Chris Wood, Global Head of Equity Strategy at Jefferies, warns that the current capital expenditure plans of US hyperscalers may be unsustainable. He characterizes the massive investment in data centers and hardware as potential malinvestment, noting that the industry is increasingly relying on debt financing rather than cash reserves.
The strategist points to the competitive advantage of Chinese open-source AI models, which currently offer significantly lower costs per token. During the week ending July 19, 2026, Chinese models processed 36.39 trillion tokens, compared to 7.39 trillion tokens handled by top US models. Models such as Moonshot AI’s Kimi K3 and Zhipu AI’s GLM-5.2 are priced at roughly one-quarter of US alternatives.
Wood suggests that the combination of falling token prices and the market share gains of Chinese models threatens the profitability of American AI providers. He advises investors to monitor future capital expenditure guidance, which is projected to reach $695 billion in 2026 and $870 billion in 2027, as any reductions would indicate a shift in market sentiment.
The report highlights that a potential pullback in infrastructure spending would impact the entire hardware supply chain. While debt-heavy growth strategies face increased scrutiny, Wood suggests that semiconductor firms may remain a more stable investment option, advocating for a focus on infrastructure suppliers amid growing economic uncertainty.
