Is artificial intelligence a bubble? For Americans, thanks to ever-higher valuations on top AI firms it’s been another week of euphoria.
The S&P 500 index hit a new high on Thursday with the world’s most valuable firm, Nvidia, marching towards a $6 trillion (£4.5 trillion) valuation.
Nice for its founder and boss Jensen Huang, one of the business leaders accompanying Donald Trump in China last week, and a big contrast to the mounting gloom on the markets in the UK.
US markets did fall back a bit on Friday, but it was nothing like the wallop that hit London. Sterling, gilts and equities all plunged on political uncertainty, in particular the prospect of Andy Burnham becoming Prime Minister.
But share prices do not go up forever, so you have to ask how long the AI boom can continue, and – if it is indeed a bubble – what happens when it pops?
To be clear, there’s an important distinction between how AI will transform the economy and what it means for financial markets.
It’s perfectly possible for a new technology to have a deep and lasting effect on the way we work and indeed live, but for investors to lose money on it if they buy at the wrong time. As a template, take the dot-com boom of the late 1990s and the crash that followed.
Prediction: Claude says there will be a nasty break in high-tech share prices in the next couple of years, but this is not around the corner
So the issue here is not the AI revolution itself. It’s the price you put on the firms at the heart of it. Can you justify those levels of investment and their valuations?
My worry is that everything has to go right. Nvidia at $5.5 trillion is on a chunky price/earnings (p/e) ratio of 46, meaning it is valued very highly by investors.
Alphabet, the holding company for Google, is worth $4.5 trillion and on a p/e of 30; Apple at $4.4 trillion is on a p/e of 36. By contrast US equities as a whole, taking the S&P 500 as the benchmark, are just over 30, though if you look at the forward p/e – that is based on profit forecasts for the next year – it is down to 22.
The historical average for the S&P 500 is 16 to 18, so a significant squeeze on company earnings would be bad news indeed. If Nvidia were to miss a profit forecast at its next set of results on Wednesday that might well be enough to trigger a wider sell-off.
Hamish McRae: The correction could be 40%
So too might a rise in interest rates. The mood in the US has shifted in the past few days from expecting no change in rates from the Federal Reserve this year to at least one upward move.
That’s a reaction to disappointing inflation figures, and the prospect of the Strait of Hormuz remaining closed for several more weeks at best – which would push up inflation through the autumn.
Longer-term US yields also climbed. If the costs of funding the huge investments needed to keep the AI enterprise moving forward increase substantially that cuts into margins.
If credit becomes not only more expensive but harder to obtain, that really could trigger the sell-off.
As for the scale of the correction, well, typically it could be an average decline of 40 per cent for the high-tech sector. After the dot-com bubble burst in 2000, technology shares, as represented by the Nasdaq Composite index, fell by 80 per cent.
But the valuations at the peak were more extreme, and the index had gone up by 600 per cent in the previous five years. So it is unlikely we are heading into that territory. However, the accepted definition of a bear market is a fall of 20 per cent and that is looking increasingly likely. But when?
No one can time downturns reliably, but my instinct is that it’s still more likely to come next year, or later, than this year.
Sponsored
Scottish Mortgage’s Tom Slater: What investing in Anthropic has revealed about AI’s future
What Claude had to say on the AI bubble
And since we’re talking about artificial intelligence, I thought I would ask Claude, Anthropic’s free-to-use AI assistant, for its assessment.
It looks at the arguments that the downturn might be in the next 12 to 24 months, or if things might run on for longer – two to four years.
Its judgment is that the ‘highest probability window for a meaningful correction based on historical tech cycle lengths’ is 2027 to 2028.
There you have it. There will be a nasty break in high-tech share prices in the next couple of years, but this is not around the corner.
That may be wrong but at least it’s sensible – and, let’s face it, a blessed relief from fussing about the fate of Sir Keir Starmer and his Chancellor Rachel Reeves.
