The rolling bubble: stop asking whether AI is one, ask which layer is repricing

On August 3, Dhaval Joshi posted on LinkedIn that AI is not one bubble waiting to pop but "a rolling SEQUENCE OF BUBBLES," in which capital leaving one deflating sector inflates the next one in line ^[1]. Joshi, until recently chief strategist for the Counterpoint strategy at BCA Research in London ^[2], argues the useful question is not whether AI is a bubble but which AI bubble is popping today ^[4].

That reframing is more useful than the bubble-or-not argument because it implies a rotation rather than a single date. Joshi's account is that investors keep misjudging, then correcting, who or what actually captures AI's value ^[5]. By his own chart, three layers have already been through the cycle ^[23]. Software rallied on the idea that AI would be a productivity tool, then crashed once investors decided AI agents threatened the SaaS subscription model itself: "So, the software boom turned to bust" ^[6]. Silver spiked on its status as the best electrical conductor for power-hungry data centres, which Joshi says "could not justify a near trebling of the silver price when there are other good conductors" ^[8] - roughly a 200 per cent advance to unwind ^[24]. Semiconductors rose on the premise of near-limitless pricing power; Joshi says chipmakers have no moats around their profits, and "astronomical margins will crash back to earth when demand and supply equilibrate, as they ultimately must. So, the semis boom is unwinding - though has further to go" ^[9].

The software episode is the one with a direct read-through for anyone selling seats: the derating was not a verdict on AI working badly, it was a verdict on the pricing model ^[6]. Joshi frames the wider problem as a margin question rather than an earnings question. He told Fortune he slightly disagreed with BCA's Peter Berezin that this is an earnings bubble, calling it a "profit margin bubble" instead, with the market now asking "How is the E high?" and whether those margins can be held ^[10].

The obvious objection is that this is just price discovery, and Fortune puts it directly: markets test a thesis, find it wrong, correct ^[12]. Joshi's answer is amplitude. "If you can make a fortune in a matter of weeks or months, and, crucially, then lose it all just as quickly or even quicker, then that constitutes a 'bubble'" ^[11]. He describes something closer to narrative contagion that briefly grips a sector and then rolls elsewhere, and notes the silver leg shows the misallocation is not confined to equities ^[13].

The consolation, for now, is mechanical: because each deflation coincides with a reinflation somewhere else, there has been no correlated selloff ^[15]. That is also the fragility. Joshi has argued since July 2023 that AI hype would not automatically become profits for the leading tech companies, drawing parallels to the 1990s ^[19], and the 1990s precedent is unkind to funders: the buildout created enormous value, but the fibre that bankrupted its original owners still carries the internet for entirely different shareholders ^[21].

Watch the liquidity that powers the rotation. A rolling sequence needs a steady stream of optimism and capital, and a deteriorating economy could slow or reverse the flow between sub-sectors ^[22]. Watch hyperscaler free cash flow, which the latest earnings season showed being consumed by capital expenditure, with Google going free cash flow negative for the first time in its history ^[18]. And note that the consensus is already crowded: Jamie Dimon has repeatedly flagged elevated valuations, Bank of America's Global Fund Manager survey has named an AI equity bubble the top tail risk, and Sam Altman, David Solomon and Jeff Bezos have all conceded something bubbly is happening ^[16] - while the pop that was due in 2025 has not arrived ^[17].