Why US Cities Can't Just Build More Trains: The Cost Anatomy

Here's a puzzle worth sitting with: the country that built the New York subway's first 28 stations in four years (1900–1904), the Hoover Dam ahead of schedule, and the interstate highway system now routinely takes a decade and world-record sums to add a few miles of urban rail. Meanwhile, cities in France, Spain, Italy, South Korea, and Turkey — democracies with unions, environmental laws, and dense historic cores — keep extending their metros at a fraction of American prices.

This isn't a story about whether trains are good (that's a values fight this desk doesn't referee). It's a cost-accounting story: when an American transit dollar gets spent, where does it actually go? Thanks to a research effort that assembled comparable data on hundreds of projects worldwide — NYU's Transit Costs Project — plus the MTA's own unusually frank self-examination, we can now open the books.

The size of the gap

Start with the flagship numbers. Phase 1 of New York's Second Avenue Subway — 1.5 miles, three new stations, opened January 2017 — cost about $4.45 billion, which the MTA itself reports as $2.95 billion per mile: the most expensive subway segment in the Transit Costs Project's global database. Phase 2, extending the line into East Harlem, has been projected at roughly $4 billion per mile as of the mid-2020s. Against that: across 900-plus projects tracked worldwide, typical subway construction lands near $350 million per mile, with wealthy peers like the Nordic countries building far below U.S. prices — the researchers found no correlation between a country's GDP per capita and its construction costs.

New York is the extreme, but it isn't an outlier so much as the front of a national pattern: the Transit Costs Project's data shows U.S. rail projects generally clustering multiples above the international norm, coast to coast. So: anatomy time. The premium decomposes into three layers — what we dig, how we manage, and how long we take.

Layer one: we dig cathedrals

The single most concrete finding in the cost autopsies is almost embarrassingly physical: American stations are enormous. On Second Avenue Phase 1, stations and related structures consumed about 77% of construction costs — the tunneling itself was a minority expense. The 86th Street station's underground box runs about 60% longer than the platform it serves, packed with full-length mezzanines, sprawling back-of-house rooms for equipment and staff, and ancillary structures; comparable international stations add just 3–20% beyond platform length, per the Transit Costs Project's New York case study.

Why so big? Partly deliberate design choices: deep-mined caverns instead of cheaper cut-and-cover excavation (digging up the street is disruptive, and avoiding disruption is itself a political choice with a price tag), generous mechanical spaces, bespoke architecture per station. Internationally, the playbook runs opposite: standardized station designs repeated across a line, shallower digs, and a tolerance for temporary street disruption in exchange for permanent savings. Madrid — long the benchmark for cheap, fast metro building — made standardization and cut-and-cover the core of its method.

Layer two: the coordination overhead

The second layer is white-collar. Phase 1 spent about $656 million on consultants — design, engineering, and construction management — because U.S. transit agencies, after decades of building rarely, have hollowed out their in-house engineering and must rent expertise project by project. The Transit Costs Project found agencies that build cheaply (Madrid, Milan, Istanbul, the Nordics) keep strong permanent technical staffs that manage contractors tightly, standardize designs, and carry lessons from one project to the next. Build rarely → lose staff → hire consultants → costs rise → build even more rarely: a self-tightening loop.

Procurement adds its own tax. American agencies have tended toward oversized, bespoke contracts that few firms can bid on — and thin competition raises prices; the researchers documented U.S. bids drawing a fraction of the bidders common abroad, with contractors pricing in the risk of working under litigious, change-order-heavy conditions. Interagency friction piles on: utilities relocated defensively, police details, betterments (unrelated wish-list items attached to the project because the money is passing by), and every affected agency holding an effective veto that must be bought out in design concessions or cash.

And then there's process time. Federally funded projects route through environmental review (NEPA) and the Federal Transit Administration's multi-year Capital Investment Grants pipeline; overlapping state and local reviews, community litigation risk, and stop-start funding stretch schedules. Time is not a neutral ingredient: every added year means more staff-years of overhead, more inflation, more financing cost, more chances for scope to grow. The GAO's recurring reviews of federal transit megaprojects return again and again to schedule as a cost driver. If this rhymes with why America is millions of homes short — many veto points, each locally reasonable, compounding into national scarcity — the rhyme is the point: it's the same institutional physics applied to different concrete.

The tunnels are cheap. What's expensive is everything America wraps around a tunnel: rooms, reviews, risk, and years.

A note on vocabulary, because it hides a lot of money: engineers split project budgets into hard costs (concrete, steel, labor — the physical build) and soft costs (design, engineering, management, insurance, financing, and administration). The Transit Costs Project found soft costs on American projects running around a fifth or more of budgets — roughly double the share common abroad — which means a meaningful slice of the U.S. premium is spent before and around the digging, not on it. "Betterments" deserve their own line in the glossary: when a transit project passes through, other agencies and utilities often condition their cooperation on upgrades they've long wanted anyway — a rebuilt water main beyond what the tunnel disturbs, a repaved street, a new substation — and the transit budget absorbs them, because the alternative is delay, and delay costs more. Each betterment is defensible; the habit, summed across a decade-long project, is a tax with no name on it.

What it isn't (mostly)

Honest anatomy requires ruling things out. It isn't mainly geology — Manhattan schist is famously good tunneling rock, and peer cities dig through worse. It isn't simply "unions" — the cheap builders in France, Italy, and Scandinavia are heavily unionized; though the Transit Costs Project did document New York-specific staffing practices (crews far larger than international norms for the same tunnel-boring work) that inflate labor costs, which is a work-rules story more than a wage story. It isn't safety regulation per se — European safety regimes are comparably strict. And it isn't uniquely public-sector incompetence — U.S. private megaprojects share some of the premium, though less extreme. The differentiator the data keeps pointing back to is institutional: agencies' in-house capacity, standardization, procurement design, and the political willingness to impose short-term disruption for long-term savings.

One more exclusion matters for reading headlines: this is a construction cost story, not an operating cost one. A line that costs too much to build can still be cheap to run, and vice versa; the two problems have different mechanics and different fixes. And the premium isn't confined to subways — the Transit Costs Project and the Eno Center for Transportation have both documented American light rail and even bus-rapid-transit projects pricing above international peers, which is telling: bus lanes involve no tunneling, no Manhattan geology, and no cathedral stations, yet the process layer — consultants, reviews, betterments, time — travels with the funding structure, whatever the vehicle.

The stakes, and the experiments

Why does a per-mile number deserve 1,800 words? Because cost is policy. At $350 million a mile, a billion dollars buys a meaningful extension; at $3 billion a mile, the same money buys a study and a shortfall. High costs quietly decide that most American transit proposals die in spreadsheets — no vote required. Whatever one thinks transit should get (a genuinely contested question — highway and transit advocates have argued over federal funding shares since the Interstate era), everyone's dollars stretch further or shorter depending on this machinery. The same logic applies federally: the FTA's grant programs, funded through the congressional appropriations machinery, effectively price how many projects the country can attempt per decade.

The encouraging part of the story, as of 2026: the problem has moved from ignored to studied to actively worked. The MTA publicly benchmarked itself against global peers and reorganized its construction arm, reporting double-digit percentage cost reductions on recent tunneling contracts; the FTA and Congress have funded project-delivery research; and several agencies have begun importing the international playbook — standardized stations, earlier utility agreements, larger bidder pools, design-to-budget mandates. Whether the U.S. premium actually shrinks is an empirical question the next decade of Census-of-concrete will answer; the mechanism to watch is simple: station size, bidder counts, and years-to-open, project by project.

Until then, the anatomy holds. When you hear that some American city "can't afford" a train, translate it: the city is paying for stations dug like cathedrals, expertise rented by the hour, risk priced by wary bidders, and time — years and years of time — at compound interest. Other rich democracies write smaller checks for the same tunnels. The difference isn't the rock, and it isn't the trains. It's everything in between, and every line of it is now, finally, itemized.

This explainer is written to stay accurate over time. Facts and figures were verified against the primary sources listed above as of August 22, 2026. If you spot an error, our corrections policy explains how we fix it.