Why Railroads Bankrupted Their Builders

Between 1865 and 1893, American investors built roughly 150,000 miles of railroad track. They built a great deal of it in places where it would never generate enough traffic to cover operating costs, let alone return capital. By the time the depression of 1893 hit, nearly a third of the nation’s railroad mileage was in receivership. The companies that built those lines weren’t stupid. They weren’t corrupt, at least not primarily. They were caught in a set of structural incentives that made individually rational decisions produce collectively catastrophic outcomes. Understanding how that happened is useful because the pattern has repeated itself, in recognizable form, at least four times since.

The core dynamic is one that economists now call a coordination failure, though the 19th-century railroad men would have recognized it by the simpler name: the race to the right-of-way. A railroad’s value depended substantially on being first. Once a company had laid track through a mountain pass, through a river valley, between two commercial cities, it controlled access to that route. Potential competitors faced a different calculation: build a parallel line and split the traffic, or cede the field entirely. The rational response for the second-mover was often to build, because even split traffic was better than no traffic, and the first mover couldn’t credibly commit to pricing the second mover out. So you ended up with two railroads between Chicago and St. Louis, then three, then four, each carrying enough freight to survive but none carrying enough to service its construction debt.

Jay Gould understood this dynamic better than anyone, which is why he spent his career not building railroads but acquiring them at bankruptcy prices. His strategy depended on the system producing bankruptcies, which it reliably did. He would buy into a struggling line, run it just well enough to prevent immediate collapse, use it to threaten the traffic of adjacent more-profitable lines, and either sell at a premium or force a consolidation on favorable terms. Gould is often described as a robber baron, a term that emphasizes his morality rather than his insight. His actual contribution was recognizing that in an industry with high fixed costs, easy capital access, and no coordination mechanism, overbuilding was structurally inevitable and the money was in the aftermath, not the building.

The investors who funded the building were not irrational by the standards of their information. Railroad stocks and bonds were the technology stocks of the 1870s: understood to represent transformative infrastructure, backed by land grants and government bonds in many cases, and genuinely producing returns in the early years when traffic was growing fast enough to service debt. The problem was that the capital kept coming even as the returns per dollar invested were clearly declining. British investors, who funded a substantial fraction of American railroad construction, had limited visibility into competitive conditions on the ground. American promoters, who had better information, had strong incentives to use it to raise capital rather than to warn capital away. The information asymmetry between builders and investors was a feature of the system, not a bug.

What’s particularly striking, looking at the period from the perspective of a full century, is how little the overbuilding was hidden. The trade press of the 1880s ran analysis showing clearly that rate wars were destroying margins, that construction in certain regions vastly exceeded any plausible traffic estimate, that consolidation was inevitable and would destroy most of the current equity value. The analysis was accurate and was widely read. Capital kept flowing anyway. The explanation isn’t ignorance. It’s that investors understood the industry was overbuilt while believing, individually, that their particular investment was the one that would survive consolidation — that they had picked the right horse in a race where most horses were going to fall.

The resolution, when it came, was the House of Morgan. J.P. Morgan spent the 1890s and early 1900s doing what he called “morganization” — imposing consolidation on bankrupt railroads, combining competing lines under unified management, establishing the cartel-like pricing stability that the industry needed to return to solvency. Morgan was explicitly contemptuous of competition as an organizing principle for industries with high fixed costs. He thought Adam Smith’s invisible hand was a fine theory for businesses where each transaction was independent, and a disaster for businesses where the cost structure required stable, predictable, long-term revenue streams. Whether or not you find his conclusions sympathetic, his diagnosis was correct.

The parallel to the first internet boom is close enough to be instructive. Between 1995 and 2001, American and European investors funded the construction of roughly 40 million miles of fiber optic cable, much of it in routes that would never carry enough traffic to generate positive returns. The logic was identical to the railroad logic: first-mover advantage, the value of controlling the right-of-way, the impossibility of credibly committing to not compete. Companies like Global Crossing and 360networks built transatlantic and transcontinental routes in direct competition with each other and with incumbents, all betting that traffic growth would outpace supply. Traffic growth did not outpace supply. The bankruptcy wave that followed in 2001-2002 wiped out roughly $2 trillion in equity value. The fiber itself, of course, is still there, still carrying internet traffic. Infrastructure built during manias often survives even when its builders don’t.

The semiconductor industry in the early 2020s followed a version of the same script, though faster and with government intervention that complicated the story. The CHIPS Act and its European and Asian equivalents created simultaneous investment surges in advanced semiconductor fabrication at a moment when the industry was already planning capacity expansions based on pandemic-era demand signals that were already reverting to trend. By 2024, there was a significant glut in mature-node chips; by 2026, leading-edge capacity was tighter than expected partly because some announced expansions were delayed after investors looked more carefully at the economics. The consolidation happened faster than in the railroad era, partly because the industry was already oligopolistic to begin with, but the pattern of coordinated overinvestment followed by painful rationalization was familiar.

What do these episodes have in common? Several things, which together constitute something close to a law of transformative infrastructure investment. First: the technology genuinely is transformative, which means early returns genuinely are good, which means capital genuinely should flow to it. This is not speculative; the railroads did transform American economic geography, the fiber did carry the internet, the chips did enable modern computing. The mania is not irrational at the object level; the infrastructure being built is real and valuable. Second: the coordination problem is insoluble within the structure of competitive capital markets. Nobody has a strong incentive to be the one who doesn’t build, because the one who doesn’t build loses. Third: the resolution requires either market concentration, regulatory intervention, or both, and it typically happens after substantial destruction of capital.

The investors who fund the building are not, by and large, the ones who profit from the outcome. The railroads that survived consolidation generated good returns for decades; the investors who owned them in 1910 did well. The investors who funded construction in the 1870s and 1880s mostly didn’t. The fiber that Global Crossing laid across the Atlantic is carrying traffic today; the owners of that fiber paid pennies on the dollar for it after the bankruptcy. Capital destruction in the building phase is not incidental to infrastructure booms. It is the mechanism that makes the infrastructure cheap enough for the next-generation users who will actually profit from it.

This is uncomfortable for capital markets ideology, which prefers a story where investment and return are tightly coupled. The uncomfortable truth is that transformative infrastructure requires massive coordinated investment that no single investor or set of investors can rationally provide in a competitive market structure, and the mechanism that actually produces that investment is collective irrationality followed by bankruptcy and consolidation. You don’t have to like this story to benefit from its outcomes. The American economy in 1900 ran on railroad infrastructure paid for by British bondholders who never got their money back. The modern internet runs on fiber paid for by Nasdaq-era equity investors who lost most of their money. Someone always pays.

The lesson that never quite gets learned is the one about the gap between building value and capturing value. The builders of transformative infrastructure almost always create more value than they capture. The railroads that linked the Great Plains to eastern markets created enormous agricultural wealth; almost none of it accrued to the railroad companies. They captured some fraction of it as freight rates, but the rate wars competed most of that away, and the rate wars were themselves a consequence of the overbuilding. The farmers who shipped grain on those lines captured far more value than the investors who built them. This is not a market failure in the conventional sense — markets cleared, goods moved, prices reflected supply and demand. It’s something more fundamental: the returns to transformative infrastructure are disproportionately social rather than private, and no private capital structure can fully capture social returns.

This is why public investment in infrastructure, which looks economically irrational from a project-level perspective — governments routinely fund infrastructure at rates of return that no private investor would accept — is actually economically rational from a social welfare perspective. The British government’s decision in the 1840s to impose some regulatory order on the railway mania probably slowed total investment; it also probably prevented some of the worst overbuilding and resulted in somewhat better outcomes for investors. The American decision to fund transcontinental railroads with land grants rather than direct capital was an attempt to solve the same problem: align private and social returns closely enough to make private construction viable without requiring the government to evaluate projects on financial criteria that would make all of them look like bad investments.

These are not ancient history problems. Every generation of transformative infrastructure investment recreates them, because the underlying economics don’t change. Fixed costs are fixed costs. Coordination failures are coordination failures. The mechanisms shift — bond markets to equity markets to sovereign wealth funds — but the structure stays the same. The railroads went bankrupt and J.P. Morgan reorganized them. The fiber companies went bankrupt and private equity rolled up the survivors. The question for any new round of infrastructure investment is not whether the pattern will repeat, but whether the institutions involved have any mechanism to exit the race before the crash, or whether they’re locked into building until the capital runs out.

Most don’t. The race to the right-of-way is hard to exit, because exiting is indistinguishable from losing until after the fact. The investors who didn’t fund fiber in 1999 looked foolish in 2000 and wise in 2002. The investors who didn’t fund early railroads looked cautious in 1875 and vindicated in 1894. But there were only a few of them, and they had to be willing to look wrong for years before being right. That’s not a temperament that gets rewarded in institutional capital management.

So we build. We build until we can’t build anymore. The infrastructure survives. The investors don’t. And twenty years later, somebody writes an analytical piece about the economics of the previous boom and how rational it all was in its own way, which is true, and how we might avoid it next time, which is probably too optimistic.

The railroads are still there. The fiber is still there. The money is gone. The ratio of created value to captured value is, and has always been, remarkably unfavorable to the people who wrote the checks.