Revolutionary technology, a flurry of blockbuster IPOs (which, some argue, boil down to a risk-transfer), massive valuations built on massive losses. The technology may be novel, but the market dynamics make bankers who experienced the rise and fall of companies like Pets.com and Webvan wake up in cold sweats. The undeniable parallels make it unsurprising that the dot-com bubble of the 90s and 2000s is frequently compared to today’s AI boom, but how seriously should we take this cautionary tale?

Déjà-valuations

Dwarfing all IPOs to have come before it, and as a precursor to the similarly gargantuan IPOs that will come after it, the recent IPO of SpaceX is an informative reading of the stock market’s barometer. Frenzied demand and a rocketing share price for the lossmaking AI company – yes, SpaceX is now an AI company – showed that there is still belief in the AI narrative, and, for now, there is plenty of money to go around.

Subsequently, SpaceX’s stock price crashed – falling around 30% from its highs, in less than 10 days. The self-administered salt in the wound then came last week with its $25 billion bond offering that, to put it mildly, was less than stellar.1 This tells an equally valuable story: the AI trade may be losing its lustre. For a market that runs on sentiment, this could spell disaster.

SpaceX had its moment as the hottest new trade in the AI space, however it is only the most recent in a long line of stock-market-phenomena. In recent times, it has become commonplace for companies within the AI-sector to see their market cap increase up to eight-fold. The Philly Semiconductor Index – a weighted index of semiconductor manufacturers – is up approximately 150% since 2025. It is undeniably a boom, but its characterisation as a ‘bubble’ is a contentious topic where the stakes have rarely been higher.

The good news is, it’s different. But that’s also the bad news

Though finding dissimilarities may help investors sleep at night, that search almost immediately uncovers an uncomfortable truth. A key difference between the dot-com bubble and today’s boom is scale. As Joachim Klement, a managing director at Panmure Liberum, pointed out in a recent article,2 in 2025 alone, inflation-adjusted-investment into technology by US firms was almost double the amount reached at the height of last generation’s bubble. He goes on to write that “the US economy is growing solely because of the tech boom”, and “over the past four quarters, 93% of US GDP growth was explained by tech investments. Even at the peak of the TMT bubble, it barely reached 60 per cent.” These figures are set to grow further, almost guaranteeing that we will not see history repeated: if there is a crash, it will be much larger.

Profitability and tangibility also differentiate today's AI front runners from the notoriously lossmaking and illusionary firms of the TMT-bubble era. Nvidia, Microsoft, Google, such firms are highly profitable stock-market behemoths with real business models and widely utilised products. Their dot-com counterparts, on the other hand, were unprofitable, their business models were unproven, their products' utility was unclear, and the narratives bolstering interest in the flurry of IPOs were, sometimes, ungenuine. This may not rule out a bubble, but it is important to stress that the structure of this boom is far more robust – on spreadsheets and in use-cases – than the one a generation ago.

Unfortunately, however, the thick wads of cash that should contain a potential blast seem to be thinning. Klement sheds light on the “numbers” behind the data-centre-investment by some of the major players in AI-boom – namely Microsoft, Meta, Amazon, Alphabet and Oracle.2 He calculates that even if costs are assumed to be zero, “implied return on investment is highly negative for all of them except Amazon.” Concluding, from this, that ”if the hyperscalers continue on the current trajectory, the AI boom will become a story of one of the largest destructions of shareholder value in history.”

Put this way, the market looks frothy, and the gravity of the situation seems hard to overstate. Even a small downturn in investment – which is contingent on continued confidence in, and demand for AI – could send the US economy into a recession and at the same time send the stock-market into free-fall. The SpaceX case-study looks to be a microcosm of what reduced ‘AI-hype’ could do to equities, and it does not bode well.

Not a bubble? Could still be bad

In a recent interview with the FT, the chief investment officer at Schroders stated their belief that the AI boom is not yet “a bubble situation”.3 A positive outlook from asset managers is reassuring, however even if the AI boom is not a ‘bubble’ in the traditional sense, it could still have major ramifications.

Chiefly, there are fears that the huge IPOs on the horizon, along with the selling of new shares and bonds by hyperscalers across the AI-space could suck in investment that would otherwise go to other parts of the economy currently calling for investment, such as renewables and defense. We must recognise that the investment-booms that accompany technological revolutions provide long-run benefits in the form of infrastructure, the 19th century railroad boom for example. However, long term benefit does not rule-out a market crash: as Robert Armstrong at the FT highlights, “in their later stages, booms are characterised by excessive confidence and malinvestment.”4

Indeed, considering previous technological revolutions, it seems an overshooting of investment is practically unavoidable. The dot-com bubble explicitly illustrates that no matter how capable a technology is of revolutionising almost every facet of human endeavour, humans can still get carried away.

Bubble or not, this eagerness and unmeasured approach is ever present in the AI-boom, in fact, it is there by design. To the companies that are spear-heading the capex tsunami, the existential risk that AI-disruption poses to their business far outweighs that of misallocated investments. Actually, Meta’s continued investments into the ‘Metaverse’ seem to indicate that to hyperscalers, the latter poses no issue at all. Furthermore, the guardrails, if any, are flimsy as the US government operates under the relatively safe assumption that regulation could stifle progress; an unappealing idea given the AI arms-race with China.

AI’s indirect revolutionisation of markets has also made the walls of a potential bubble somewhat thinner. In recent decades, ‘Big Tech’ firms – more than any other sector – have consistently bought-back shares, being a main force behind the ever-shrinking supply of publicly available equity. The trillions invested in the ‘AI-build-out’ have greatly reduced free cash-flows for the very ‘Big Tech’ firms leading the ‘de-equitisation’ trend, meaning that if firms want to continue buying back shares, they will have to borrow to do so. The fear being that the increased cost may reduce, or entirely remove, an important portion of demand, which could act as an airbag in the case of a crash.

As with all preceding booms, a correction is most likely on the horizon. Contingent not on the AI boom being a bubble, but, instead, on the nature of investment cycles. They run on sentiment and feed on balance sheets, with their hunger depending on risk. Excess strain on any one of those elements, and they die. The true extent of the damage is anyone’s guess, but, as Armstrong put it, “investment cycles turn, and the bigger they are, the more wrenching the turn is.”4

Different engine, same cliff

The mechanics are slightly different, but the premise is a tale as old as time. The technology could change everything so the risk to reward ratio is undefined. Getting there first means everything. The incentives align with the actions: full speed ahead.

And if we need to pump the brakes? Well, there are none anyway.

Footnotes

  1. William Gavin, SpaceX’s new bonds are flashing a warning sign, as investors pump the brakes on AI frenzy (opens in a new tab), MSN, 26 June 2026. ↩

  2. Joachim Klement, The impossible maths of the AI boom (opens in a new tab), Financial Times, 20 May 2026. ↩ ↩2

  3. Alan Livsey, AI and tech groups not in ‘a bubble situation’, says Schroders CIO (opens in a new tab), Financial Times, 17 June 2026. ↩

  4. Robert Armstrong, AI is revolutionising the stock market (opens in a new tab), Financial Times, 13 June 2026. ↩ ↩2