The AI Investment Boom and Its Collapse

Technological breakthroughs have repeatedly triggered enormous investment booms, followed by excessive optimism, overinvestment and sudden collapses. The pattern appeared in American canal construction in the 1830s, Britain’s railway frenzy in the 1840s, the United States’ 1920s enthusiasm for electricity and automobiles, and the dot-com boom of the late 1990s.

The current AI boom is expanding at an even faster pace. Global investment in AI infrastructure could reach $700 billion in 2026 and nearly $1 trillion in 2027. The Bank for International Settlements says investment has already grown fourfold in the three years since the boom began in 2023, compared with the period immediately before it.

A small number of companies are expected to survive the winner-takes-all competition. Chipmakers such as Nvidia and AMD, hyperscalers including Google, Amazon, Meta, Microsoft and Oracle, AI developers such as OpenAI and Anthropic, and newer cloud providers such as CoreWeave and Nebius are all continuing to invest aggressively.

In 2025, 1,246 AI companies operated across 32 countries. The United States accounted for about 700 and China for about 250, while South Korea had 15. Yet AI companies represented 40% of U.S. stock-market capitalization and 39% of South Korea’s, the highest shares among the countries cited. AI capital spending also accounted for 26% of total corporate investment in South Korea in 2024, compared with 23% in the United States.

The article argues that the boom carries the seeds of its own weakness. As investment capacity expands, companies must meet increasingly high productivity and profitability thresholds to sustain the boom. Overinvestment by one company can reduce competitors’ profitability, creating an economic externality. The boom also combines real-economy and financial overheating through circular finance, including cross-investments, loans and guarantees. Because much of the debt-funded infrastructure consists of specialized chips, data centers and power facilities, a simultaneous contraction in demand could force companies to sell those assets cheaply. The competitive pressure driving the boom may therefore intensify, rather than spread, its risks.