Not One Bubble but Many: How a Strategist Reads the AI Investment Market

Asking whether artificial intelligence is in a bubble is the wrong question, says Dhaval Joshi, former chief strategist at London-based BCA Research. The right question, he argues, is: which AI bubble is bursting today? Unlike the classic model — one enormous bubble that inflates and then pops all at once — Joshi describes "rotating bubbles" that swell and deflate in sequence: investors misjudge which part of the AI chain will capture the value, then correct that mistake, and attention rotates to the next segment.

The theory accounts for several market moves in recent months. Software stocks (SaaS — software delivered to customers as a cloud service) climbed on the promise that AI would boost workplace productivity, but prices tumbled once AI agents began to threaten the subscription model itself — a sell-off dubbed the "SaaSpocalypse." Silver followed a similar arc: its price surged on demand from data centers for the conductive metal, then slid as it became clear the rally couldn't be justified given the availability of alternative conductive materials. Semiconductor stocks also rallied on the assumption of virtually unlimited pricing power, yet Joshi argues these companies are not sufficiently insulated from competitors and their fat margins won't hold.

Joshi calls this a "margin bubble" rather than a "revenue bubble": the issue isn't that share prices are unjustified relative to earnings — the question is whether companies can sustain today's elevated margins over the long run. If an asset spikes sharply over a few weeks or months and then collapses just as fast, he says, that is a genuine bubble; ordinary market self-correction is never that abrupt or that quick.

Joshi is not alone. Jamie Dimon has repeatedly warned about richly valued assets, and Bank of America's fund manager survey flagged an "AI stock bubble" as the market's biggest risk. Even OpenAI's Sam Altman, Goldman Sachs CEO David Solomon and Amazon founder Jeff Bezos have acknowledged bubble-like signs. According to Reuters calculations, capital expenditure at Microsoft, Alphabet, Amazon, Meta and Oracle is expected to exceed their free cash flow by 2027; Google's free cash flow has already turned negative for the first time in the company's history.

Joshi expects AI capital spending to peak in late 2026 or the first half of 2027. He sees three scenarios for who ultimately captures the value AI creates: first, corporations with powerful competitive moats such as Amazon or Google take it all; second, highly paid "star professionals" — lawyers or consultants, for instance — win by cutting their costs with AI; third, competition grows so intense that nobody holds on to margin and the real winner is the ordinary consumer, as prices fall. As a vivid illustration he points to the 600% jump in a year in the price of the all-but-obsolete, 20-year-old DDR3 memory chip — the equivalent, he says, of paying $50,000 for a beat-up 2007 Toyota Corolla.