The canal that halved the price of coal
The Duke of Bridgewater’s canal carried coal from his mines to Manchester. Once it opened, the price of coal halved, and the Duke’s profits soared (Federal Reserve Bank of New York, Liberty Street Economics, 2014). A horse could pull about a ton on land but up to 30 or even 50 tons on a boat (Federal Reserve Bank of New York, Liberty Street Economics, 2014).
From 1 canal to 21 a year
Parliament approved one new canal in 1790, six in 1791, seven in 1792 and twenty-one in 1793. Grand Junction Canal shares jumped from £100 to £472.75 in October 1792 and were back near £100 by 1795 (Federal Reserve Bank of New York, Liberty Street Economics, 2014). Some canals earned profits for years; others were abandoned or never started (Federal Reserve Bank of New York, Liberty Street Economics, 2014).
Railway Mania followed in the 1840s. In 1846 alone, Parliament authorized 4,538 miles of new line (Odlyzko, University of Minnesota, 2010). Investors suffered heavy losses when the boom broke (Odlyzko, University of Minnesota, 2010).
The asset outlived the bubble
Across the Atlantic, the Erie Canal opened on 26 October 1825. Between 1815 and 1825 the cost of shipping wheat from Buffalo to New York City fell from about $100 to $10 a ton (New York State Museum). The roughly $7 million canal paid for itself in tolls within about nine years (New York Heritage, Erie Canal at 200).
The NY Fed’s write-up compares canal overbuilding to the fiber-optic glut of the dot-com bust: overinvestment that looked premature but “perhaps prudent in the long run” (Federal Reserve Bank of New York, Liberty Street Economics, 2014).
What the popular version gets wrong
Popular version Busts wiped out the value.
What the record shows They wiped out many shareholders, not the infrastructure. Canals and rail lines kept working after the price crashes (Federal Reserve Bank of New York, Liberty Street Economics, 2014) (Odlyzko, University of Minnesota, 2010).
What it means for today
AI data centers are this era’s canals. McKinsey estimates about $6.7 trillion of data-center capital spending will be needed by 2030 (McKinsey, 2025). History says expect overshoot somewhere, and favour the companies that own the scarce inputs (chips, memory, power) and the customers. See Superintelligence.
The Future: Superintelligence →
Companies we own that connect to this story (tickers only, never position sizes):
AI accelerators.
High-bandwidth memory.
Powered data-center sites.
On-site power.
AWS, among the largest data-center builders.
Building its own AI data-center capacity at large scale.
Sources
- Federal Reserve Bank of New York, Liberty Street Economics, 2014
- New York State Museum
- New York Heritage, Erie Canal at 200
- Odlyzko, University of Minnesota, 2010
- McKinsey, 2025
Last reviewed: 22 Oct 2026. Historical figures are quoted as the sources publish them; where sources disagree, we say so.
Not investment advice. History is a lens, not a forecast. This page is opinion and education, researched and presented with the help of AI analyst personas, and may contain errors. The EarlyConviction team holds positions in NVDA, MU, IREN, BE, AMZN, META and may buy or sell at any time. Read the disclaimer.