3 hours ago
9
Jefferies strategist Chris Wood thinks America’s AI spending spree is heading somewhere uncomfortable. In the October 9, 2026 edition of his “GREED & fear” newsletter, he warned that the boom could end in massive capital destruction across the US AI sector.
His culprit is not weak demand for artificial intelligence. It is competition from inexpensive Chinese open-source models, which he expects to keep winning market share while the price of AI output slumps.
The spending that worries Jefferies
Wood is Global Head of Equity Strategy at Jefferies, and his concern starts with scale. US hyperscalers have issued combined capital expenditure guidance of approximately $695 billion for 2026.
That figure is projected to rise to $870 billion in 2027. Hyperscalers are the giant cloud operators that rent out computing power, and capex covers the data centers, chips and other hardware they buy to do it.
Wood’s verdict on much of this outlay is blunt. He calls it “malinvestment,” an economist’s term for money sunk into projects unlikely to earn back what they cost.
He also points to how the bills are being paid. In his analysis, the industry has shifted significantly toward debt financing rather than funding the buildout from cash.
That distinction carries real weight. Losing your own cash on a bad bet stings, but borrowing heavily to make that bet can create credit risks and potentially set the stage for a market correction.
China’s token surge
The competitive pressure Wood describes shows up clearly in usage data. Chinese AI models processed 36.39 trillion tokens during the week ending July 19, 2026.
Top US models handled 7.39 trillion tokens over the same week. Tokens are the small chunks of text that AI models read and generate, which makes them a rough gauge of how much the models actually get used.
Price helps explain the gap. Models such as Moonshot AI’s Kimi K3 and Zhipu AI’s GLM-5.2 cost roughly one-quarter per token of some US alternatives.
Wood connects those dots directly. Slumping token prices combined with Chinese models capturing significant portions of the market make the projected US hyperscaler capex unsustainable, in his view.
A different kind of AI race
Much of the conversation around AI has framed it as a contest of capability: who builds the smartest model first. Wood’s argument shifts attention to economics, specifically who can deliver usable AI at the lowest cost per token.
Open-source models matter here because they can be adopted widely without the same licensing arrangements as closed systems. If developers can get comparable output for a fraction of the price, the premium that justifies enormous infrastructure budgets gets harder to defend.
What this means for investors
For investors, the most immediate pressure point may be the balance sheet. If Wood is right, debt-heavy growth strategies could face increased scrutiny as market conditions turn less favorable for that approach.
Margins are the second concern. Lower-cost Chinese alternatives could dampen profitability for American AI providers, which in turn could make the tech investment landscape more volatile.
Semiconductor firms are one group flagged as potentially standing to gain in that environment, with Wood advocating for “picks and shovels” investments as a cautious approach amid increasing uncertainty.
There is also a tension worth noting. The same capex that Wood views as malinvestment is revenue for the suppliers building the infrastructure, so any pullback would ripple through the hardware supply chain rather than staying contained at the hyperscalers.
Several signals will show whether Wood’s thesis is gaining traction. The first is future capex guidance: any trimming of the approximately $695 billion figure for 2026 or the $870 billion projection for 2027 would suggest companies are hearing the warning.
The second is usage data. If Chinese models continue to process tokens at volumes like the 36.39 trillion recorded in mid-July, while per-token prices keep sliding, the pressure on US pricing power will be harder to dismiss.
The third is the financing mix. Watch whether hyperscalers keep leaning on debt markets to fund their buildouts, because that is where Wood sees the risk of capital destruction turning into credit risk.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·