Printing a house is the least useful thing a 3D printer can do on a building site. The useful thing is printing the twelve parts nobody will ever cut a mould for.
I should say what I have and have not done. I have not printed a building. I drew wood execution details for an interior design firm, and what stayed with me is how many of those parts would be made once, by hand, for one project, and never again. That gap is what this post is about.
The economics, stated plainly
Conventional manufacturing has a large fixed cost and a small marginal one. Tooling, moulds, dies, line setup: pay once, then each additional unit is cheap. Unit cost falls steeply with volume and rises steeply with variety, because every variant needs its own tooling.
Additive manufacturing is the mirror image. Almost no fixed cost, a significant marginal one. It does not reward volume much. It also does not punish variety at all: the thousandth part can differ completely from the first at no extra cost, because the only thing that changed is a file.
The two sit at opposite ends of one axis. Injection moulding wins wherever volume is high and variety is low. Printing wins wherever volume is low and variety is high, and it wins there by default, because in that region the alternative is not a cheaper process.
The alternative is nothing at all.
What “nothing at all” looks like
Economy of scale gets described as making things affordable, and it does. It also decides what exists.
A factory tools up for the case that recurs often enough to repay the tooling. Everything else is not produced expensively. It is not produced. The demand is real, the buyer is willing, and there is no supplier, because the volume never justified the mould.
The grand version of the idea is a parts library that a town treats as infrastructure. The ordinary version is three jobs that turn up in any week of keeping old buildings in use:
- Odd opening – Print the frame to the opening, because altering an old opening to suit a catalogue window damages the building, and bespoke joinery costs more than the window.
- Dead supplier – Print the handle, bracket or vent whose maker moved on three product generations ago, because that small part otherwise condemns a larger assembly that was fine.
- Site jig – Print the template or alignment guide for one detail on one site, because the crew will use it a few dozen times and nobody will ever tool for it.
Add the fittings matched to one person’s body and one dwelling, and the shading device correct for this facade at this latitude, and the abandoned region starts to look like most of what maintaining a building actually requires.
Why local matters more than it should
Building components tend to be bulky and low in value for their size. For that class of object, logistics is often a large share of delivered cost, and it grows with distance and with how awkward the item is to handle. A bracket is cheap. A bracket shipped individually to a remote site is a delivery with a bracket attached.
Producing on or near site collapses that share. It also collapses lead time, which for a repair is often the binding constraint rather than price. A part available next week and a part available in three months are different products, and the second one often loses to demolition.
The design side is the harder half
I would have guessed the machine was the hard part. It is the easy part.
A useful component is a definition with the dimensions and conditions exposed, so that producing the variant for this opening, this orientation, this person is a matter of supplying parameters rather than modelling from scratch. That is the difference between a printed part and a printable product line, and it is entirely a software problem.
It is the same capability I argue for in why design offices should be writing this code themselves. A parametric component library is generative design applied at the scale of a fitting rather than a building: same discipline, much faster feedback, and a better place to learn it than a building is.
The definitions also have to be shared. The value of a community parts library is that the fifth community needing a version of the same thing starts from the fourth one’s work. The economics only close if design effort amortises across everyone using it, precisely because manufacturing effort does not.
That is the part I would watch. The machines are getting cheaper on their own. The shared definitions are not getting written by anybody.
Where it fits with mass production
To be fair to the other side, because overclaiming here damages the credible version of the argument: the integrated, repeatable interior systems I argued for in affordable housing are a mass-production problem. Bathrooms and kitchens, designed once and made many times. Printing is the wrong tool for those and would make them worse and more expensive.
Printing handles what is left after that. The adaptations, the repairs, the parts for the building that is already there and is not going to be replaced.
Look far enough ahead and that residual stops looking like one. Most of the buildings that will be standing decades from now are standing today, most of them are non-standard in some way, and almost none of them are served by a supply chain optimised for new construction at the median.
If something is not for sale, that has never meant nobody wanted it. Which part on your last site did someone make by hand because no one would ever sell it?