AI is revolutionising the stock market
Big Tech no longer prints money; it needs it. Are the companies irrational in abandoning old business models for something less profitable and more uncertain — and what happens when confidence dips?
Financial Times UK 13 Jun 2026
By Robert Armstrong
Artificial intelligence will change the way we all live and work, but, even putting that aside, the AI boom will leave a permanent mark on the markets and the economy.
The huge surge in investment the technology has triggered, and the rush to finance it, will alter the landscape in profound ways. This week’s SpaceX flotation is just the latest sign: we are facing regime change not just in technology but in finance.
In finance, the regime change begins with a handful of companies that have towered over stock markets for 20 years or more. Consider just four: Alphabet, Microsoft, Meta and Amazon. Until recently, you could argue that the first three had the most profitable business models in corporate history: a combination of unassailable barriers to entry, nearzero marginal unit costs, manageable investment requirements and growing target markets.
These three have printed money: in 2024, together they generated free cash flow (cash profits after investment spending) of over $200bn, four times the level of a decade before.
Amazon is a slightly different case: its ecommerce business is low margin and its cloud computing business is investmentheavy. But its dominance in its markets helped it become a cash machine over time: $33bn in free cash flow in 2024, against $2bn 10 years before.
Together, the four groups have been a major force pushing US markets ever higher. But when these software/internet companies decided to become AI “hyperscalers,” everything changed.
Wall Street analysts expect capital investment at the four, overwhelmingly spent on data centres, to rise sixfold between 2023 and 2027, to a staggering $815bn — with the result that free cash flow is expected to fall by 70 per cent.
Are the companies irrational in abandoning their old business models for something less profitable and more uncertain? No: they are betting that if they do not invest in AI, barriers to entry will fall and their old models will be destroyed.
The wager may ultimately be proved wrong, but it is logical enough. (It is interesting, though, that another cash machine company is betting the other way. Apple thinks its consumer hardware/services combination can take advantage of AI without investing in models and data centres of its own.)
These cash machines are, at least for now, spitting out much less cash. This will have big consequences for markets.
The most obvious is the departure from the market of the biggest single buyer of those companies’ shares: the companies themselves. In the past 10 years, Alphabet, Meta and Microsoft have bought back almost $800bn of their own shares. (Amazon rarely does buybacks.)
When such groups buy their own equity, they are supporting Nasdaq, the S&P and many other bourses, since the ups and downs of Big Tech stock often determine the movements of share indices across the world.
Indeed, one of the major trends of recent decades has been “de-equitisation” as big companies bought back their own equity and many smaller groups were bought by private equity, using mostly debt.
But now, if the Big Tech groups want to buy back their shares, they will increasingly have to borrow to do it. As a result, a major source of support may have left the whole market.
“When a correction happens — which it will — it is harder for these guys to prop their stock up than it was before, as when Meta bought lots of Meta shares back after its stock crashed [in 2021-2022],” says Robert Buckland, who popularised the concept of de-equitisation as a strategist at Citigroup.
Will the rising valuations and prices that characterised the de-equitisation era persist in an era of re-equitisation?
Alphabet, Microsoft, Meta and Amazon could, if they wanted to, still pay for their AI investments out of their profits. Other companies, soon-to-be-public and public, cannot.
SpaceX is the most famous example: it burned through $9bn in cash just in the first quarter of this year. Early success for its IPO this week was all but locked in by imminent demand from passive investors — demand Nasdaq accelerated by changing its index inclusion rules.
Observing SpaceX’s capital haul, others will follow. Alphabet, Google’s parent, is already raising $85bn by selling new shares; Meta, which owns Facebook, Instagram and WhatsApp, maybe next. OpenAI and Anthropic are unlikely to wait long. If things go as hoped,
just those five companies could raise $400bn this year. Others seem likely to jump on the bandwagon Would half a trillion dollars put undue strain on investors’ appetite, triggering selling of other stocks and undercutting the bull market? In a purely monetary sense, the answer is no.
The market has plenty of buying capacity. Investors and institutions have $8tn sitting in money market funds in the US, a sum that has been rising briskly for years. And the total market cap of the US stock market is over $70tn.
The larger point is that market regime changes are not purely, or even mostly, about buying capacity. They are dominated by a larger economic cycle, which is in turn largely defined by mood and confidence. The economic cycle that has built up around AI is much larger than the capital-raising cycle, and is potentially more significant for markets as well as the economy.
The cycle is tricky to measure: economic data does not distinguish neatly between tech and other sectors, or betwteen AI tech and other tech. But the broad upswing can be tracked.
Skanda Amarnath of Employ America, a think-tank, adds together investment in software, computer equipment and industrial equipment, as well as personal expenditures on hardware and software. Together those amount to over $2tn or nearly 7 per cent of GDP, he calculates. On this reckoning, the current AI/tech/data centre boom is about the same size, relative to GDP, as the 2007 housing boom at its peak, and bigger than the late 1990s tech boom.
And it is no coincidence that those previous booms coincided with market peaks followed by famously painful corrections. “Business cycles are about risk, about balance sheet commitment [to make investments],” Amarnath says — and high risk appetite is a transitory phenomenon. This does not mean AI is a bubble, just that investment cycles turn, and the bigger they are, the more wrenching the turn is.
This helps explain why Noah Weisberger of BCA Research, finds that historically “IPO waves are usually followed by more muted returns and [valuation] pressures . . . That does not necessarily mean that returns turn negative or that [valuations] fall, but it does suggest the most rapid phase of the rally is likely behind us.” He adds that very large IPOs tend to be followed by market returns that are “well below [the historical] average and more volatile”.
The AI boom is big enough to have an impact on the global macroeconomy, too. The massive demand for investment in the US will draw in savings from the rest of the world, deepening the US current account deficit. The higher demand for global savings will, at the margin, push up interest rates — just at a time when the world is also trying to finance big investments in the energy transition and defence. These non-AI investments could be crowded out.
Investment booms are — in general and over the long run — good. The railroad boom and bust left the US with lots of useful track; the telecom boom and bust left it with loads of fibre optic capacity. But in their later stages, booms are characterised by excessive confidence and malinvestment.
Certainly confidence is high right now. A vivid measure of current confidence is Wall Street analysts’ projection for future free cash flow at the four hyperscalers we started with. It is expected to bottom out in 2027 — and then grow at a rate much faster than anything in the companies’ history.
Wall Street estimates for more than a year or two in the future are always somewhat speculative. But it takes some pretty committed speculation to conclude that these companies will increase profits faster after entering a new, more capital-intensive and possibly more competitive line of business.
“When share prices go up enough, they become Giffen goods [where demand rises with price]. You get the wrong cost of capital, people pay themselves a lot of money and they over-invest,” says Ian Harnett of Absolute Strategy Research. “It becomes an abusive relationship” between investors and companies, he adds.
If the investment boom keeps expanding, the relationship will be abusive soon enough.
The break-up will be messy, in the style of 2001 or 2008. The economy and the market will survive, but will be left with scars. After the US housing bubble popped, residential construction never fully recovered, despite the dire need for new homes. After the dotcom bubble burst, stock valuations kept falling for 10 years.
Whatever AI technology does for us directly, its secondary effect on the market will be significant: de-equitisation to re-equitisation; a deepening of global saving and investment imbalances; and possibly a bust that will end with assets far beyond tech being revalued.
We are watching an old era die and a new one being born.
Data visualisation by Ray Douglas and Jonathan Vincent
