What Will Make the Market Crash?

The largest stock market crashes have occurred in September or October. The Panic of 1907 began in mid-October, the Wall Street crash of 1929 saw its biggest falls on 24 and 29 October, the ‘Black Monday’ crash of 1987 was on 19 October, while the collapse of Lehman Brothers that triggered the banking crisis occurred in September 2008.

Are we heading for a similar crash? By any conventional indicators, stock market valuations are overheated. The problem is that this has been the case for some months, and those ignoring the warning signs have profited.

The investment boom in AI is believed by many to herald such a huge boost to business productivity that traditional indicators of company valuation are no longer valid – but of course the phrase ‘This time is different’ is itself a warning sign. It is the title of a book on investment bubbles and crashes, written by economists Carmen M Reinhart and Kenneth S Rogoff and published in 2009, shortly after the start of the financial crisis.

There are two strong indicators that this time is really no different to earlier bubbles. The first is that at least some of the investment in AI may be misplaced.

 There is little doubt that the rapid development of highly powerful AI tools holds the potential for significant productivity gains.

 It is less certain, however, that the big bet on massively increasing the capacity of data centres in the quest for an ultra-high level of synthetic intelligence is going to pay off in a direct way.

A study by the Massachusetts Institute of Technology in August reported that 95% of AI pilot schemes did not result in better business performance. In mid-September JP Morgan Asset Management warned that valuations in AI were ‘stretched’, such that only a small disappointment in earnings could prompt a sell-off. 

The scale of investment in data centres to power AI is in the order of $3tn, around half of which comes from the capital of big tech companies, but a high proportion is funded by private credit, which is comparatively opaque, and is linked to the banking system.

It is a near-certainty that many venture capital-backed AI start-ups will fail, but the extent of this and the impact on the wider economy is difficult to gauge.

Evidence is emerging of productivity gains from AI – but these emerge from smarter and more focused use of bespoke AI tools, allied to the most intelligent human direction. Small language models (SLMs) may be more effective than LLMs in many business applications, which is great news for those firms that get it right – but less so for the investors who have bet big on scaling up.

It is all but inevitable that a major new technology will feature a bubble. It has occurred with railways in the 19th century, and dotcom firms and supportive infrastructure in the late 1990s. Even when the bubble bursts, it is only a serious problem for the wider economy if it affects the banking industry such that loans dry up for other parts of the economy, heralding a recession.

The second major risk factor is the pro-cyclical behaviour of the President of the US in slashing interest rates and encouraging speculation. Just before the financial crash of 2007-08, Chuck Prince, then the CEO of Citibank, famously said that as long as the music is playing, you need to dance. President Donald Trump is now the one trying to keep the music playing.

President Trump encourages stock market investment, and authorised the US government to purchase a 10% stake in the chip manufacturer Intel – an extraordinary decision.

He has also pressured the Federal Reserve to lower interest rates, expressing a desire both for ultra-low rates and a weaker dollar. The official US interest rate was duly cut in mid-September, by 25 basis points, to 4-4.25%.

It is unusual for interest rates to be reduced to very low levels in non-recessionary conditions. Inflation is not very high, but it is above the nominal target of 2% and in August it edged upwards to 2.9% from 2.7%. There are, however, indicators of credit delinquency and other signs of financial stress among some consumers.

There is likely to be at least one more cut of the same amount before the end of 2025, and probably two. Nominally, the Federal Reserve is independent of the White House, but President Trump has made his desire for lower rates very public. 

The courts have so far paused his efforts to remove Lisa Cook from the Federal Board. His own nominee for the board, Stephen Miran, has been approved.

There is another risk factor: Less reliable economic statistics. In early August, President Trump fired Erika McEntarfer, Commissioner of the Bureau of Labor Statistics, complaining that the employment data, weaker than expected, were incorrect. In addition, budget cuts at the Bureau have meant a reduction in the data points that feed into the official statistics.

The US is exhibiting some of the features more normally associated with emerging economies: A President over-reaching his authority, compromised independence of key institutions, concern over the accuracy of economic data. 

This does not mean we are about to witness economic meltdown and hyper-inflation in the US, given the depth and strength of its internal economy, but there are signs of weaker long-term stability.

A near-certainty is the continued increase in US public sector debt, and erosion of the value of paper money. The gold price has risen from $2,600 per ounce less than a year ago to around $3,800 per ounce by late September.

 Gold now forms a greater proportion of central bank reserves than US Treasuries for the first time since 1996.

As regards the investment bubble, is this time different? In many respects, no. Will there be a market crash in October? It is never possible to be certain, but there are many red warning signs.