The London markets are awake to a familiar scent. It is the smell of overpriced optimism. It is the scent of capital chasing a story rather than substance. And today, the story is artificial intelligence. The question is not whether the AI sector is transformative. It is. The question is whether the market has priced in a decade of earnings in a matter of months. The answer, for anyone who has watched a boom before, is deeply uncomfortable.
Consider the numbers. The Nasdaq 100, weighted heavily by the ‘Magnificent Seven’ tech names, has surged over 50% in the past year, driven largely by AI enthusiasm. The forward price-to-earnings ratio for the tech sector has blown past 30, a level that historically precedes corrections. The yield on the 10-year gilt, meanwhile, has drifted above 4%, a silent warning that the cost of capital is rising. When bonds offer a decent risk-free return, equity multiples become harder to justify. It is basic arithmetic.
Yet the narrative persists. Every earnings call now features the word ‘AI’ as a talisman. Chief executives, desperate to justify their valuations, spin tales of productivity miracles and revenue streams that have yet to materialise. The market, always a credulous beast in bull cycles, laps it up. But the sceptics are sharpening their pencils. Goldman Sachs, not normally a Cassandra, has warned that the AI trade is looking ‘crowded’. That is polite banker speak for ‘top heavy’.
Let us examine the balance sheets. The biggest beneficiaries of the AI boom are not the software firms but the hardware suppliers: the makers of chips and data-centre equipment. Their profits are real, but the market has extrapolated from a steep slope into a vertical cliff. Meanwhile, the supposed end users of AI many enterprise software companies are spending heavily on AI infrastructure with little to show in revenue. This is classic capex froth, reminiscent of the dotcom era when companies built networks that consumers would eventually use, but the investors who paid for the wires lost their shirts.
The parallel with the dotcom bubble is inevitable and instructive. In the late 1990s, the internet was genuinely revolutionary. It did change everything. But the stocks that rose 1,000% then fell 90% were not wrong about the technology. They were wrong about the timing and the distribution of profits. The same dynamic is at play today. AI will indeed reshape industries. But the companies that dominate the conversation today may not be the ones that harvest the profits tomorrow. The infrastructure plays, like semiconductor makers, are more defensible. But their multiples already discount a golden age that may take years to arrive.
What could pop the bubble? A hawkish turn from central banks would be the classic trigger. The Federal Reserve and the Bank of England have signalled caution on rate cuts as inflation proves sticky. Higher rates for longer compress valuations, especially for stocks with distant cash flows. Alternatively, a profit warning from a bellwether such as Nvidia would shatter the consensus. If future earnings disappoint, the unwind could be violent. Liquidity in the AI sector is thin; a stampede for the exit would be messy.
The more worrying possibility is a loss of faith. Bubbles are psychological constructs. They require buy-in from a broad base of investors. If retail money starts to flow out of AI ETFs and into money market funds, the reversal could be self-reinforcing. Margin debt, the fuel of bull markets, has risen sharply. A margin call cascade is ugly when it begins.
None of this means AI is a mirage. It is not. The technology holds genuine promise. But market prices are disconnected from intrinsic value. For the prudent investor, the wise move is to trim exposure, lock in some profits, and wait for the correction that inevitably follows euphoria. Timing the crash is impossible, but recognising the risk is essential. The City of London has seen this play before. The name of the sector changes, but the music always stops. The question is whether you are still dancing when it does.
