Stalled Data Centres and the Railway Bust of 1873
![empty formal interior, natural lighting through tall windows, wood paneling, institutional architecture, sense of history and permanence, marble columns, high ceilings, formal furniture, muted palette, A vast, disused legislative chamber, marble floor cracked and veined with creeping ivy, where a rusted railway rail emerges from beneath a pile of dusty blueprints and abruptly fractures into a tangle of severed glass-fibre cables, sunlight streaming horizontally through tall, grimy windows, casting long shadows that slice across the room like timelines derailed, atmosphere heavy with silence and suspended consequence. [fal-ai/z-image/turbo] empty formal interior, natural lighting through tall windows, wood paneling, institutional architecture, sense of history and permanence, marble columns, high ceilings, formal furniture, muted palette, A vast, disused legislative chamber, marble floor cracked and veined with creeping ivy, where a rusted railway rail emerges from beneath a pile of dusty blueprints and abruptly fractures into a tangle of severed glass-fibre cables, sunlight streaming horizontally through tall, grimy windows, casting long shadows that slice across the room like timelines derailed, atmosphere heavy with silence and suspended consequence. [fal-ai/z-image/turbo]](https://cdn.digitalrain.dev/thelongview/viral-images/24c2199e-a5ef-49f0-964b-db6a9ee6780a_viral_2_square.jpg)
Data-centre investment has grown tenfold since 2019, but local political friction has so far altered approvals, not cancellations. The 1873 railway collapse was triggered by capital withdrawal; today’s headwinds are procedural. Without evidence of sustained spending reductions, the recession link remains conditional.
In 1873, European investors who had bought large amounts of American railway bonds pulled their money out after a financial crisis at home, and a railway boom that had laid more than 33,000 miles of track since 1868 gave way to a recession lasting five years. A Hong Kong columnist has placed that episode at the centre of an argument about the present. He contends that the Democratic wave in the US midterms, which pushed the semiconductor index down more than 3 per cent intraday before it recovered, will halt enough data-centre projects to tip the economy over.
The scale he gives explains why the argument carries weight. Data-centre construction spending ran at close to $100bn on an annualised basis in September, ten times its level five years earlier, and AI-related jobs rose by 750,000 over the same period. That $100bn covers land preparation, civil works and electrical fit-out only. The columnist says every dollar of civil spending brings $3 to $5 of investment in chips, AI racks and software, which puts total data-centre investment at about $600bn a year. He cites Goldman Sachs' estimate that AI investment contributes 1.9 per cent to economic growth in 2026, set against brokers' forecasts of 2.3 per cent growth this year and 2.2 per cent next. On his reading, AI investment now takes a larger share of GDP than at any point since the railway construction frenzy of the 19th century.
That earlier boom is the parallel he relies on. Railways became the largest employer after agriculture and absorbed almost all of the country's steel, timber and machinery capacity. Banks and financial institutions channelled household savings into railway bonds, and those bonds depended on European buyers. When that capital left in 1873, the bubble burst. The resemblance is in the concentration: a single category of investment carrying a large part of the economy, financed through a narrow group of buyers.
The break comes in the trigger. The columnist's chain starts with local politics, not with capital leaving. Opposition to data-centre construction helped Democrats win office in the main data-centre hubs, from governors down to local councils. But those officeholders also answer for local jobs, tax revenue and the economy, so in practice they have tightened approvals, increased procedural transparency and bargained with developers for community benefits and commitments on utility bills. Developers have paid and invested in public relations, and projects have gone ahead. His conclusion is that politics is mainly a matter of distributing benefits, so the sell-off in chip stocks is a delay, and investors will return once projects clear.
Two other threads in the piece bear on that conclusion. The columnist suggests that fewer projects could let the large cloud providers raise their prices. He argues, however, that compute costs are rising, and that the pressure falls on model developers such as Anthropic and OpenAI, whose token prices keep falling while the cost of compute for training models climbs. He calls that an unhealthy environment. He also points to Masayoshi Son, who has put tens of billions of dollars into OpenAI and joined a $500bn compute infrastructure plan. SoftBank, which the columnist describes as a proxy for OpenAI, has seen its share price halve from its June high.
What the argument does not yet establish is the link between a slowdown in approvals and a fall in spending large enough to show up in GDP. The 1873 collapse was a sudden withdrawal of capital. The columnist's own account describes a process of delay, in which projects restart once local arrangements are settled, and no figure is yet available for how many projects have been stalled, cancelled or merely postponed. The recession scenario therefore rests on the direction of political pressure, not on observed cancellations. The question the parallel leaves open is how long the delay would have to last before it reached the spending figures.
† Analysis: the writer's conclusion from the reported facts, not a fact stated in the source.
- The author infers that because the 1873 collapse was triggered by a sudden withdrawal of foreign capital, the current data center situation—characterized by political delays rather than immediate capital flight—lacks the same immediate recessionary trigger, suggesting the recessionary risk is theoretical rather than imminent.
- The author concludes that the sell-off in chip stocks is likely a temporary market reaction to political uncertainty rather than a long-term structural failure, based on the premise that developers are successfully navigating local political requirements through public relations and community concessions.
Published November 15, 2026