AI-Related Stock Concentration Reaches Second Highest in History, Only Behind Railroad Boom
Barchart cites data from Bank of America and Bianco Research indicating that the concentration of AI-related stocks has risen to the second highest level in history, only behind the railroad boom of the 1880s in the United States.
Bank of America estimates that the "AI Big 10" (including the Magnificent 7, Broadcom, AMD, Micron) accounts for nearly 40% of the S&P 500 market capitalization; a broader group of 41 AI-related companies (including direct AI, utilities, and capital equipment firms) makes up about 49%.
During the peak of the railroad stocks in the 1880s, they accounted for about 63% of total market capitalization, the Nifty Fifty around 36%-40%, the Japanese stock market in the late 1980s about 44%, and the TMT bubble in 2000 around 40%-44%. The current AI concentration has surpassed most historical peaks.
The expected capital expenditure of major cloud providers is projected to rise from about $330 billion at the beginning of 2025 to $1 trillion over the next 12 months, accounting for approximately 3% of U.S. GDP, higher than the 2.2% peak during the railroad construction in the 1870s and 1.1% during the tech bubble of the 1990s.
Since summer, the average price of AI tokens has dropped by about half; a Bank of America fund manager survey shows that about 45% of respondents expect an AI bubble by 2027, with 37% net overweighting tech stocks, the highest since 2020.
Most investors lost money during the historical railroad boom, while a few survivors became long-term core assets; the current rise in concentration is accompanied by rapid capital expenditure expansion and price pressures.
Market Mechanism: Institutional and retail funds continue to flow into a few leading AI companies and their supply chains, with event-driven further concentration benefiting computing power and infrastructure companies, while facing pressure from the relative performance of non-AI sectors and risks of declining future capital returns.
Source: Public Information
ABAB AI Insight
During the peak of U.S. railroad investments in the 1880s, railroad stocks once accounted for about 63% of total market capitalization, with capital flooding into rail construction through bonds and stocks, ultimately triggering the Panic of 1873 and long-term defaults; Bank of America's strategy team (including Michael Hartnett) has repeatedly compared the current AI concentration to that period as an extreme historical case and tracks the weight changes of the "AI Big 10."
In terms of capital pathways, funds are primarily concentrating upstream through the capital expenditures of major cloud providers (expected to reach around $1 trillion) and the semiconductor supply chain, while institutional net overweighting of tech stocks has risen to a six-year high; the motivation is to secure computing power and model leadership, but this has been accompanied by a decline in token prices and compressed return expectations.
Drawing parallels with the British railroad boom of the 1840s (where figures like George Hudson maintained high valuations through capital payments) and the U.S. railroad bubble of the 1870s, technology was implemented but most investors lost money; currently, AI is in a phase of capital-intensive expansion rather than a widespread profit realization period, with the industry's position similar to the early stages of excessive infrastructure construction.
Structurally, this represents capital concentration: market pricing power and fund flows are dominated by a few AI-related companies, with the mechanism being that high expected capital returns attract more leverage and allocation, forming a self-reinforcing concentration cycle until overcapacity or slowing demand growth breaks this cycle, similar to many historically technology-driven infrastructure cycles.
ABAB News · Law of Cognition
- The higher the concentration, the faster the transmission of volatility.
- Technological success does not equate to investor success.
- Prices peak before capital expenditures do.