The AI Bubble: When the Machines Meet Economic Reality

When even The Economist begins warning readers to expect an artificial intelligence crash, it is worth paying attention. In its leader of 12 November 2026 (US time), entitled "Why you should expect an AI crash," the magazine argues that the extraordinary enthusiasm surrounding AI may be approaching a painful encounter with economic reality. The technology may be revolutionary, but revolutionary technologies have a long history of producing financial manias in which investors confuse genuine technical progress with the certainty of extraordinary profits.

The Economist's argument is not that artificial intelligence is a fraud, nor that its collapse as a financial investment would necessarily produce another global financial crisis. Rather, it distinguishes between a potentially severe stock-market correction and a systemic banking catastrophe. A dotcom-style crash is entirely plausible, it suggests, while a repetition of the 2008 financial disaster is considerably less likely. That distinction is important, but it should not obscure the serious economic consequences that could follow if the present investment boom loses momentum.

The historical parallels are obvious. Railways, electricity, radio, automobiles and the internet all inspired enormous enthusiasm among investors. Many of these technologies ultimately transformed civilisation, but the financial fortunes of their early promoters were frequently disastrous. The railway mania of the nineteenth century produced vast overinvestment, while the dotcom bubble at the turn of the millennium destroyed trillions of dollars in market valuations. The internet survived and flourished, but that was little consolation to investors who had purchased shares in companies whose business models could not justify their prices.

Artificial intelligence may be following a similar trajectory. The extraordinary capabilities of large language models have encouraged expectations that almost every sector of the economy will be transformed. Corporations are investing enormous sums in specialised processors, data centres, electricity supplies, networking equipment and supporting infrastructure. Investors have rewarded companies associated with this expansion with valuations that assume substantial future earnings. The difficulty is that expectations of future profits are not the same thing as profits already earned.

There is a fundamental economic question that technological enthusiasts sometimes neglect. Who will ultimately pay for all this computing capacity, and how much will they be willing to pay? AI systems can already perform valuable tasks in programming, document preparation, research, administration and other fields. Yet the existence of useful applications does not automatically establish that customers will generate sufficient revenue to justify the entire infrastructure being constructed around them.

Competition may intensify the problem. As AI systems become more capable, their services may also become cheaper. Open-source models, smaller specialised systems and efficiency improvements could place downward pressure on prices. What is excellent news for consumers may prove considerably less attractive to investors who have financed expensive computing infrastructure on the assumption of sustained high margins.

There is also the problem of technological obsolescence. A conventional factory may remain productive for decades, but advanced AI computing equipment can lose its competitive advantage remarkably quickly. A company spending billions on today's processors must consider the possibility that tomorrow's systems will deliver comparable performance at a fraction of the cost. The financial risk is not merely that AI will fail to develop. It is that AI will develop so rapidly that expensive investments become outdated before they have earned an adequate return.

The Economist nevertheless sees reasons to doubt that an AI crash would necessarily become another 2008-style catastrophe. Much of the boom has been financed through equity rather than the enormous structures of highly leveraged debt that helped produce the global financial crisis. If share prices collapse, investors suffer losses, but those losses do not automatically trigger the same chain of bank failures, forced liquidations and frozen credit markets associated with a systemic banking panic.

This is a reasonable distinction, although it should not be converted into a guarantee of safety. Equity losses can have substantial economic consequences even without a banking collapse. Falling share prices reduce household wealth, undermine business confidence and discourage new investment. Pension funds and retirement portfolios may suffer. Companies that expanded employment and construction programmes in anticipation of continuing AI investment could suddenly reverse course.

The wider economic consequences might be especially serious because the AI boom has become so important to American technological investment and stock-market expectations. A sharp reduction in data-centre construction, processor orders and associated expenditure could affect businesses far removed from the companies developing AI models. Electricity infrastructure, construction, commercial property and specialised manufacturing are among the sectors exposed to changes in investment demand.

Nor should one assume that equity financing eliminates all financial danger. Some infrastructure projects rely upon borrowing, and the relationships between technology companies, investors, lenders and specialised financing vehicles can be complex. A major collapse in expected revenues could reveal financial weaknesses that were less apparent during the period of rapid expansion. The absence of an obvious 2008-style mechanism is reassuring, but it does not mean that every channel of financial contagion has been eliminated.

The more interesting question concerns the relationship between technological progress and economic value. Artificial intelligence could become one of the most important inventions in history while simultaneously disappointing the shareholders of companies currently associated with it. Indeed, widespread adoption might eventually depend upon AI becoming so inexpensive that many present business models cease to be attractive.

There is a further paradox. If artificial intelligence achieves the productivity revolution promised by its advocates, it may undermine the employment and income structures upon which consumer demand depends. Businesses may reduce labour costs through automation, but the broader economy still requires people with sufficient purchasing power to buy goods and services. A technology that transforms production without an accompanying adjustment in the distribution of purchasing power could generate economic and political tensions far beyond the fortunes of technology stocks.

This issue deserves particular attention from critics of the existing financial order. Modern economies are already characterised by substantial concentrations of corporate ownership and technological power. If AI greatly increases productive capacity while the financial rewards flow disproportionately to a small number of corporations and asset owners, the resulting problem will not simply be one of efficiency. It will concern who controls the new productive machinery and who benefits from its output.

The social-credit concern with the relationship between production and purchasing power becomes relevant here, as advanced by C. H. Douglas: what happens when society acquires the capacity to produce more with less human labour, while access to the resulting abundance remains dependent upon established mechanisms of income distribution? Artificial intelligence could make this question considerably more urgent.

The political response to an AI crash would also be revealing. During the expansion, technological corporations celebrate market innovation and private enterprise. If enormous investments subsequently fail, pressure may emerge for governments to protect strategically important infrastructure, subsidise national AI champions or otherwise cushion the consequences. The public should be entitled to ask why private investors ought to retain the gains of speculative enthusiasm while taxpayers are expected to absorb its losses.

None of this establishes that an AI crash is imminent. Financial bubbles can persist longer than sceptics expect, and genuine productivity gains could eventually justify investments that presently appear extravagant. The timing, severity and economic consequences of any correction remain uncertain.

Nevertheless, The Economist has identified a central weakness in the prevailing enthusiasm. The belief that a technology will transform civilisation does not prove that every company investing in it will become profitable, or that current market valuations are rational. History repeatedly demonstrates that the most important inventions can coexist with some of the most spectacular financial mistakes.

The artificial intelligence revolution may continue long after the speculative excitement surrounding it has disappeared. The machines could become more powerful, more useful and more deeply integrated into everyday life, even as investors discover that they paid far too much for the privilege of financing their development.

That would not be the end of artificial intelligence. It would simply be another demonstration that technological brilliance does not abolish economic reality.