Ahmedabad Times

Jefferies’ World Head of Fairness Technique Chris Wooden is warning that the AI buying and selling will finally be damaged not by a sudden collapse in demand for chips, however by a market-wide realisation that hyperscalers and main AI labs can not earn an ample return on the huge capex they’re enterprise. He sees considerations over “malinvestment” as the particular threat that may lastly set off the tip, or at the least a painful pause, within the AI growth.

In his newest publication Greed & WorryWooden describes the continuing AI construct‑out as “probably the most dramatic capex cycle” he has ever seen, pushed by hyperscalers and foundries racing to ramp knowledge‑centre and compute capability. TSMC, for example, has lifted its capex steering for 2026 to about US$56 billion from US$41 billion final yr, with Jefferies’ Taiwan accomplice Fubon Analysis now projecting US$65–70 billion of capex in 2027.

This surge in funding is already translating into growth‑like macro circumstances in Taiwan, with actual GDP progress hitting 14.55% yr‑on‑yr in 1Q26 and export orders up 53.4% yr‑on‑yr within the three months to Might. Wooden notes that AI‑associated demand now accounts for an estimated 31% of TSMC’s revenues in 2026, underscoring simply how concentrated the cycle has develop into in AI infrastructure.

Additionally Learn | 3 AI shares outweigh all of India: Why this focus is sounding EM alarm bells

Jevons Paradox and “Picks and Shovels” Winners

Wooden frames the AI demand story by means of the lens of Jevons Paradox: as token prices fall and effectivity improves, whole compute consumption rises as an alternative of falling. “The elevated demand triggered by cheaper costs must be good for the picks and shovels performs,” he writes, arguing that DRAM and reminiscence suppliers are the first fairness beneficiaries three and a half years into the AI capex arms race.

He cites Micron CEO Sanjay Mehrotra’s remark that “reminiscence has developed from a peripheral element into the core engine driving productiveness within the AI period,” including that the massive three DRAM makers now have ample leverage to lock in lengthy‑time period gross sales agreements. Micron has already signed 16 strategic buyer agreements overlaying roughly 20% of its DRAM quantity and a 3rd of its NAND quantity, sometimes with 5‑yr tenors — proof, in Wooden’s view, of structural change within the business.

The Commoditisation of AI Fashions

A key contextual threat is the speedy commoditisation of huge language fashions, notably within the shopper and, more and more, company markets. Wooden highlights the launch of GLM‑5.2 by Hong Kong‑listed Z.ai, previously Zhipu AI, noting that knowledgeable sources describe the brand new mannequin as “virtually equal to Anthropic” for company use at only one quarter of the fee per token.
This comes towards a backdrop of a backlash towards “tokenmaxxing” and explosive progress in cheaper Chinese language fashions on platforms comparable to OpenRouter. Within the week ended 21 June, prime Chinese language AI fashions processed 21.37 trillion tokens on OpenRouter, up from 4.37 trillion in late April, versus 5.76 trillion tokens for the highest US fashions. That shift in quantity, he argues, is already signalling a commoditised panorama and mounting stress on the economics of premium Western AI suppliers.

Malinvestment in AI

Wooden is express that the important thing vulnerability within the AI commerce just isn’t a traditional semiconductor oversupply shock however the eventual recognition that the hyperscalers and main AI labs will fail to earn a passable return on their funding. “GREED & concern is personally satisfied that considerations about malinvestment would be the almost certainly set off for an finish to the AI commerce, or at the least for a protracted pause to refresh,” he writes.
The hazard, in his view, lies in round funding preparations and aggressive capability growth constructed on optimistic monetisation assumptions. He factors to buildings comparable to Nvidia financing OpenAI in order that OpenAI can in flip purchase extra Nvidia chips — a suggestions loop that works so long as traders are keen to bankroll the ecosystem however may unwind sharply as soon as doubts over lengthy‑time period returns take maintain.

Why Conventional Provide‑Facet Dangers Are Secondary

Traditionally, semiconductor cycles have tended to finish with abrupt will increase in provide and stock gluts. Wooden believes the present cycle is completely different. “The important thing level to notice for now could be that that is the way in which the cycle is almost certainly to finish relatively than due to a sudden enhance in provide, as has historically been the case in semiconductors,” he argues, emphasising that DRAM makers now command far higher pricing energy.

The dominant three DRAM producers have been in a position to negotiate multi‑yr strategic buyer agreements, and Wooden relays market chatter that Korean reminiscence makers are already regretting locking in lengthy‑time period phrases as a result of they count on chip costs to rise additional. In such a structurally tight business, the extra believable finish‑of‑cycle set off just isn’t oversupply however investor capitulation over capital self-discipline and earnings visibility within the AI stack above reminiscence.

Regardless of these structural considerations, Wooden emphasises that there’s “zero signal of AI capex slowing” but. He hyperlinks ongoing spending to US banking deregulation underneath the Trump administration, citing Alvarez & Marsal’s estimate that latest regulatory modifications will unlock US$2.5 trillion in further lending capability throughout the US banking system, together with US$1.1 trillion unlocked within the final two quarters.

Portfolio Implications For Wooden

Importantly, Wooden just isn’t calling for an instantaneous collapse in AI‑linked equities; as an alternative, he’s re‑positioning in the direction of {hardware} and reminiscence names that he believes will stay lengthy‑time period beneficiaries even when the AI commerce endgame is outlined by capital‑return disappointments increased up the stack.

He’s lifting publicity to tech {hardware} throughout GREED & concern’s mannequin portfolios, including SK Hynix and Kioxia with preliminary 4% weightings within the international lengthy‑solely e book and rising the allocation to Samsung Electronics. Alphabet and Alibaba are being faraway from the worldwide portfolio, reflecting a deliberate tilt away from huge‑cap platform performs in the direction of “picks and shovels” beneficiaries of the AI capex cycle.

(Disclaimer: Suggestions, options, views and opinions given by the consultants are their very own. These don’t symbolize the views of The Financial Occasions)