By Mike Dolan
LONDON, June 30 (Reuters) – Although some deem it blasphemous to suggest the AI boom is another bubble, the world’s central banking forum and watchdog has warned against getting swept up in the frenzy, just as investors appear increasingly reluctant to second-guess it.
In its latest annual report on the state of world finance, the Basel-based Bank for International Settlements on Sunday didn’t doubt the scale of the AI buildout — but it did fret about the hangover when the spending crests.
Even as the U.S.-Israeli war on Iran rumbled in the background, U.S. chipmaker stocks staged a record 75% rally in the second quarter of 2026 — driven by yet another wave of rising capital expenditure forecasts from hyperscaler giants racing to build out AI infrastructure, stoking supply bottlenecks and chip shortages in the process.
That pushed U.S. earnings growth estimates for 2026 to nearly 25%. Forecast AI capital spending by the five biggest hyperscalers is approaching $1 trillion this year, and Goldman Sachs reckons the cumulative total could be $7.6 trillion by 2031.
For the umpteenth time, is this a bubble, anyone?
Not a bit of it, claims one of the tech sector’s biggest investors and cheerleaders. Only last week, SoftBank boss Masayoshi Son suggested it offended the gods of digital superintelligence to even think such things.
“It’s blasphemy against AI if you say it’s a bubble,” Son said. “It’s just the beginning. AI’s potential will be unlocked.”
With hundreds of billions of dollars at stake if it all comes good, perhaps he’s bound to say that.
But other investors, maybe wearily, are beginning to concede that the explosion in spending and stock prices may not be the bubble they once feared after all.
Deutsche Bank said on Tuesday its latest quarterly client survey showed the lowest perceived bubble risk since 2021 for the so-called Magnificent Seven megacap stocks — mostly those companies doing the big AI spending.
For the broader U.S. tech sector, however, those risk perceptions remained as elevated as they have been for the past two years.
That divergence may explain why, even as chipmakers posted their best quarter on record, June was the worst month ever for the Mag 7 since the grouping was conceived three years ago.
To be sure, bearish capitulation cuts both ways as we hit the half-year mark.
For many chipmaker stocks, the revenue numbers are simply very real, and share price moves may be rationally reflecting that.
Micron Technology, now valued at $1.25 trillion, has seen its stock more than treble since March. But revenue estimates have risen just as sharply, leaving its 12-month forward price/earnings ratio virtually unchanged this year and, at just eight times, less than half its level two years ago.
Even if a near-trebling of the share price of still-loss-making Intel stands out as something of an outlier, valuations of chipmakers such as Broadcom and Qualcomm remain historically contained.
AI EATS ITSELF?
And yet, heresy or not, it falls to financial stability watchdogs such as the BIS to detail what could go wrong for the wider markets and economy. And, just like SoftBank’s Son, maybe they’re bound to do that too.
The BIS report dwelt mainly on risks to the sustainability of the pace of investment spending, compounded by a race among a small group of firms built on the belief that only a few players with superior technology will ultimately dominate market share.
The watchdog warned that fierce competition could lead firms to pour too much money into AI projects whose returns remain uncertain, leaving the whole sector exposed if the expected payoffs fall short.
With competitive pressure driving capex ever higher, it said, the sector’s overall payoff after investment costs could shrink or even turn negative in adverse scenarios.
“Disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust, with potential knock-on effects on financial conditions.”
Another warning flag detailed by the BIS was supply bottlenecks in power generation, electricity grids and memory chips. These could force firms doing the big spending to overcommit via longer-dated contracts to secure scarce supplies, exposing them to over-investment and making them more vulnerable to any demand disappointment down the line.
On that score, the report reserves its sharpest warning for the risk that AI ultimately eats itself.
In an extreme scenario in which AI proves as capable of replacing human work and intelligence as its biggest acolytes suggest, the risk is that more income is diverted from labour into further AI investment. Taken far enough, workers’ share of national income could fall towards zero, leaving fewer people with the purchasing power to buy what the economy produces.
Anticipating that demand problem, the BIS suggests, forward-looking firms would eventually pull back from further investment.
“Productivity stalls not because of technological limitation, but because the demand to justify further capacity expansion is missing. The demand bottleneck becomes the binding constraint.”
Blasphemy or not, an awful lot of money in this AI arms race is still based on faith.
(The opinions expressed here are those of Mike Dolan, a columnist for Reuters.)
Enjoying this column? Check out Reuters Open Interest (ROI), your essential new source for global financial commentary.
Follow ROI on LinkedIn, and X.
And listen to the Morning Bid daily podcast on Apple, Spotify, or the Reuters app. Subscribe to hear Reuters journalists discuss the biggest news in markets and finance seven days a week.
(by Mike Dolan; Editing by Marguerita Choy)
