Somewhere between the modular factory tour and the AI permitting demo, an old promise is being repeated. Build faster, build cheaper, and housing becomes affordable. The promise is a century old, and it’s being made again this year with fresh conviction. The complicated part is where the spotlight is aimed: the promised land of cheap builds keeps colliding with a mechanism that most of its promoters have never priced.
For an architect betting on efficiency to bring volume and relevance, that collision is worth understanding before the bet is placed.
The Bottom Line
Production savings in constrained housing markets flow to land bids and developer margins. Prices in those markets are set by the scarcity of permitted locations rather than by the cost of raising a building on them, so a cheaper building mostly results in a more expensive site.
Instead, automated plan reviews, digital permitting, and code standardization are technological layers that could actually lower housing prices for consumers. That layer is being refined in Singapore, Vienna, Los Angeles, and Minneapolis, but it brings its own issues for practitioners, such as drawing IP, liability, and the interpretation of the code itself. Even there, the effect is conditional. Automated review, for example, while merely accelerating the development of a fixed stock of permitted sites, delivers a productivity gain for developers and municipalities, thereby increasing potential supply.
The Mechanism
Developers price sites by working backwards. Expected sale value minus construction cost minus profit equals what they can bid for the land. When construction costs fall, that residual rises, and competing developers bid it straight into the land price. The savings arrive in the seller's pocket long before any tenant sees a lease.
The long-run record shows how completely land dominates this arithmetic. Real house prices across fourteen advanced economies have roughly tripled since 1950, and about 80 percent of that increase traces to land, while construction costs have gone sideways. American construction productivity has been flat since 1948, with land-use regulation identified as a principal cause.
Los Angeles now offers the mechanism in its purest form: a market where the permission itself carries a price tag. Around one in five land listings in the city comes with ready-to-issue permits attached, and a new MIT working paper by Evan Soltas and Jonathan Gruber finds developers pay a 50 percent premium for that pre-approved land, about $48 per square foot, with permitting alone explaining roughly one third of the gap between the city's home prices and its construction costs. On a typical site, the approval is worth around $770,000. No factory saving competes with that number.
Prefabrication has tested this arithmetic before. Manufactured homes peaked at roughly one-third of new American single-family homes in the early 1970s, in what the Minneapolis Fed calls the first and only manufactured housing boom, and factory-built housing now sits in the low single digits of the market.
Tokyo and Sweden run the two control experiments. Tokyo pairs industrialized building with permissive zoning and fast approvals, grows its housing stock by close to two percent a year, roughly double London or New York, and keeps prices comparatively stable. Sweden industrialized the factory and left the constraint standing: around 84 percent of Swedish detached homes use prefabricated elements and construction productivity sits above the OECD average, yet the OECD's 2025 survey still finds persistent shortages in Stockholm, Gothenburg and Malmö, identifies land value as the main driver of prices, and aims its recommendations at zoning and permitting speed rather than at the factories.
Market Signals
Denver builds the experiment twice
West Holden Place, the largest modular multifamily project in Denver, claims 20 to 25 percent construction savings, and rents a one-bedroom at $1,295 against a metro median around $1,615. The assumed headlines have been that modular housing breakthroughs decrease housing prices, making accommodation more affordable. However, the multifamily project units also carry income restrictions and the project is financed through city and state affordability programmes, so the below-market rent is engineered by subsidy as much as produced by the factory, and the developer told the Denver Gazette that Colorado's apartment market now confronts builders with challenges that have little to do with construction costs. He is right: after a decade of record conventional supply, metro vacancy reached a 16-year high of 7.6 percent and average rents fell 4.8 percent. The final word hasn’t been said on the project effects, but the well-known “correlation vs causation”-problem could lead to faulty assumptions.
Minneapolis moves the constraint
Minneapolis paired the end of single-family-only zoning with the repeal of parking minimums and corridor upzoning under its 2040 plan. Pew finds that the city grew its housing stock by 12 percent from 2017 to 2022, while rents rose by 1 percent, compared with 4 percent stock growth and 14 percent rent growth in the rest of Minnesota. The heavy lifting came from apartments along transit corridors rather than from the celebrated duplex reform, which has added few units.
Singapore treats permission as infrastructure
CORENET X, a roughly S$450 million national platform integrating 16 regulatory agencies, uses BIM models as the submission format and runs automated code checks at the country scale. It is the mature precedent: a state handling permission processing the way it handles ports and power.
Vienna teaches the code to a machine
The EU-funded BRISE project trained an AI on the Vienna building code to pre-check BIM submissions against a baseline in which a permit took up to a year, with the city projecting that approval times would be cut by up to half. The project's own journal warns that the model cannot simply be copied: Prague, Berlin or Zurich would each need an individually trained AI, and the realistic scaling path runs through models built at the regional or national level.
Los Angeles pre-checks the rebuild
After the Eaton and Palisades fires, California deployed Archistar's voluntary AI plan check for rebuilds in Los Angeles city and county, and the state reports average staff review time down 54 percent for like-for-like rebuilds submitted with the tool. The vendor now publishes its own rebuttals to concerns about IP, opacity and data security, while a quoted architect concedes the contractual frameworks are still to be written.
Intelligence Brief
The practical question is how a practice reads project volume correctly and prices its exposure honestly.
On volume, watching the permission layer is important to understand upcoming demand. Published permit turnaround statistics, pre-check adoption announcements, and zoning reform dockets in the jurisdictions, when read together, say more about near-term workload than any factory's regional capacity. Automation arriving alongside zoning, as in Minneapolis, signals supply and fee volume. Automation arriving within a fixed zoning envelope signals faster answers for the same number of projects, and the Los Angeles permit market suggests that part of even that gain capitalizes into land. A faster “no” is still a “no”.
The exposure sits on three fronts. One front, drawings and models submitted through third-party review platforms, raises IP issues and data security questions that neither contract templates nor case law has settled, as noted in Common Edge's analysis of AI liability. The stamp retains liability even when the model passes the plan, and no insurer has yet been tested for a failure traceable to a missed automated check. Resolving those questions in writing before the first submission costs an afternoon. Being the test case costs considerably more.
Second, the jurisdiction lock is structural. Every reviewer model is configured against one base code with its local amendments, so non-standard assemblies, bio-based systems, and anything the training model underrepresents will encounter friction at machine speed rather than a conversation with a code official. Documenting code intent explicitly at submission and requesting human review when the model returns ambiguous flags need to become standard practice; automated review might otherwise negatively impact, e.g., vernacular construction.
Third, scope. If review automates, standardization pressure moves upstream into design, and the margin on producing compliant volume compresses. The margin on authoring systems, writing the pre-design brief, and translating messy client and community intent into permit-legible form holds far better. That is the same direction as my previous post, on the pre-design turn, pointed to previously.
Modular construction has genuine advantages: speed, cost predictability, and hedging against site labour shortages, all of which are worth more when financing remains generous. Those benefits buy buildable projects; however, I would argue that affordable end prices are bought somewhere else entirely, at the permission counter, and the firms watching that counter will read the upcoming development better than the firms watching the factory gate.
After researching this weeks topic, it’s clear that the perspectives on modular constructions economic impact is divided. If you disagree with this post, you are welcome to drop a line below and tell me your take! And, as always, see you next week!
-Johan