Why factory housing has not scaled

The claim, and the count

Factory-built housing is presented, reliably and in every decade, as the technology about to solve the cost of housing. The pitch is durable because the logic is genuinely appealing: move the work indoors, and construction becomes manufacturing, with the productivity gains manufacturing has delivered everywhere else.

The United States has counted the result since 1992, and the count is not ambiguous. In 2025 the country completed 1,005,000 single-family houses. Modular was 14,000 of them, and the panelised and precut category was 16,000. The remaining 975,000, which is 97 percent, went up on site. [1]

The trend is worse for the pitch than the level is. The high water mark was 2002: 46,000 modular units, in a year of 1,325,000 completions, which put the share at 3 percent. They have not exceeded 3 percent since 2007 and have rounded to 1 percent every year since 2018. Across three decades in which building science improved, labour got scarcer and more expensive, software transformed every adjacent industry and the case for building indoors strengthened on every axis, the share of American houses built in factories fell by roughly two thirds.

Something is holding this down that is not technology, and it is not new.

A federal audit already answered this, in 1976

In 1969 the Department of Housing and Urban Development launched Operation Breakthrough. Its purpose, in the words of the audit that followed, was to improve the process of providing housing for lower income families by demonstrating the value of industrialized factory built housing construction methods and by eliminating or reducing barriers to industrialized housing construction. [2]

It was not a small effort and it was not naive. It funded prototype systems, built demonstration sites, and pushed on the regulatory obstacles that fragmented the market. Much of what it pushed on, statewide building codes in particular, moved.

The General Accounting Office reported on it in November 1976. Its finding was that the programme did not prove the marketability of most of its sponsored housing construction methods, and it named three reasons for the outcome: unexpected decreases in the housing market, the suspension of HUD subsidised mortgage housing programmes that had been intended as the market for the output, and a lack of cost savings potential in some of the systems. [2]

Read the first two together, because they are the same reason stated twice. The programme could not create a market large enough and steady enough to support the production volumes that factory construction requires to be efficient. The technology was not the binding constraint. Demand was.

Why volume is the whole problem

A factory is a fixed cost. A building, equipment, a trained workforce and a management structure exist whether or not anything is being produced, and they have to be paid for out of what is produced. The economics only work above a threshold volume, and below it the fixed costs are spread across too few units and the factory loses money on every one.

A site-built operation has almost the opposite cost structure. Its capacity is subcontractors, and subcontractors go home when work stops. A builder that completes eighty houses this year and thirty next year does not carry the cost of the missing fifty.

Now put that against the demand pattern in the table above. Between 2006 and 2011, national single-family completions fell from 1,654,000 to 447,000, a drop of nearly three quarters in five years. [1] A site-built builder shrinks through that. A factory cannot: it either runs far below the volume its economics require, or it closes.

That asymmetry is the mechanism. Housing demand is deeply cyclical, and factory production is the business model least able to survive a trough. Every downturn removes plants, and the plants do not come back when demand does, because the capital required to rebuild capacity has to be justified against the next trough rather than the current peak. The 1976 audit called this an unexpected decrease in the housing market. Fifty years of data suggest the decrease is a permanent feature and the surprise is the thing that needs explaining.

Where it works, which is the part that gets ignored

The national share conceals a real and stable success, and it is visible only when the completions data is split by who is building.

The Census Bureau separates houses built for sale, which a builder puts up on its own land and sells finished, from contractor-built houses, which an owner commissions on land they already hold. [3] In 2025 those were 721,000 and 130,000 houses respectively.

In the for-sale market, modular units came in under 500, a figure the Bureau flags rather than prints. In the commissioned market, 8,000 of the 130,000 houses were modular, a share of 6 percent, and in the Northeast the same measure reached 16 percent. [1]

That is a factor of more than twenty between two segments of the same national market, and it points directly at what the method is actually good for. An owner building one house on one lot wants a fixed scope, a fixed price and a defined completion. A factory delivers all three, and the alternative on offer is a general contractor pricing a bespoke job with local subcontractors. The value is real and buyers pay for it.

A production builder wants none of those things. It already has crews, repeat plans and land inventory. Its competitive advantage is flexibility: changing plan mix as the market moves, phasing to match absorption, adjusting specification to hit a price point. A factory takes that flexibility away and adds a fixed set date and a supplier with its own capacity constraints. The exchange is a bad one, and the for-sale numbers say builders have been making that judgement consistently for thirty years.

The three obstacles that get blamed, ranked by how much they explain

Regulatory fragmentation is the usual first answer: a factory serving several states faces several sets of adopted codes and amendments, which limits standardisation. This is real and it is the one Operation Breakthrough attacked hardest. It is also the one that has improved most since, and modular share fell anyway. It is a cost, not the binding constraint.

Financing is the second, and it is stronger. A construction lender advances against work in place on the security. A factory needs payment before anything exists on the borrower's land. That mismatch is structural rather than a matter of lender conservatism, and it falls hardest on exactly the buyer segment where modular otherwise works.

Transport and geography is the third and the most underrated. Modules are large, and the cost and permitting of moving them rises with distance, which caps the radius a plant can economically serve. That radius has to contain enough demand to fill the plant, which is why modular share is highest in the dense Northeast and lowest in the West, where the same radius encloses far fewer houses. It is also why a national modular strategy is harder than it looks: the business is regional whether anybody wants it to be or not.

What would actually change the number

Not a better factory. The output of American modular plants is not the problem, and the audit reached that conclusion in 1976 when the plants were considerably worse than they are now.

What would change it is demand that does not collapse. A buyer who commits volume across a cycle, rather than in a peak, changes the calculation that closes plants in a trough. That is why the interesting recent activity in offsite construction is in segments with institutional buyers, in multifamily and in public and social housing, rather than in detached houses sold one at a time. Those buyers can commit to volume, and volume is the input the model has always been short of.

It also implies something uncomfortable for the pitch itself. If the constraint is demand stability rather than technology, then a factory housing venture is not really a manufacturing business with a construction problem. It is a demand aggregation business with a factory attached, and the ones that fail are usually the ones that had the second half right and the first half missing.

Fifty years of federal data and one federal audit point in the same direction. The reference sheets on The Build Sheet are written on that basis, which is why they describe what factory construction is good at rather than what it is going to become.