12th October 2026
History is full of investors who convinced themselves that prices could only go up. Artificial intelligence may genuinely transform the world, but that does not mean today's extraordinary investment boom will end happily for everyone.
In 1637, Dutch tulip prices collapsed after a period of extraordinary speculation. In 1720, Britain's South Sea Bubble burst, leaving investors nursing heavy losses. At the turn of the millennium, the dot-com boom ended in a spectacular technology stock-market crash.
Today, another investment frenzy is under way. This time, the object of excitement is artificial intelligence.
Billions of pounds and dollars are pouring into AI companies, computer chips and data centres. Technology giants are committing enormous sums to infrastructure, while investors are betting that artificial intelligence will transform everything from office work and education to medicine, manufacturing and government.
There are good reasons to believe that AI will change the world. But there is an important question that investors should not ignore.
Could artificial intelligence prove to be a revolutionary technology while the financial boom surrounding it turns out to be a bubble?
History suggests that the two outcomes are perfectly compatible.
When enthusiasm outruns reality
Financial bubbles rarely begin with an obviously ridiculous idea. More often, they start with a genuine opportunity.
Investors identify something new that could transform the economy. Early successes attract attention, rising share prices attract more buyers, and the prospect of enormous profits becomes increasingly difficult to resist.
Eventually, the story can become more important than the numbers.
Investors stop asking how much profit a business is likely to make and start concentrating on how valuable it might become. Companies are judged on the size of their ambitions rather than their ability to turn those ambitions into cash.
Rising prices then appear to confirm the original excitement. More investors join in, sometimes borrowing money to avoid missing out.
The danger is that expectations become so high that even a successful business can disappoint.
When confidence finally weakens, investors discover that an asset bought at an inflated price does not become a bargain simply because its underlying technology is impressive.
That is the lesson repeated throughout financial history.
Tulip mania: the power of a story
The Dutch tulip mania of the 1630s has become the classic illustration of speculative excess.
Rare tulip varieties attracted intense interest, and prices for some bulbs rose to extraordinary levels. Buyers became willing to pay sums that were difficult to justify by the practical or commercial value of the flowers.
When confidence broke in 1637, prices collapsed.
The popular version of the story sometimes exaggerates how widespread the mania was, but its underlying lesson remains useful. When people believe that an asset will become more valuable simply because other people are prepared to pay more for it, speculation can become self-reinforcing.
AI is obviously very different from a tulip bulb. It has practical applications, paying customers and the potential to increase productivity.
But the psychology can be similar. The fact that a technology is exciting does not establish what a company developing it is worth.
The South Sea Bubble: promises of extraordinary wealth
Britain's South Sea Bubble of 1720 provides another warning.
The South Sea Company became the focus of enormous investor enthusiasm amid expectations about its commercial prospects and its role in managing government debt. Its shares soared before collapsing, causing substantial losses.
The episode involved political influence, financial arrangements and expectations that proved far more optimistic than reality justified.
The comparison with AI is not exact. Today's technology companies operate in a very different economic and regulatory environment.
Nevertheless, the South Sea story illustrates how enthusiasm, powerful backers and promises of future wealth can persuade investors to overlook difficult questions about how the expected returns will actually be delivered.
AI investment has its own version of that question: who will ultimately pay for the extraordinary infrastructure being built, and will those payments produce enough profit to justify the cost?
The dot-com crash: the closest comparison
The technology boom of the late 1990s is arguably the most relevant historical parallel.
The internet was changing the way people communicated, shopped and did business. Investors correctly recognised its enormous potential, but many became convinced that almost any company associated with the internet was destined for success.
Businesses attracted enormous valuations despite having limited revenues, no sustainable profits or uncertain commercial prospects.
When confidence broke in 2000, technology shares plunged. The Nasdaq Composite eventually fell by roughly 78 per cent from its March 2000 peak to its October 2002 low.
Many internet businesses disappeared. Investors lost fortunes.
Yet the internet itself continued to develop and eventually transformed the global economy.
Companies such as Amazon survived and became major businesses, although investors who bought near the peak had to endure a devastating fall before any eventual recovery.
The lesson is particularly relevant today. AI could become as important as the internet without every company investing in it becoming a profitable investment.
The extraordinary cost of the AI race
The sums now being committed to AI are staggering.
OpenAI, the company behind ChatGPT, is expanding rapidly but faces enormous computing costs. Recent reports have highlighted differences between previously anticipated revenue figures and newer estimates, alongside projections of very large cumulative cash outflows over the coming years.
Those figures need to be interpreted carefully. Differences in accounting and the treatment of revenue earned through cloud partners can complicate comparisons, while projected spending is not the same as money already lost.
But the broader question remains.
AI developers need powerful chips, data centres, electricity and vast computing capacity. The infrastructure must be paid for before the eventual financial returns are fully known.
Microsoft, Google, Amazon and Meta are among the companies investing heavily in the technology. Nvidia has become a central supplier of the advanced chips needed to run many AI systems. Other companies, including Anthropic and China's DeepSeek and Alibaba, are competing to develop powerful models and services.
If demand continues to expand and customers are willing to pay enough, the investment could prove highly profitable.
If customers adopt AI more slowly than expected, or if competition forces prices down, some of that infrastructure may generate disappointing returns.
The risk is not that nobody wants AI. It is that the industry may spend too much, too quickly, in anticipation of profits that arrive later or prove smaller than expected.
A bubble does not require every company to fail
This is where the debate becomes more complicated.
Some of the largest AI investors have established businesses that generate substantial revenues and profits. Microsoft, Google and Amazon are not dependent entirely on the success of a single AI product.
Nvidia is selling equipment that customers genuinely need today. AI services are already being used by businesses and individuals, and the technology is improving.
These facts make the current situation different from the most speculative corners of the dot-com era.
But they do not rule out a bubble.
A company can be excellent and still have an overvalued share price. A technology can be transformative and still fail to generate the profits investors have already assumed.
If markets have priced in years of exceptional growth, a relatively modest disappointment can trigger a large fall in share prices.
The eventual winners may survive and prosper, while investors who paid too much for their shares suffer substantial losses.
Could the consequences reach Britain?
A major fall in US technology shares would not remain an American problem.
British pension funds, workplace pensions, private pensions and ISAs can all have exposure to US shares and global investment funds containing major technology companies.
If those shares fall sharply, investment values in Britain can fall too.
There could also be wider consequences if the correction undermines business confidence, reduces investment or exposes financial institutions to losses on loans made to companies building AI infrastructure.
A stock-market correction would not automatically cause a recession or a banking crisis. Much would depend on the scale of the losses, the amount of borrowing involved and the condition of the wider economy.
But if an AI downturn coincided with high energy prices, expensive government borrowing and weakening consumer spending, the combination could be more damaging than any one problem on its own.
For ordinary savers, the sensible response is not to assume that a crash is imminent. It is to understand how much their investments depend on a relatively small group of large technology companies and whether their portfolios are diversified enough to withstand a substantial fall.
What should investors watch?
There are several warning signs worth following.
Are AI companies meeting their revenue targets? Are customers paying enough for the services they use? Are major technology firms continuing to increase investment because demand justifies it, or because they fear falling behind competitors?
Are chip orders and cloud revenues supporting the forecasts being used to justify current share prices? Are data-centre projects generating adequate returns? And are lenders becoming more concerned about the debts and commitments associated with AI infrastructure?
Perhaps the most revealing question is whether spending on AI is producing measurable improvements in business productivity and profits.
If it does, today's investment may prove justified. If the industry struggles to turn its technological achievements into sufficient commercial returns, the market could become much less forgiving.
The technology may be real. The bubble may be real too.
History does not tell us when an AI bubble might burst, or even establish that the entire market is currently in one. Markets can remain expensive for years, and today's investment could ultimately support a much larger and more productive economy.
But history does offer a clear warning.
Tulip buyers became caught up in the belief that prices would keep rising. South Sea investors were captivated by extraordinary commercial promises. Dot-com investors correctly recognised the importance of the internet but often badly misjudged the value of the businesses they backed.
Today's AI investors face a similar challenge: separating the potential of the technology from the financial assumptions built around it.
The real danger is not that artificial intelligence will fail to change the world. It is that investors may already have paid too much for the future they expect it to deliver.
If that happens, the technology could continue advancing even as share prices collapse.
And for people whose pensions and savings are tied to the stock market, that distinction could make all the difference.