite differences in material lot, ambient conditions, operator and post-processing, with evidence left behind for every unit.

This is hard enough inside one factory; across a distributed network it is a different order of magnitude. Phase I writes it into the deliverables: identify and validate candidate polymer and metallic weapon system components, then produce them using validated processes. Validation first, production second — the sequence is itself the specification.

The cost of qualification

Qualification cost is not additive, it is multiplicative

The volume of qualification work is set by how many machine-material-part combinations have to be qualified. In a single-fleet network, a part qualified once in one material carries that status to other sites along with the configuration. In a mixed estate, every additional machine type multiplies the combination count. This is why the number of machines a programme owns says almost nothing about whether it can take on an order that demands traceability.

It also explains why Phase I puts a secure encrypted digital workflow architecture and production using validated processes on the same deliverables list, aligned to Cybersecurity Maturity Model Certification (CMMC) and National Institute of Standards and Technology (NIST) cybersecurity standards. The first governs whether the file leaks. The second governs whether the part can be made and proven. A network exists only when both hold; with one missing, what remains is a set of machines with network access.

Not a procurement preference

The advantage of one fleet shows up in the qualification ledger, not the datasheet

Converging a fleet onto one platform intuitively feels like lock-in. Where quality evidence has to be delivered, the arithmetic runs the other way: qualifying once and letting the network inherit it, against qualifying at every site, differs by the schedule and auditability of the whole programme.

It is also why equipment suppliers on programmes like this are rarely selected on single-machine specifications. No Phase I deliverable concerns the maximum part size of any one printer. The deliverables are the secure end-to-end transfer of Smart Product Recipes (SPRs) and the digital quality assurance instructions that accompany each part produced. The equipment has one question to answer: can it reproduce the same recipe at a different site.

On the division of labour: Automation Alley leads the design and validation of the secure digital manufacturing environment; Markforged supplies industrial additive manufacturing systems; Fraunhofer USA handles applied research; NCDMM, the prime contractor, also manages America Makes and operates seven locations nationwide. Automation Alley describes the combination as bringing together additive manufacturing technology, cybersecurity and manufacturing process validation.

Greater China and Viet Nam

Turning demonstration sites into capacity is not a question of the next machine

Taiwan, Hong Kong and Viet Nam have each seen additive manufacturing demonstration sites, clusters and defense self-reliance initiatives appear over the past two years. The publicly visible discussion sits mostly on machine counts, build volumes and material options — the fields on a datasheet.

The Phase I list points at a different set of fields. How many qualified processes do your sites share? Move a part to another site and does it need re-qualifying? What format of evidence does each unit leave behind, and who can retrieve it?

None of these require a budget cycle or another purchase to answer. What they decide is whether, when an order requiring traceability arrives, those sites constitute a network or a collection of unrelated machines.

CONCLUSION

Start by counting the processes in your fleet

Do not start with equipment. Take the list of sites you already have and count three numbers: how many machine models, how many material specifications, and whether the same part runs on the same parameters at every site.

If those three numbers add up to well above one, what you have is machines, not yet a network.