Why construction productivity keeps falling — and what it would take to fix it

Summary

Construction productivity has declined persistently not because of regulation alone, but because the industry structurally blocks the feedback loops that drive efficiency in other sectors. That is the central argument of Brian Potter, senior infrastructure fellow at the Institute for Progress and author of Origins of Efficiency (Stripe Press), whose thinking is shaped by fifteen years as a structural engineer and a front-row seat to one of the sector's most expensive failures.

The Katerra Autopsy

Potter joined Katerra, a factory-based construction startup, during its peak and left roughly a year before it declared bankruptcy after burning through approximately $2.3 billion in venture capital, much of it from SoftBank. The company's thesis — that applying electronics-manufacturing discipline to building construction would compress costs — was not wrong in principle, but it was incomplete. A long graveyard of predecessors had attempted the same "Henry Ford of housing" pitch and failed. Potter's conclusion is that the structural barriers to efficiency in construction persist regardless of who is running the company.

What Actually Drives Efficiency

Across dozens of industries, Potter identifies a consistent pattern. Cost declines require high, repeatable volume; opportunities to iterate on process; and the ability to transfer learning back into production. Manufacturing nails, shipping containers, and Model T automobiles all followed this arc. Construction blocks nearly every one of these mechanisms. Projects are bespoke, sites are non-repeatable, and each build starts largely from scratch.

Regulation is a contributor but not the whole story. Construction cost curves are rising in countries with radically different regulatory regimes, pointing to deeper structural constraints around customization and scale.

Solar vs. Nuclear as a Case Study

The same framework explains energy cost trajectories. Technologies that can be manufactured at very high volumes with minimal site-specific customization — solar panels and batteries — have followed steep downward cost curves with no visible floor. Nuclear reactors, built in small numbers with heavy site-specific engineering, sit in the expensive quadrant and show no comparable learning curve. Potter is bullish on solar and storage capturing a large share of U.S. energy supply, contingent on regulatory interference not disrupting buildout momentum.

Data Centers and the NIMBY Risk

The current data center construction boom is historically unprecedented in scale, with data center capital spending now exceeding office building construction. Potter is cautiously optimistic that the execution mindset being developed — building large physical infrastructure quickly — could transfer to other sectors. The risk is political. Virginia, historically the most permissive major data center market, is seeing organized resident opposition for the first time. Potter flags this as a leading indicator of a broader NIMBY mobilization that could constrain the buildout.

Tacit Knowledge as a Strategic Asset

Potter endorses the argument, recently surfaced in writing by Dan Wang, that industrial process knowledge is effectively irreplaceable once lost because it lives in people and relationships rather than in documented systems. The learning curve in manufacturing is inseparable from physically doing the work. His illustration is Wernher von Braun during the early U.S. space program. Von Braun warned military leadership that the Navy's rocket would fail and his would succeed. Leadership ignored him, the Navy rocket exploded on the pad, and the U.S. eventually launched its first satellite using von Braun's design. Both the Soviet and U.S. early rocket programs were built substantially on accumulated German World War II rocket expertise — process capital that could not simply be replicated from a blueprint.

The policy implication for investors and executives is direct. Offshoring low-margin manufacturing does not just shift production; it permanently erodes the process knowledge base that future innovation compounds on.