For much of the last two years, one beguiling idea has propelled financial markets ever higher. Artificial intelligence (AI), investors told themselves, would unleash productivity gains so spectacular that it could justify almost any valuation. From cloud computing giants like Oracle to obscure software installers, anything with a tenuous connection to large language models rode a wave of irrational exuberance.
Capital has flooded in, while investor scepticism has been almost non-existent. Commercial viability has been seen as a detail that will be sorted out in due course. Profits, it has been assumed, would inevitably be forthcoming once the vast infrastructure was in place.
That assumption is now being sorely tested. The turbulence that has roiled markets in recent weeks has been widely cast as a correction to speculative excesses. Yet this interpretation misses a more fundamental dynamic. What has unfolded is not merely a sell-off but a rotation; a reassessment of which technologies merit investment and which do not. Investors are not rejecting AI wholesale, but repricing who in the end will actually benefit from it.
The main US stock index, the S&P500, is pretty much flat year to date. This hardly suggests panic. But as the old Wall Street maxim goes, there is no such thing as a stock market, only a market of stocks. Beneath the placid surface, the picture is actually far more dramatic. Technology-heavy indices such as the Nasdaq have fallen further. Some other favoured sectors, such as software, are suffering double digit losses. A market that had, until recently, treated all AI-related stocks as profit machines is exercising greater discrimination.
AI is not just a one way bet
Several forces have converged to trigger this shift.
The first is AI’s gradual commercialisation. The technology has not, actually, disappointed. Rather ironically, in some areas it seems to work too well. An example has been the collapse in analytics and software firms after Anthropic, an AI company, launched productivity tools that could theoretically make much of what they do obsolete.
Second, there is a broader flight back to safety. Capital is flowing towards parts of the economy that were long neglected. Industrial companies, utilities, commodity producers and banks – sectors often dismissed as dull – have quietly outperformed since late last year. The old economy, it seems, is having a comeback.
Software, publishers, data‑services companies, financial information providers, alternative asset managers and gaming stocks have all dropped sharply on rising fears of AI‑driven disruption. So far this year, the software sector is down 16% while the unloved Stoxx Europe 600 index, crammed with companies such as commodity producers, utilities, industrials and financials, is up 4%.
Many asset managers have argued that the ultimate beneficiaries of AI will not be the companies that merely build the tools, but those that apply them most effectively. Manufacturers that optimise production, logistics firms that cut waste, healthcare providers that improve diagnostics; in the longer run, these may reap greater rewards than software vendors selling generic platforms.
Recent market movements suggest that this logic is gaining traction. Capital is shifting from enablers to users, from infrastructure to application, from promise to those that are delivering performance.
Third is the ongoing reassessment in so-called American exceptionalism. The capricious geopolitical, monetary and economic policies emanating from Trump’s White House continue to unsettle investors. What was once unthinkable – non US-based investors hedging US dollar exposure – is now openly discussed. For Trump, this must be deeply unsettling. Unlike the furore over tariffs, from which he simply “chickened out” and let markets bounce back, this rotation out of US tech is not something he can control.
Fourth comes a reckoning of Big Tech and crypto. For years the “Magnificent 7” mega-cap tech stocks have driven market returns. Since last October, however, they have flatlined and are now slightly negative. Furthermore, within this elite cohort, fortunes have diverged sharply. Tesla and Microsoft have tumbled on concerns about their enormous capital expenditures, while Alphabet has rallied on hopes it can successfully adapt its legacy Google search business for the AI era.
What was once a straightforward bet on technology – allocate to any of the Magnificent 7 and forget about it – has become rather more difficult. Lavish AI spending no longer guarantees investor approval. Last year, it seemed that management promises to spend ever more on AI infrastructure were invariably rewarded. Now pledges to increase capex spend are met with alarm. Investors want evidence of returns, not merely promises of future dominance. They are starting to demand to see the business models that will supposedly justify these extraordinary outlays.
In the last earnings season, which culminated last week, four of the largest technology companies in the US forecast capital spending totalling roughly $650-billion (R10.3-trillion) in 2026 – a staggering sum earmarked for new data centres. The outlays planned by Alphabet, Amazon, Meta and Microsoft, all chasing dominance in the emerging AI market, constitute an unprecedented spending bonanza. Each company’s estimate for this year approaches or exceeds its combined budgets for the previous three years. Any one of them would set a record for capital spending by a single corporation over the past decade.
On 28 January, Meta announced full-year capital expenditure could reach $135-billion (about R2.1-trillion), a leap of 87% from 2025. Microsoft on the same day reported a 66% increase in second-quarter capital spending, surpassing estimates. The market baulked, triggering a 13% drop, and the second-largest single-day decline in market value for any stock in history.
Alphabet on Wednesday, 4 February, rattled investors when it unveiled capital-spending forecasts of as much as $185-billion (R2.9-trillion), exceeding not merely analyst estimates but the outlays of vast swathes of American industry. Its shares were down 5% in minutes. Amazon on Thursday went even further, announcing plans for $200-billion (R3.1-trillion) in capital expenditure for 2026, sending its shares tumbling 8% as well.
Collectively, the quartet has lost more than $950-billion (R15.15-trillion) in market value since announcing their latest earnings and forecasts. Meanwhile, companies manufacturing AI computing hardware, such as Nvidia and Broadcom, have been steady or up.
Finally, there is cryptocurrency. The broader reassessment of risk tolerance augurs poorly for token prices and for “Bitcoin treasury” companies that buy them as an investment. Strategy, one such business, reported a $17-billion (R271-billion) operating loss in its latest quarterly results. The stock, which nearly joined the S&P 500 last year, fell a further 17% last Thursday, and is down roughly 70% from its post-election peak. What was saluted as corporate innovation now looks increasingly like reckless speculation.
An overdue rebalancing
Nevertheless, this rebalancing appears overdue. For too long, financial markets have acted as though profits from AI were inevitable. Reality is more complicated. Technology results in progress and hopefully higher productivity, but can make investors blinkered to the risks. Not all capex will be rewarded with sufficient returns on investment.
For many investors the lesson has been sobering. The era when virtually any company could attract investment merely by invoking AI is ending. What comes next will demand discernment, patience and a clearer understanding of where true value lies. The era of tough choices is beginning. DM