In his newsletter ‘GREED & fear’, Wood writes that the new quarter has opened “with much talk of ‘AI fatigue’ as investors look out for a peaking out of momentum and rotation into cheaper ‘value’ names which have not been part of the AI trade,” citing Tencent as one Asian example. He argues that sharp pullbacks in Korea’s AI leaders are “both natural and healthy” after “hyperbolic moves,” with the Kospi now down 22% from its 19 June peak and single‑stock leveraged ETFs on SK Hynix and Samsung Electronics dropping around 30% from asset highs.
Wood highlights just how extreme the AI run‑up has been: since the start of 2023, a market‑cap‑weighted basket of Micron, SK Hynix and Samsung Electronics has surged about 760%, versus a 180% gain for a basket of Alphabet, Amazon, Meta and Microsoft. “Long‑term GREED & fear would still rather own the DRAM makers,” he says, adding that “the demand for compute can keep growing even as the cost of tokens collapse,” while he has “no idea which of the hyperscalers, if any, are going to be successfully monetising their AI capex.”
Also Read | Korea’s pain will be India’s gain? Why Nifty bears betting on Kospi crash may get disappointed
Trillion‑Dollar AI Capex, Thin Monetisation
Jefferies underscores the macro scale of the AI build‑out, estimating that the four major US hyperscalers will spend about US$700bn on capex this year and more than US$800bn next year, rising to over US$1tn in 2027 when Oracle, Anthropic, OpenAI and neo‑cloud players are included. This US$1tn figure equates to roughly 3% of US GDP, around 22% of US non‑residential fixed investment and an estimated 33% of total pre‑tax profits of all US non‑financial companies, underlining Wood’s description of AI as “the mother of all cycles.”Yet, he warns that the financing and accounting of this capex arms race are increasingly stretched. The four hyperscalers have lifted projected capex to a massive 92% of projected operating cash flow, collectively issued US$169bn of bonds so far this year and accumulated US$662bn of future data‑centre lease commitments that remain off balance sheet, with total undiscounted lease obligations nearing US$969bn.Why Jefferies Is Rotating Toward India
Against that backdrop, Jefferies is deliberately tilting its Asia Pacific ex‑Japan asset‑allocation toward markets less dominated by AI momentum. In its latest GREED & fear note, the firm assigns India a 12% recommended weight versus a 10.9% benchmark weight in the MSCI AC Asia Pacific ex‑Japan index, giving India a positive mismatch of 1.1 percentage points.
Despite a correction in memory stocks, GREED & fear remains Underweight Taiwan and only Neutral on Korea, having cut Korea’s neutral weighting from 24.6% to 20.8% since late June, while maintaining exposure to smaller ASEAN markets largely “just to maintain a presence there.” Wood’s message is that markets like India, which host “cheaper ‘value’ names which have not been part of the AI trade,” are well positioned to benefit from any sustained rotation out of momentum AI names.
China: Valuation Play in the Rotation
China is the other key leg of Jefferies’ rotation. Wood states that “it is too late to sell MSCI China or indeed Hong Kong,” arguing that this is “precisely the area that should benefit from any mean reversion out of momentum AI names,” a view he credits to Jefferies’ global head of quantitative strategy, Desh Peramunetilleke.
MSCI China has de‑rated sharply to just 10.6 times 12‑month forward earnings, down from 13.9 times in October 2025 and 18.5 times in early 2021, even as the CSI 300 is up 11.9% in the first half of 2026 while MSCI China is down 14.9% in US dollar terms. Wood concedes that falling household loans and rising retail non‑performing loans are “one area of concern,” but maintains a base case that consumption is stabilising at a lower share of GDP and that China will avoid a self‑feeding negative equity cycle in residential property, leaving consumer and domestic‑demand stocks already pricing in much of the macro strain.
Bigger Than Dot‑Com – And Now Rotating
To frame the AI cycle historically, Jefferies notes that US investment in information‑processing equipment and software has climbed to 4.88% of nominal GDP in 1Q26, surpassing the 4.46% peak reached at the height of the dot‑com boom in 4Q00. Wood stresses that the earnings from this capex boom are “front‑end loaded” in favour of picks‑and‑shovels suppliers, while hyperscalers spent US$130bn on capex in 1Q26 but booked only US$41. 6bn of depreciation and amortisation, making profits look technically overstated.
With the Hyperscalers‑4 index underperforming the S&P 500 by 11% since early May and AI leaders well off their highs, Wood argues that investors can no longer ignore the risks around monetisation, financing and political pushback to data‑centre projects. “So long as the AI capex arms race continues, the beneficiaries will remain the picks and shovels trade (i.e. the people being paid for the capex, not the companies spending the money),” he concludes, as Jefferies repositions toward India, China and other Asian markets poised to gain from a long‑overdue rotation out of AI momentum.
(Disclaimer: Recommendations, suggestions, views and opinions given by the experts are their own. These do not represent the views of The Economic Times)
