Demography and technology are boosting growth in the sector. Ozge Brinkworth sees potential for companies that can either introduce entirely new products or enhance existing ones to improve efficiency and clinical outcomes.
Article last updated 5 December 2025.
From baggy jeans reappearing on catwalks to the Gallagher brothers reuniting, throwbacks to the 1990s have been hard to avoid recently.
The stock market has been no exception. The AIled surge in US tech stocks is drawing inevitable comparisons with the almighty dotcom boom of the second half of that decade – not to mention warnings about the way it ended.
Yet history never repeats itself exactly. Just as there will never be anything quite like Knebworth ’96 for Oasis fans, so today’s tech boom is not a perfect mirror of the dotcom boom. In navigating today’s market, we think there are important lessons in both the similarities and the differences with that era. The similarities with the dotcom boom are obvious. Bold claims from evangelistic CEOs about the potential of a new technology to revolutionise every aspect of dayto- day life are highly reminiscent of the 1990s internet revolution. And the valuation of the US stock market has climbed to levels last seen in the dotcom period, as measured by prices relative to companies’ profits. As in the 1990s, tech stocks have led the ascent and command particularly high valuations.
But what about the differences?
Investment inches up
The dotcom boom was very broad. It saw a surge in IT investment far above its long-run trend – big enough to drive investment in the economy more generally much higher.
Turning to today, for all the billions of dollars spent on AI so far, the current increase is nothing like as large, relative to the size of the US economy. Total US private investment rose from around 12% of GDP in 1994 to a peak of nearly 15% in 2000. In contrast, it has essentially been flat recently, inching up by just 0.2pp of GDP since the November 2022 launch of ChatGPT, the AI chatbot.
The current wave of spending is also remarkably concentrated in the hands of far fewer companies. Five ‘hyperscalers’ – Amazon, Alphabet (Google), Meta, Microsoft and Oracle – will together invest over $350bn this year, and probably much more in 2026. They account for about two-thirds of all the investment spending by listed companies with a strong connection to AI.
The dotcom boom was also fuelled by debt in a way that the current AI wave, so far at least, is not. Debt in the US private sector rose, relative to the size of the economy, through the 1990s. And firms’ balance sheets became more stretched. Their ability to service their debts deteriorated as they took on debt to fund internet-related expansion. Debt directly helped charge the rally in the stock market too. Margin debt – borrowing to buy stocks – surged, sending the market soaring further. Exuberance wasn’t limited to publicly traded stocks either. Private equity fundraising exploded, as did mergers and acquisitions (M&A); both were largely funded by debt.
But today, the hyperscalers spending so much on AI are extraordinarily profitable. They have net cash flow not far below $600bn this year – and rising rapidly. Their investments have been funded primarily out of those profits, so net issuance of debt has been minimal and balance sheets remain extremely strong. Looking at the US as a whole, business (and household) debt there has been falling relative to the size of the economy since the post-pandemic reopening.
Notwithstanding some eye-watering AI-related activity in private markets, there’s been no broader surge in either private equity or M&A activity either.
Mighty IT: Investment in IT is rising as a % of GDP, but the dotcom surge was much more extreme.
Vigilance is vital
Pointing out these differences is not a call for complacency, however. Instead, it’s a challenge to look more closely at where the risks lie. It’s important to be vigilant about debt funding. It creates inherent fragility if borrowing costs rise or lenders’ sentiment sours. Debt almost certainly amplified the dotcom boom and bust, fuelling more spending on the way up, and the subsequent bust, with many more underwater investments and bad debts to work through on the way down. That hit the stock market. So it’s positive that so far, there’s been no such debt surge connected to AI investment. But experience suggests we should keep a close eye out for that changing. We should also be wary if activity in private equity markets or M&A starts to take off as both did before the dotcom peak, or if there’s a wave of AI-related companies floating on the stock market.
On the other hand, the concentration of AI investment in a few hands creates a different set of risks not present to the same extent in the 1990s. The decisions of just five companies have a disproportionate bearing on the entire AI supply chain. The business models of at least three of the other 10 largest companies in the world, by market cap, are heavily dependent on the hyperscalers’ AI-related demand – and that’s just the tip of the iceberg.
Deep pockets
Understanding the whole AI ecosystem requires an understanding of how hyperscalers behave and why. Their deep pockets, including cash reserves of hundreds of billions of dollars between them, suggest that financing isn’t an immediate constraint on their grand investment plans.
But these plans could slow down for other reasons. There’s a risk, for example, that the hyperscalers struggle to make money from their new investments as quickly as hoped. That would sharply shrink the gap between the cash they generate and their investment spending. Investors are already braced for this gap to narrow below its previous trend in 2025–26, before starting to grow again. But if it contracted by much more than expected, the hyperscalers might scale back their investment plans. This would ripple through the whole supply chain.
With that in mind, we need to monitor what’s happening to the hyperscalers’ cash flows, the guidance they provide on their investment plans, and how they intend to fund their investment. This is crucial to understanding the direction of the AI industry more generally.
In conclusion, there are valuable lessons to be learned from looking back to the 1990s, especially about the importance of debt. But it’s equally important to analyse the risks to this market on its own terms. This means keeping a particularly close eye on the behaviour of the key players that drive it.
Mind the gap: Hyperscaler free cash flow (cash flow minus investment) is expected to narrow, then widen
