AI and the Material World (May 2026)

Theme and Background

This chapter discusses the extreme dominance of the AI capital cycle over the U.S. economy and global equity markets, as well as the investment risks arising from this excessive concentration. The report notes that AI-related capital expenditure now accounts for over 40% of U.S. GDP growth, while non-AI capital expenditure has declined in absolute year-on-year terms. Over the same period, weak U.S. consumer confidence, gasoline prices approaching $5 per gallon, and sticky interest rates have made the AI theme "too big to fail" at the economic and fiscal revenue levels. The author argues that capital is flowing from all non-AI sectors into AI trades, creating a crowding-out effect.

Core Thesis

The author's core investment thesis: investors should be wary of the risk of an AI-driven drawdown, rather than continuing to chase excess returns from the AI theme. Contrarian judgments include: although the AI theme is already viewed by the market as a "certainty" asset, the collapse in the cost of capital behind it is undermining the sector's capital cycle, while passive investing and over-allocation to AI in private markets will exacerbate concentration risk. For many "picks-and-shovels" stocks in the supply chain (e.g., semiconductor foundries, electronic component manufacturing), share prices have already far exceeded what effective governance can protect for shareholders, and no longer offer a margin of safety.

Key Arguments and Data

U.S. real non-residential private fixed investment trends show AI-related investment rising rapidly from a 2023 baseline of 100 to approximately 140 by 2026, while other non-residential investment declines to around 98

• AI capital expenditure as a share of GDP growth: Over 40% (Source: U.S. Bureau of Economic Analysis, Jefferies).
• Non-AI capital expenditure (absolute): Declining year-on-year over the same period.
• Macroeconomic environment: Weak consumer confidence, gasoline prices near $5/gallon, sticky interest rates.
• Cost of capital: Sharply lower for AI-focused companies; relatively higher for non-AI companies, creating a "crowding-out effect."
• MSCI Emerging Markets Index concentration: The weight of South Korea and Taiwan has been rising since 2010; the weight of semiconductors, technology hardware, and equipment has also risen in tandem.
• Upcoming AI IPOs: IPOs by OpenAI, Anthropic, SpaceX, etc., are expected to be rapidly included in global indices, further increasing the weight of AI themes in passive allocations.

The combined weight of South Korea and Taiwan in the MSCI Emerging Markets Index has risen from approximately 25% in 2010 to approximately 45% in 2025

The weight of semiconductors, technology hardware, and equipment in the MSCI Emerging Markets Index has risen from approximately 10% in 2010 to approximately 38% in 2025

Companies/Assets Covered

The price-to-book ratio of Asian tech stocks has risen from approximately 2x in 2016 to approximately 6.3x in 2025, representing a significant valuation expansion

• TSMC: Largest position in the portfolio, held for over ten years; management is respected, but the share price is considered to have moved well beyond a safe range.
• Delta Electronics, Mediatek: Bought at the pandemic bottom in 2020; management is also respected; however, the author judges that their share prices have also exceeded what effective governance can protect for shareholders.
• OpenAI, Anthropic, SpaceX: Upcoming IPOs, expected to be rapidly added to indices and become new concentrated targets for passive capital.
• Chinese internet companies, Indian companies (multiple sectors): Persistently underweighted due to the "quality trap"—technology penetration lowers barriers to entry for branded consumer goods.
• Chinese real estate, basic materials, industrials, energy, telecoms: Account for over one-third of the portfolio, with a focus on asset-heavy sectors and areas that do not incentivize new capacity.
• High-yield financial businesses (industry consolidation type): Account for 25% of the portfolio, providing immediate cash flow.
• More than half of the portfolio is in Asia (excluding tech-intensive Asian markets), with an overall underweight to "themes and dreams" and a heavy allocation to "real cash flows."

Investment Implications

For investors, the author recommends broadly underweighting the obvious valuation-bubble picks-and-shovels stocks in the AI supply chain, and rotating toward assets that can generate returns through dividends and modest reinvestment, without relying on high-growth assumptions. Key areas include: Chinese real estate, basic materials, industrials, energy, telecoms, and other fixed-asset-intensive industries, as well as high-yield financial businesses under industry consolidation. The author believes that opportunities exist in non-AI sectors (i.e., "everything else"), even though this means having to bear short-term relative performance pressure. For passive allocators, the increasing concentration on AI themes warrants caution regarding the amplifying effect on portfolio volatility.