Chris Wood Warns: Malinvestment Risk Could Trigger the End of AI Trade

The relentless surge in Artificial Intelligence spending is facing a critical warning from one of Wall Street's most seasoned strategists. Chris Wood, Jefferies’ Global Head of Equity Strategy, suggests that the AI boom won't end due to a lack of demand, but rather a looming crisis of profitability.

The Threat of Malinvestment Over Oversupply

Unlike traditional semiconductor cycles that typically collapse due to sudden inventory gluts or supply shocks, Wood argues that the AI trade faces a unique structural risk: malinvestment. He believes the "endgame" for the AI boom will be triggered when hyperscalers and leading AI labs realize they cannot generate adequate returns on the massive capital expenditures (capex) they are undertaking.

Wood highlights a potentially precarious feedback loop where companies like Nvidia provide financing to entities like OpenAI, which then use that capital to purchase more Nvidia chips. While this cycle fuels rapid growth, it relies heavily on optimistic monetization assumptions. If investors begin to doubt the long-term earnings visibility of the AI stack, this circular funding model could unwind sharply.

Massive Capex and the Rise of Memory Giants

The scale of current investment is unprecedented. Wood describes the ongoing build-out as "the most dramatic capex cycle" he has ever witnessed. A primary example is TSMC, which has lifted its 2026 capex guidance to approximately $56 billion, with projections from Fubon Research suggesting it could reach $65–$70 billion by 2027. This surge has significantly boosted Taiwan's economy, with real GDP growth hitting 14.55% year-on-year in Q1 2026.

Interestingly, while the software and model layers face risks, the "picks and shovels" of the industry—specifically memory suppliers—are seeing structural benefits. Wood points to the Jevons Paradox, where increased efficiency leads to even higher total consumption. This has turned DRAM and memory from peripheral components into core engines of AI productivity. Major players like Micron are already securing long-term leverage, with 16 strategic agreements covering 20% of their DRAM volume and a third of their NAND volume.

The Commoditization of AI Models

A significant headwind for Western AI providers is the rapid commoditization of Large Language Models (LLMs). Wood notes that the cost of high-performing models is plummeting, particularly with the rise of efficient Chinese models.

Data from OpenRouter reveals a massive shift: in late June, top Chinese AI models processed 21.37 trillion tokens, a significant jump from 4.37 trillion in April. This volume far outpaced the 5.76 trillion tokens processed by top US models. As models like Z.ai’s GLM-5.2 offer performance nearly equal to premium Western providers at one-quarter of the cost, the pressure on the economics of "premium" AI services continues to mount.

Key Takeaways

  • The Core Risk: The AI trade is more likely to end due to a realization of "malinvestment" and poor returns on capital rather than a traditional chip oversupply.
  • Memory Resilience: While AI software faces commoditization, memory manufacturers (DRAM/NAND) are gaining structural pricing power and long-term contract stability.
  • The Feedback Loop: Investors should monitor the circular relationship between chipmakers financing AI labs, as this cycle is highly sensitive to shifts in investor sentiment.