The phantom of the 1999 dot-com crash is haunting the Square Mile. This week, as the FTSE 100 clawed to modest gains, the tech-heavy Nasdaq Composite endured what analysts are calling a ‘reality correction’ on an almost biblical scale. The cause? A sudden, collective panic that the artificial intelligence boom may have been priced for a utopia that does not yet exist. City of London regulators have been forced to issue an unusual statement urging ‘measured confidence’ as bellwether stocks in AI infrastructure and cloud computing haemorrhaged value.
For the uninitiated, the AI rally of 2023-24 felt like a second gold rush, except the gold was made of ones and zeros. Nvidia, the chipmaker that became a household name through its GPUs powering the world’s neural networks, saw its market capitalisation soar past the GDPs of most G7 nations. Software companies, cloud providers, and even firms merely whispering ‘machine learning’ saw their valuations double, triple, sometimes quadruple. It was a narrative so compelling that even cautious pension funds began to treat it as a sure bet.
But last Tuesday, something cracked. A series of earnings reports from mid-tier AI firms revealed revenue that, while growing, was wildly outpaced by expenditure on training data, server farms, and the ever-more-elusive talent of PhDs in deep reinforcement learning. The market blinked. Then it ran. In a single trading session, the Solactive AI Index shed 7.2 per cent. This was not a dip. This was a rear-guard action.
From my vantage point, having watched three previous tech cycles from the San Francisco trenches, this feels hauntingly familiar. In 1999, the promise was a ‘new economy’ where dot-coms would replace brick-and-mortar. In 2024, the promise is a ‘cognitive revolution’ where AI systems will unlock trillions in productivity. The common denominator is gravity. Valuations detached from fundamentals, a phenomenon I have termed ‘algorithmic irrationality’, eventually snaps back.
Yet this is not a simple echo. Unlike the dot-com era, where many ‘pure plays’ were little more than websites with a stock ticker, today’s AI firms are producing real tools. I have personally tested language models that write passable poetry and image generators that can fabricate anything from a corporate headshot to a work of art. But the revenue models remain nascent. The enterprise adoption is real but incremental. We are asking the market to price a technology that is still fumbling for a business model beyond chatbots and coder assistants.
The Bank of England’s Financial Policy Committee has not yet intervened, but the language from Threadneedle Street has shifted. In a speech last Thursday, Governor Andrew Bailey warned of ‘elevated risk premia’ and ‘information cascades’ in technology stocks. This is code for a bubble. When central bankers start talking about behavioural finance, the party is over.
So what happens next? Three scenarios. The first, the ‘soft landing’, is a rotation out of pure-play AI into diversified tech and value stocks, deflating valuations gradually. The second, the ‘correction’, is a 20-30 per cent sell-off across the sector, cleansing the froth but preserving the fundamental growth story. The third, the ‘crash’, mirrors 2000: a cascading panic that claims not only the overvalued but also the solid players, as confidence evaporates faster than server memory.
My instinct, shaped by years of watching the valley go through cycles of hype and hangover, is that we are in the second scenario. The fundamentals of artificial intelligence are genuine. The technology will reshape industries. But the market, in its relentless hunger for the next big thing, has paid a 50 per cent premium for a future that is five years away. The correction is not a rejection of AI. It is a rejection of speed.
The City of London’s plea for calm is not a lie. It is a necessary balm for a market in shock. But the warnings are real. If you hold AI stocks, do not panic. But do not pretend this will be painless. The algorithm, it turns out, cannot be tasked with pricing itself. That job belongs to the humans. And we have a notorious track record of mistaking a moat for a flood.
