AIに歴史からの警告！膨張した過去の歴史とは？

AI Boom Draws Historical Warning — BIS and Prominent Investors Sound the Alarm

The Nikkei Shimbun ran a report headlined "A Historical Warning for AI: The Ballooning Investment and the Risk of Reversal." The AI investment wave has been lifting the global economy — semiconductor demand tied to data-center construction has buoyed Taiwan and South Korea, and power-plant additions are accelerating. At the same time, major technology companies are increasingly tapping the corporate bond market to plug funding gaps, and skepticism about the sustainability of the boom is growing.

In the United States, books on financial panics are selling briskly. Titles chosen by former President Obama, including *1929: Inside the Crash*, and a book on the global financial crisis of 1873 — tracing the contraction that followed the flood of capital into U.S. railroad construction — have drawn particular attention. Prominent investor Ray Dalio has warned that a debt crisis will occur within two years. The BIS devoted its 2026 Annual Report to comparing today's AI boom with four past investment surges, concluding that the current pace of investment expansion is "abnormally fast." Goldman Sachs projects that AI-related capital expenditure by Microsoft, Alphabet, Amazon, Meta, and Oracle will reach ¥189 trillion in 2027, up 50% from this year.

Lessons from Four Historical Investment Bubbles

**Example 1: The U.S. Canal Boom (1830s)** The success of the Erie Canal in the 1820s — linking the East Coast to the Great Lakes and generating toll revenue from moving inland crops cheaply to New York — prompted other states to launch their own canal projects in rapid succession. Most proved unprofitable, and many states were unable to repay funds drawn from public finances. Capital on a scale comparable to the AI boom was deployed in a short window, and the whole episode unraveled within seven or eight years.

**Example 2: Britain's Railway Mania (1840s)** After the Liverpool–Manchester Railway succeeded in the 1830s, shareholders received annual dividends of around 10% — highly attractive against a UK gilt yield of just 3%. Hundreds of new railway schemes obtained parliamentary approval. Shares were sold on an instalment basis, however, requiring buyers to make ongoing payments after purchase. As obligations mounted, investors sold stakes in other railways to fund the calls; selling begat more selling, and the bubble collapsed. One lasting legacy of the era: rail gauges across the country were standardized at roughly 1.4 meters, a standard carried forward to modern track.

Across the Atlantic, the transcontinental railroad (Central Pacific, East-West link completed 1869) was being built at the same time. To raise capital, bankers sold bonds paying 7.3% interest through newspaper advertisements and sales brokers, pitching them to German immigrants and European investors — going so far as to name a city in North Dakota "Bismarck" to aid European marketing. When Austrian stock prices plunged in 1873 the contagion spread, U.S. bonds were dumped, and the railroad company went bankrupt. The names of railroad magnates such as Leland Stanford lived on through Stanford University, and the predecessor of the Dow Jones — a transportation index of eleven stocks, nine of them railroads — was born in this era.

**Example 3: The U.S. 1920s Bubble and the Great Depression** Concurrent breakthroughs in electricity, the automobile, and aviation — from the Wright Brothers' first flight in 1903 to Lindbergh's transatlantic crossing in 1927 — pushed stocks dramatically higher. Margin trading, which allowed investors to control ten times their capital with a 10% deposit, supercharged the ascent. On Black Thursday, October 24, 1929, a surge of sell orders on automobile stocks overwhelmed the market. Prices temporarily recovered, but the passage of tariff legislation in June 1930 triggered a second collapse, ultimately driving the market roughly 90% below its peak.

**Example 4: The Dot-Com Bubble (around 2000)** A flood of investment into internet-related companies ended in a crash, yet the fiber-optic networks laid during the boom lived on as foundational infrastructure for the internet economy that followed.

The Key Risk That Makes This AI Boom Different

All four past bubbles shared one feature: the physical assets survived. Canals, railways, and fiber-optic cables continued to serve economies for decades after the financial wreckage was cleared. The AI boom's core assets — GPUs and data-center equipment — are different. While the shells and cooling systems of data centers will remain, the GPUs themselves can become obsolete in as little as a few years. That is the structural risk analysts say has no precedent in history.

There is also the question of how long it takes AI to translate into corporate profits. The historical analogy given is factory electrification: when electric motors first arrived, operators simply swapped large steam engines for large electric motors and missed the point; it took years before industry figured out that placing small motors on individual machines was where the real efficiency gains lay. AI may follow the same curve. In the meantime, the question is whether companies that have committed ¥189 trillion in capital expenditure can generate returns before the reckoning arrives. The risk that bonds issued by emerging cloud companies trace the same arc as debt in past financial collapses is being watched closely.

Markets on the Day: Nikkei Plunges More Than 1,400 Points Before Recovering

On Monday, October 19, 2026, the Nikkei Average opened down 504 points at 64,819, rattled by renewed turbulence at AI cloud provider CoreWeave and the resulting sell-off in U.S. equities on fears over AI-related debt. Rising oil prices added to the negative sentiment, and losses widened to more than 1,400 points at the session's worst, briefly breaking below 65,000. The market found a floor when semiconductor testing equipment leader Advantest rebounded more than ¥2,000 from its intraday low of ¥32,420 to close at ¥34,450, pulling the broader index back to near 65,000. Cybersecurity stocks were also active, with trading volume in some names surging from around 100,000 shares to more than 40 million shares over just a few days amid a run of news reports on customer data breaches — though investors were cautioned that small-cap growth stocks in the segment carry the risk of hitting daily loss limits.