A large share of what a house costs was never about what it is made from. It is about how we decided to put it together, and that share is the one nobody is optimising.
Land, zoning, finance and supply are real, and this post leaves them to the policy people. The question architecture has mostly declined to ask is how much of the cost is a consequence of how we design and assemble the thing, and I think the answer is more than the industry finds comfortable, and most of it is only reachable through computation.
The cost of a house is a stack, not a number
Attacking “the cost of housing” fails because it is a stack of loosely coupled quantities rather than one: land, structure, envelope, services, interior fit-out, on-site labour, the interfaces between trades, finance during construction, and, across the building’s life, energy and maintenance.
That structure tells you what kind of win is available. No single intervention halves the total. But a stack of ten items, each improved by a modest percentage, compounds. Ten percent off six of them is not ten percent off the house.
This is exactly the shape of problem that rewards optimisation and punishes intuition. Each layer has its own physics, its own constraints, and its own trade-offs against the others. Thicker insulation costs material and buys operating cost. Deeper structure costs volume and buys span. A human designer can hold two or three of those relationships at once and reason well about them. Nobody holds ten simultaneously, which is why the industry optimises them one at a time, in sequence, and captures a fraction of what’s there.
It’s the same reason dependency order beats a ranked list in any other kind of work. The items aren’t independent, and treating them as independent throws away the part that compounds.
Why this couldn’t be done before, and can be now
Design optimisation in building was technically possible for decades and practically absent, because each iteration cost a person’s week, so an office evaluated three or four massing options and called that the space. I have written about why that barrier just moved, and the argument below assumes it has: thousands of variants evaluable against real objectives, with the professional still setting the objectives and picking the answer.
The part nobody is optimising: the interior
The structural and envelope arguments are at least understood, even where they aren’t applied. The interior is where the unexploited value sits, and the reason is economic rather than technical.
A house interior is assembled from components manufactured independently, by different firms, to different standards, and joined on site by different trades. The kitchen isn’t designed with the wall. The wall isn’t designed with the services. The floor meets the wall at a junction neither party specified, and both assumed the other would handle.
Every one of those meetings is a tolerance stack, a coordination cost, a sealant line, and a future failure.
Architecture has almost no economy of scale in exactly the place where scale should be available: the functional interior items that appear in every dwelling and could be designed once, properly, by someone paying attention.
What Apple actually did, and why it applies here
The instructive part of Apple’s integration is that designing the components together let them delete interfaces, and the vertical supply chain is a side effect of that.
A unibody enclosure is fewer parts, so there are fewer joins, less to misalign, less to loosen, less to fail, and less to inspect. The quality comes from the interfaces not being there.
Housing has the same opportunity and treats it as impossible, because the fragmentation is assumed to be the nature of the industry rather than a choice somebody made. Design a bathroom as a system (surfaces, drainage, ventilation, services, storage and lighting resolved together) and a long list of recurring defects stops being managed and starts being absent.
The gains are the ones occupants actually feel. Fewer things that go wrong. Less to clean. Less to maintain. Less that degrades visibly in the first five years, which is the period during which people decide how they feel about where they live.
The simplest case: get rid of the internal corners
If I had to demonstrate the whole argument with one detail, it would be internal corners.
Look at any room that’s been lived in for a decade. The lines where surfaces meet have gone dark. Skirting to wall, wall to ceiling, around the frames, along the worktop. Those lines were crisp at handover and now they read as grime, because that’s exactly what they are.
The mechanism is five separate problems stacked in one geometry:
- It’s a low-velocity zone, so airborne dust settles there and isn’t disturbed
- It’s a capillary geometry, so moisture wicks in and stays
- It concentrates cleaning contact, so finishes wear differentially along the line
- It’s a movement joint between two elements that expand at different rates, so the sealant cracks and the crack fills
- It’s a shadow line, so the eye is drawn to it before anything else in the room
The result is that a building’s perceived age is carried disproportionately by its junctions. Not the surfaces. The lines between them.
The remedy is known and in daily use elsewhere. Laboratories, hospitals and food production facilities have coved junctions as standard, because in those buildings a dirt-trapping corner is a contamination risk rather than an aesthetic one. The geometry isn’t exotic: it’s a fillet, and it eliminates the settling zone, the capillary path and the shadow line in a single move.
Housing doesn’t do it, almost entirely because the trade structure makes a continuous junction between two separately-supplied elements difficult to procure. Design the elements together and it stops being difficult.
That is one detail, and there are many like it, individually unremarkable and collectively enormous.
What the 3D printing activity actually tells us
There’s a lot of construction-scale printing work at the moment, and it’s worth reading honestly rather than as a signal that the problem is being solved.
The activity concentrates where the margin is: expressive facades, feature elements, prestige projects, demonstrators. Very little of it concentrates where housing need is greatest. That is a market doing what markets do, allocating a novel capability toward the products where it earns most.
Which means waiting for that activity to reach affordable housing on its own isn’t a plan. The capability arrives there when someone deliberately aims it. And the interesting near-term application is printing the low-volume, high-variety components mass production will never serve: the complement to the integrated, mass-produced interior above, not a replacement for it.
The horizon
None of this arrives at once. It’s a couple of decades of layers landing gradually (structural optimisation, envelope tuning, integrated interior systems, local manufacture of the parts that don’t justify a factory), each worth a few percent, with the combination worth considerably more than the sum suggests.
The precondition is that design iteration becomes cheap enough to explore properly. That stopped being a technical barrier recently enough that most of the profession hasn’t noticed yet.
The uncomfortable implication for architects, myself included, is that affordable housing looks the way it does partly because we designed at the resolution the tools permitted.
We called the tool’s resolution a constraint, and the constraint was ours. Which junction in your last drawing set did nobody design?