Chris Wood’s Warning: Why Malinvestment Could End the AI Boom

The massive surge in Artificial Intelligence spending is creating one of the most dramatic capital expenditure (capex) cycles in history, but it carries a hidden danger. Jefferies’ Global Head of Equity Strategy, Chris Wood, warns that the AI trade's downfall won't be caused by a chip shortage or oversupply, but by a crisis of profitability.

The Looming Threat of Malinvestment

According to Wood’s latest "Greed & Fear" newsletter, the primary risk to the AI ecosystem is "malinvestment." He argues that the trade will likely face a painful pause or end when the market realizes that hyperscalers and leading AI labs cannot generate adequate returns on their massive investments.

A significant concern is the existence of circular funding loops. Wood points to scenarios where companies like Nvidia may finance AI labs like OpenAI, which then use that capital to purchase more Nvidia chips. While this creates a momentum-driven feedback loop, it remains highly vulnerable to investor skepticism regarding long-term earnings visibility and capital discipline.

A Massive Capex Cycle Driven by Infrastructure

The scale of investment currently being poured into AI infrastructure is unprecedented. TSMC is a prime example of this concentration; the company has lifted its 2026 capex guidance to approximately US$56 billion, up from US$41 billion last year. Further projections from Fubon Research suggest capex could reach US$65–70 billion by 2027.

This spending is fueling massive macroeconomic shifts in regions like Taiwan, where real GDP growth hit 14.55% year-on-year in Q1 2026. Currently, AI-related demand is estimated to account for 31% of TSMC’s total revenues for 2026, highlighting how deeply the global economy is now tethered to the AI build-out.

The Commoditisation of AI Models

Adding to the pressure on margins is the rapid commoditisation of Large Language Models (LLMs). As high-quality models become available at a fraction of the cost, the "moats" surrounding premium Western AI providers are shrinking.

Wood notes the rise of efficient Chinese models, such as Z.ai’s GLM-5.2, which reportedly offers performance nearly equal to top-tier US models like Anthropic but at just one-quarter of the cost per token. Data from OpenRouter shows a significant shift: in late June, top Chinese AI models processed 21.37 trillion tokens, vastly outpacing the 5.76 trillion tokens processed by leading US models. This trend suggests that the software layer of AI is becoming a low-margin commodity business.

Moving Toward the "Picks and Shovels"

Despite these risks, Wood does not predict an immediate collapse. Instead, he suggests a strategic shift toward the "picks and shovels" of the industry—specifically memory and hardware.

Unlike the software layer, DRAM and memory suppliers are gaining significant leverage. For instance, Micron has already signed 16 strategic customer agreements covering 20% of its DRAM volume and one-third of its NAND volume, often with five-year terms. This structural change allows memory makers to command pricing power, making them safer bets even if the broader AI capex cycle faces a reality check.

Key Takeaways

  • The Risk Factor: The AI trade is vulnerable to "malinvestment," where investors realize hyperscalers cannot earn sufficient returns on their massive capital expenditures.
  • Commoditisation Pressure: The rapid emergence of low-cost, high-performance Chinese AI models is commoditising the LLM market and squeezing margins for Western providers.
  • Strategic Pivot: While software margins may shrink, hardware and memory providers (like SK Hynix and Samsung) are securing long-term advantages through strategic, multi-year supply agreements.