Chinese artificial intelligence (AI) models could trigger 'mass capital destruction' in United States (US) markets, warned Christopher Wood, Global Head of Equity Strategy at Jefferies, in his weekly investor note GREED & fear.
AI Investment Strategy
According to Wood, the most sensible approach to investing in the AI sector is a 'diggers and shovels' strategy—that is, betting on companies that profit or have a high probability of directly benefiting from hyperscalers' spending.
He noted that the key factor has always been time, specifically the point when the market begins to worry about the return on investments made in AI. Wood would like to know how much the monthly revenue growth of Anthropic and OpenAI slowed down in June, as corporations reduced the practice of encouraging employees to experiment with AI models regardless of cost.
Volatility in Asian Markets
Tuesday was marked by a sharp drop in South Korea's Kospi index by almost 11 percent, as investors began to show concern over the AI bubble. Trading was temporarily suspended during intraday sessions when the index reached its lowest level since April, closing 10.8 percent lower at 6023.66. Major losers included Samsung Electronics (down 13.4 percent) and SK Hynix (down 14.7 percent).
Jotivardhan Jaipuria, founder and CEO of Valentis Advisors, stated that the rise in AI-related stocks happened too quickly and too early, and now investors are questioning companies' capital expenditure plans and Return on Capital Employed (ROCE).
Growing Popularity of Chinese LLMs
Meanwhile, Wood reported that Chinese large language models (LLMs) are gaining popularity. According to his data, leading Chinese AI models processed 36.39 trillion tokens on the global aggregation platform OpenRouter in the week ending July 19, significantly higher than the 4.37 trillion tokens processed at the end of April.
This figure was compared to 7.39 trillion tokens processed by leading American models on OpenRoute for the same week. Wood emphasized the growing realization that China has become a technological equal to the US in both AI and many other fields. He added that these trends have not yet been factored into the markets, and the growing narrative around AI faces increasing credit risks.
Financing AI Spending
Wood also noted that AI capital expenditure in the US over the past year has been financed less by hyperscalers' cash flow and more by debt. Hyperscalers issued more investment-grade debt obligations than the US energy sector. They became the largest issuer, attracting $194 billion in debt since the beginning of the year compared to $55 billion in investment-grade issuance in the energy sector. This is why his base case is that US markets have already reached a historical peak relative to the global stock market. Wood advised closely monitoring the relative and absolute performance of hyperscaler stocks.
