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

The global AI boom is currently fueled by the most dramatic capital expenditure (capex) cycle ever seen, but a looming structural risk could trigger a sudden market shift. Jefferies’ Global Head of Equity Strategy, Chris Wood, warns that the end of the AI trade will not come from a chip shortage or oversupply, but from a crisis of confidence regarding returns on investment.

The Threat of Malinvestment and Circular Funding

Unlike traditional semiconductor cycles that end due to inventory gluts, Wood argues that the AI trade faces a unique psychological and economic threat: "malinvestment." The primary risk is that hyperscalers and AI laboratories may fail to generate adequate returns on the massive amounts of capital they are deploying.

Wood highlights a potentially fragile feedback loop within the ecosystem. He points to structures where companies like Nvidia may finance AI labs like OpenAI, which in turn use that capital to purchase more Nvidia chips. While this circularity drives short-term growth, it creates a house of cards that could collapse if investors begin to doubt the long-term monetization capabilities of the AI stack.

Massive Capex and the Concentration of Wealth

The scale of current investment is unprecedented. TSMC has raised its 2026 capex guidance to approximately $56 billion, up from $41 billion last year, with some projections suggesting $65–$70 billion by 2027. This surge has turned Taiwan into a macro powerhouse, with real GDP growth hitting 14.55% year-on-year in Q1 2026.

AI-related demand is now so concentrated that it is estimated to account for 31% of TSMC’s revenues in 2026. This level of concentration underscores how much the global economy is currently betting on a single technological vertical.

The Commoditisation of AI Models

Another significant pressure point is the rapid commoditisation of Large Language Models (LLMs). As high-quality models become cheaper, the premium pricing power of Western AI providers is being challenged.

Wood notes that Chinese models are gaining significant traction. On the OpenRouter platform, top Chinese models processed 21.37 trillion tokens in late June, a massive jump from 4.37 trillion in April, significantly outperforming the 5.76 trillion tokens processed by top US models. This shift suggests that as the "cost per token" falls, the economics of maintaining expensive, proprietary models will face intense scrutiny.

Shifting Focus to Memory and Hardware

Despite these warnings, Wood is not predicting an immediate crash. Instead, he is re-positioning portfolios toward the "picks and shovels" of the industry—specifically memory and hardware.

He cites the Jevons Paradox, where increased efficiency leads to higher total consumption. As computing becomes more efficient, the demand for DRAM and NAND memory increases. Companies like Micron have already secured long-term strategic agreements for a significant portion of their volume, giving memory makers substantial leverage and pricing power that traditional chipmakers often lacked during previous downturns.

Key Takeaways

• The Primary Risk: The AI trade is most vulnerable to "malinvestment" concerns—the realization that massive capex isn't yielding sufficient profits.
• Commoditisation Pressure: Rapidly improving and cheaper AI models, particularly from China, are threatening the high-margin economics of leading Western AI firms.
• The Memory Hedge: While software and model providers face margin pressure, hardware and memory manufacturers (like SK Hynix and Samsung) remain the structural beneficiaries of the build-out.