We Asked a Renowned Financial Historian What Investors Get Wrong About Bubbles

Railway mania set the template

UK railway capex hit 10% of GDP in the 1840s with duplicative lines (three routes between London and Peterborough), causing railway stocks to crash 60% despite the technology's long-term economic benefits.

Infrastructure builders rarely win

During the dot-com bubble, telecom companies funded the heavy infrastructure spending but weren't the eventual winners, while even Amazon fell over 90% before recovering, proving early winners are nearly impossible to identify.

Disrupted assets outperform during crashes

Canal stocks actually outperformed railway stocks between 1845-1850 because they were already beaten down while railway valuations had further to fall.

Prisoner's dilemma drives overinvestment

Companies invest massively in new technologies because failing to participate risks letting competitors achieve monopolies, even though collective overinvestment destroys returns for the entire sector.

AI arms race follows familiar pattern

Microsoft's partnership with OpenAI to challenge Google's search monopoly triggered a competitive response forcing hyperscalers to commit to 'rather go bankrupt than lose' levels of spending.

Low barriers guarantee fragmentation

Unlike telephones which consolidated into monopolies early, technologies with low entry barriers (autos had 2,000 US companies, aircraft were built in Omaha) attract excess capital that competes profits away.

Demand is consistently overstated

Railway traffic was forecast to triple in 5 years during the 1840s bubble, while dot-com era claims that data traffic doubled every 2 months (versus actual 6 months) drove massive fiber overbuild and bankruptcies like WorldCom.

Extended depreciation masks true costs

AI chip depreciation schedules have doubled from 3-4 years to 6+ years, temporarily inflating reported profits while technological obsolescence continues, setting up future write-downs if demand disappoints.

Earnings depend on continued misallocation

Current strong tech earnings rely on sustained investment, but if demand falls short, the shift from expansion to depreciation recognition will trigger profit collapses similar to the 2000 telecom bust.