Stock market bubbles are nothing new. From tulips and the new world, to the Wall Street Crash, to black Monday and to the Global Financial Crisis: markets have a habit of getting quite high and then falling quite fast. Stock market bubbles have many causes, but some have been brought about by the emergence of new technologies.
New technologies change the world, but what tends to happen is that their ability to change the world gets overhyped. Investors’ and businesses' expectations of their ability to make money become bigger than their actual ability to make money: when everyone realises this, the crash happens. The only major crash of this nature within living memory is of course the ‘dot-com’ bubble in the early 2000s: before this there were bubbles related to radio, electricity, the railways and canals.
Recently there have been a number of commentators warning that we are in a bubble induced by the new wonder technology of Artificial Intelligence. This includes the Bank of England, Head of the IMF Kristalina Georgieva and CEO of JP Morgan, Jamie Dimon. It is certainly true that the US stock market has been doing well of late, often somewhat inexplicably given concerns about Donald Trump’s tariffs. This is at least in part attributed to rising prices for AI-related stocks. Another possible sign of an AI bubble (other than how everyone is constantly talking about how world changing AI is), are that the income that AI firms are actually getting from selling their product is small compared to the huge capital investments being made in data centres, chips, electricity power grids and the technology itself. A further concern among market analysis is the emerging circular nature of some AI investments. For example, Nvidia has invested in several start-up companies which are in turn major purchasers of Nvidia’s own chips. These feedback loops can artificially inflate demand for a product.
AI related stock market events recently included when head of Oracle Larry Ellison became the richest man of the world (for a few hours) on 10th September 2025 after the company reported a surge in demand for its data centres from AI companies. More negatively however, US tech stocks took a tumble in January 2025, when news of the success of far more cheaply developed AI from a Chinese company DeepSeek made markets question US dominance of the sector.
Bubbles are associated both with rapidly rising share prices but also rising ‘price to earnings ratios’: how many times the company’s annual income you are paying for the share. A share price that is rising because actual income is rising is less of a concern than when share prices are rising far ahead of income. Under this measure the possible AI bubble can quite easily be described as ‘not that bad’, at least when compared to the dot-com crash. The current average price to earnings ratio of the ‘magnificent 7’ companies (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla) is 55, or 38 if you exclude Tesla which is somewhat of an outlier1. Whilst paying 38x what something is earning may still sound quite high, it is relatively small compared to the dot-com boom which saw price to earnings ratios of the companies involved reach around 200x.2
Due to the passage of time, price to earnings ratios comparisons are not realistic for the railway or other previous bubbles. However on some available measures the current situation seems, again, not that bad. 3%-4% of current US GDP has gone into AI investment over the last four years: comparable British GDP figures from the railway boom in the 1840s were 15% - 20%3 . Other differences are in the investments and investors. The magnificent 7 and other major players are predominantly large and well established companies, with strong balance sheets and deep pockets. This differs from the dot-com crash which was characterised by lots of new, smaller, upstart companies that didn’t have much resilience when things went badly. The current investment is largely equity rather than debt financed: equity holders are more concentrated amongst the wealthy and are likely to be more able to bear losses. Previous railway booms were, by contrast, predominantly financed by debt from banks3. This concentration in a few key market players had predictably negative effects on the wider economy when it all went pop. AI also depreciates much faster than railways do, meaning costs are in generally being recognised in a more up front way.
So I would personally conclude that, whilst some sort of US market correction feels inevitable, we are more in a stock market ripple than a bubble. Perhaps only time will fully tell whether AI changes our lives as much as the internet, radio or railways did; or whether it will just continue to be a very energy intensive way of getting some cute pictures of kittens. I think however it is key to remain vigilant as we typically do not learn from history and are doomed to repeat it. If price to earnings ratios, share prices and other measures typically associated with bubbles show signs of rising to previous, ultimately unsustainable, highs then buyers should beware.
1 Bloomberg
2 Teeter, Preston; Sandberg, Jorgen (2017) ‘Cracking the enigma of asset bubbles with narratives’. Also Marlborough Chart of the Week: Don’t Believe the Hype – AI: bubble or only at ‘base camp’? 13th October 2025.
3 The Economist, 13th September 2025, ‘Glitch in the matrix’
Related Insights
What Does the US Government Shutdown Mean for the Economy?