Christopher Wood, global head of equity strategy at Jefferies, has signaled a strategic shift in his "GREED & fear" newsletter, citing "AI fatigue" as investors reconsider the massive capital expenditure (capex) flowing into artificial intelligence. Wood is pivoting exposure toward value-oriented markets, specifically India and China, as he anticipates a mean reversion away from crowded hyperscaler trades.
Signs of Fatigue in the AI Trade
The AI-driven rally, which dominated global markets throughout 2023 and early 2024, is showing clear signs of cooling. Wood notes that investors are increasingly sensitive to the lack of clear monetization strategies for the massive AI investments made by major US hyperscalers.
The correction is particularly visible in the semiconductor sector. According to Wood, the Kospi index has experienced significant volatility, dropping roughly 22% from its June 19 peak. Leveraged ETFs tied to industry leaders like SK Hynix and Samsung Electronics have seen even sharper declines, shedding approximately 30% from their recent highs. Despite these pullbacks, Wood maintains a preference for "picks and shovels" hardware suppliers—such as DRAM manufacturers—over the hyperscalers themselves, arguing that the demand for computing power will persist even if the cost of AI tokens collapses.
The Scale of the AI Capex Arms Race
Jefferies estimates that the four primary US hyperscalers—Alphabet, Amazon, Meta, and Microsoft—are on track to spend approximately US$700bn on capex this year. That figure is expected to climb beyond US$800bn next year and exceed US$1tn by 2027 when accounting for broader industry participation from firms like Oracle, OpenAI, and Anthropic.
This expenditure represents a significant portion of the broader US economy. Jefferies analysts point out that the US$1tn annual spend equals roughly 3% of US GDP and accounts for nearly 22% of total US non-residential fixed investment. Wood describes the situation as the "mother of all cycles," but cautions that the financing behind this growth is becoming stretched. The four largest hyperscalers have seen their projected capex rise to 92% of their operating cash flow, while relying on significant bond issuance and off-balance-sheet data center lease commitments, which Jefferies estimates are nearing US$969bn.
Strategic Rotation Toward India and China
In response to these risks, Jefferies is reallocating its Asia Pacific ex-Japan portfolio to favor markets with lower exposure to AI momentum.
- India: Jefferies has set its recommended weight for India at 12%, a 1.1 percentage point increase over the MSCI AC Asia Pacific ex-Japan benchmark. The firm views Indian equities as a source of "value" names that have remained largely insulated from the speculative AI frenzy.
- China: Wood argues that it is too late to exit Chinese and Hong Kong markets, suggesting they are primary candidates for mean reversion as capital flows out of AI-heavy sectors. MSCI China currently trades at a forward earnings multiple of 10.6, a significant discount from its 18.5 multiple in early 2021. While Wood acknowledges concerns regarding household debt and retail non-performing loans, he maintains that domestic demand and consumption stocks have already priced in much of the current macro-economic strain.
Historical Context and Future Outlook
To illustrate the extremity of the current cycle, Jefferies notes that US investment in information-processing equipment and software reached 4.88% of nominal GDP in the first quarter of 2026, eclipsing the 4.46% peak recorded during the height of the dot-com boom in the fourth quarter of 2000.
Wood emphasizes that the benefits of this cycle are currently concentrated in the supply chain. He concludes that as long as the AI arms race continues, the most reliable returns will likely be found among the "picks and shovels" providers, rather than the companies absorbing the immense costs of the infrastructure build-out. Investors are now forced to weigh the long-term potential of AI against the immediate risks of thin monetization and rising political pushback against large-scale data center projects.
Keep reading