``` Learning from history, when does a capital expenditure boom turn into a bursting bubble? ```
From 19th-century railroads to 21st-century artificial intelligence, every major technological innovation in history has triggered a capital expenditure boom, but the frenzy often ends with the bursting of a bubble.
In a special report published in November this year, “When Capex Booms Turn Into Busts: Lessons From History,” BCA Research reviews four typical capital expenditure booms, reveals the core logic of boom to bust, and issues a warning for the current AI frenzy.
The report summarizes five common laws: investors ignore the S-shaped curve of technology adoption, revenue projections underestimate price declines, debt becomes the core reliance for financing, asset prices peak ahead of investment downturn, and capital expenditure collapses and economic recessions intensify each other. These patterns are already emerging in the current AI sector—technology adoption rates stagnate, token prices plummet over 99%, corporate debt surges, GPU leasing costs decline.
Based on historical comparative analysis, BCA Research concludes: The AI boom is following the historical path of bubbles, and is expected to end in the next 6 to 12 months. The report suggests investors maintain neutral stock allocations in the short term, moderately underweight equities in the medium term, and closely monitor forward-looking indicators such as analyst forecast revisions, GPU leasing costs, and corporate free cash flow.
The report specifically notes that the current economic environment is adding further concerns: US job openings have dropped to a five-year low. If the AI boom fades and no new bubble cushions the impact, the potential economic recession ahead could be even more severe than the aftermath of the 2001 internet bust.
Lessons from History: The Trajectory of Four Capital Frenzies
BCA states that at its core, a capital expenditure boom is collective optimism about the commercialization prospects of new technology. But history repeatedly shows that such optimism is often disconnected from the realities of technological rollout, and ends in collapse amid supply-demand imbalances, accumulated debt, and inflated valuations.
The 19th-century British and American railway booms demonstrated the destructive power of overcapacity.
The report notes that the success of the Liverpool-Manchester railway in 1830 ignited an investment frenzy in Britain, with railway stock prices nearly doubling between 1843 and 1845.
By 1847, railway construction spending soared to a record 7% of the UK’s GDP. Tightening liquidity eventually triggered the October 1847 financial crisis, with the railway index plunging 65% from its peak.
The report says the American railway boom climaxed in the Panic of 1873, which forced the New York Stock Exchange to close for ten days, and between 1873 and 1875, corporate bond defaults reached 36% of face value.
After US railway mileage peaked at over 13,000 miles in 1887, overcapacity caused transport prices to crash, and by 1894 about 20% of American railway mileage was in bankruptcy administration.
The electrification boom of the 1920s exposed the fragility of a pyramid-like capital structure.
The report notes that household electrification rates rose from 8% in 1907 to 68% in 1930—but this process was concentrated in cities.
Wall Street was deeply involved, and utility company stocks and bonds were promoted as “safe assets for widows and orphans.” By 1929, holding companies controlled more than 80% of US electricity generation.
The report says that after the stock market crash in 1929, the largest utility group, Insull, went bankrupt in 1932, reportedly wiping out the life savings of 600,000 small investors. US electricity utility construction spending peaked at about $919 million in 1930, then plunged to $129 million by 1933.
The late-90s internet boom proved that innovation does not guarantee profits.
BCA notes that between 1995 and 2004, US non-farm business productivity grew at an annualized rate of 3.1%, far exceeding later periods.
But the proportion of tech-related capital expenditures in GDP soared from 2.9% in 1992 to 4.5% in 2000, and over-investment put enormous pressure on corporate balance sheets.
The report points out that free cash flow in the telecom sector peaked at end-1997 and then declined, with a sharp drop in 2000. After rising sixfold between 1995 and 2000, the Nasdaq Composite plunged 78% in the next two and a half years.
Multiple oil booms perfectly illustrate the recurring cycles of supply-demand imbalance.
BCA notes that after massive oil reserves were discovered in east Texas in 1930, daily output soared past 300,000 barrels in just 12 months, but the Great Depression sent oil prices crashing to 10 cents a barrel.
In 1985, Saudi Arabia abandoned output limits, sending oil prices to $10 a barrel.
Between 2008 and 2015, the US shale oil boom boosted crude output from 5 million to 9.4 million barrels per day, but in 2014 OPEC refused to cut output, causing oil prices to tumble from $115 per barrel in mid-year to $57 at year’s end.
Five Common Laws: The Inevitable Path from Boom to Bust
Reviewing the rise and fall of four typical booms, BCA Research summarizes five common laws that provide key benchmarks for judging the direction of the current AI frenzy. Specifically:
Law 1: Investors ignore the S-curve of technology adoption.
Technology adoption never advances linearly, but follows an S-shaped curve of “early adopters – mass adoption – laggards.” Stock prices usually rise early, peaking midway through mass adoption, when adoption rate growth turns negative.
Currently, the AI sector is showing these features: most companies claim to plan increased AI use, but actual adoption rates are stagnating, and some indicators have even slid in recent months. This divergence between “intent and action” is a typical signal of technology adoption entering the latter part of the second phase.
Law 2: Revenue forecasts underestimate price declines.
New technologies are initially scarce and command premium prices, but as they become widespread and competitive, prices inevitably plunge. Between 1998 and 2015, internet traffic grew at an annualized rate of 67%, but the price per unit of data transmission dropped sharply. Since solar panels debuted, their prices have continuously fallen—down 95% since 2007 alone.
The AI industry is repeating this pattern: since 2023, the launch of faster chips and better algorithms has driven token prices down more than 99%. Although new applications like video generation are emerging, user willingness to pay is still not clear.
Law 3: Debt becomes the core reliance for financing.
Early on, companies can usually fund capital expenditure with retained earnings, but as investment scales up, debt gradually becomes the main source of funding.
In October 2025, Meta announced a $27 billion data center financing deal through an off-balance-sheet SPV; Oracle, after securing $38 billion in loans, raised another $18 billion via bonds, bringing total debt close to $96 billion.
Even more concerning are “new cloud vendors” like CoreWeave; by October 2025, CoreWeave’s credit default swap rate had jumped from 359 basis points at the start of the month to 532 basis points.
Law 4: Asset prices peak ahead of investment downturn.
Historically, asset prices such as stocks peak before actual investment spending begins to decline. Even when investment comes off its highs, absolute levels can remain elevated, worsening overcapacity. This means that investors who wait for clear “investment downturn” signals often miss the best timing for action.
Law 5: Capex collapse and economic recession intensify each other.
Tech bubbles usually burst in two stages:
Stage one is the fade of speculative hype, and the emergence of overcapacity; stage two is the collapse of capital expenditure dragging down the overall economy, worsening corporate earnings and creating a vicious cycle.
The report points out that the 2001 US recession wasn’t triggered by deteriorating economic fundamentals, but by the collapse in capital expenditure after the internet bubble burst. The rise of the housing bubble in 2002 temporarily offset the impact, but it’s unclear whether a new bubble will soon replace the coming bust in AI.
Risk Signals in the AI Boom: The Turning Point in the Next 6-12 Months
Based on analysis of historical patterns, BCA Research believes the AI boom is following the bubble path and is likely to end within the next 6-12 months. This judgment is based on multiple risk signals already evident in AI.
From a technology adoption perspective, AI’s real-world rollout is already lagging behind capital’s fevered expectations; corporate adoption rates are stagnant, and consumer willingness to pay for AI applications is still unproven.
In terms of pricing, token prices have plummeted, showing deflationary pressure, while the commercial value of new apps like video generation remains in doubt.
From the perspective of debt risk, AI-related firms' financing structures are increasingly reliant on debt, and credit risks are beginning to emerge at some companies.
The report recommends closely tracking four forward-looking indicators:
First, analyst revisions to expectations for future capex—should rising forecasts flatten out, it might signal danger;
Second, GPU rental costs, which have begun to decline after May 2025;
Third, the free cash flow status of hyperscale companies, which, though still high, is starting to deteriorate;
Fourth, the “metaverse moment”—when an AI company announces a big project but shares fall instead, clearly marking a sentiment shift.
For investors, BCA Research advises taking a “moderately defensive” approach: Keep a neutral allocation to equities in the next three months, moderately underweight stocks over 12 months, and increase defensiveness in the months ahead.
Specifically, closely track the four indicators above, avoid waiting for clear investment downturns to adjust passively, and consider defensive sectors and high-quality bonds to hedge against large swings in AI-linked assets.
Risk Disclaimer and Exemption ClauseThe market carries risks, investment requires caution. This article does not constitute personal investment advice, nor has it considered individual users' specific investment objectives, financial situation, or needs. Users should consider whether any opinions, views or conclusions herein fit their own circumstances. Invest accordingly at your own risk.