Jefferies' global head of equity strategy argues that hyperscaler data center spending could end in "massive capital destruction" if the market shifts toward cheaper open Chinese models while token prices keep falling. Here are the figures behind the claim, which parts are opinion, and what it means for a company buying AI.
Chris Wood, Jefferies' global head of equity strategy, published his analysis in the GREED & fear newsletter (October 9, 2026 edition, according to Crypto Briefing), and The Register picked it up on October 13. His argument: capital spending on data centers by U.S. hyperscalers could end in "massive capital destruction" if the market moves toward cheaper open Chinese models while per-token prices keep falling. It is an analyst's opinion backed by data, not a verified prediction.
Models such as Moonshot AI's Kimi K3 and Zhipu AI's GLM-5.2 cost, according to the analysis, about a quarter as much per token as some U.S. alternatives. Jefferies also notes that the sector is already the largest issuer of investment-grade debt in the U.S.
Partial measurement. OpenRouter is an aggregator; it reflects usage from those who route through it, not the whole market, and cheap tokens don't equal revenue or delivered value. Concentration, not bankruptcy. Half the backlog depending on two labs that, according to Jefferies, still lose money is a counterparty risk, not a crisis that has already happened. Accounting gains. The fact that the profit increase comes from non-operating items suggests part of the improvement is investment gains rather than operations; that alone doesn't prove a problem either. Wood himself suggests watching three signals: changes in capex guidance, the trend in token volume and price, and reliance on debt markets.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel