Equities resilient despite semiconductor swoon
The summer semiconductor swoon has been the defining feature of equity markets over the last two months as leadership has rotated sharply away from the tech darlings of the spring towards defensive sectors such as healthcare and consumer staples. Last week, despite its volatile nature, provided a useful case study for the current environment, highlighting several important themes: market breadth, earnings, Chinese competition and hyperscaler capex.
Market breadth: It hasn’t been unusual in recent months for most companies to gain on days where the index has fallen. Indeed, since the end of May the US market has delivered a return of -1.2%, yet more than 60% of companies have generated positive returns. This broader pattern was evident again last week as semiconductor stocks were the worst performing major industry globally while much of the broader market, particularly consumer sectors and financials, performed well. Interest rate-sensitive and cyclical sectors underperformed as bond yields rose on the back of the Fed’s reluctance to hike rates.
Earnings: About a third of the way through the Q2 reporting season, the US market remains on track for a ninth consecutive quarter of double-digit earnings growth. Corporate fundamentals, particularly in the US, remain exceptionally strong. Year-on-year earnings growth is currently running at 37%, admittedly flattered by mark-to-market gains (reported earnings looking stronger due to assets or investments rising in value on paper) driven by the strong equity market performance through the spring, but analyst expectations for Q3 and Q4 earnings are also rising. Earnings growth remains broad based, with small and mid-sized companies delivering solid results. Strong earnings within technology have largely failed to halt the rotation out of the sector as investors remain focused on the scale of investment required for the AI buildout and the potential returns that investment may ultimately generate, particularly given growing competition from China.
Chinese competition: The infamous DeepSeek question remains unanswered: how much of the AI opportunity will ultimately be captured by China? Western companies are already using lower-cost Chinese AI models, while competitive Chinese releases continue to put pressure on western providers, contributing to a rapid pace of innovation and falling assumption costs. Similar concerns are emerging across the semiconductor industry. Investors increasingly fear that Chinese competitors could cap industry profitability by expanding supply and driving down prices. Apple has reportedly been lobbying US policymakers to allow Chinese-made chips to be used in products sold outside the US, highlighting how rapidly Chinese manufacturers are advancing across the semiconductor value chain.
Hyperscaler capex: Data centre projects have been under increasing scrutiny during the summer swoon as they continue to deploy capital at an extraordinary pace. In aggregate, free cash flow has turned negative, forcing increased reliance on debt markets to fund the AI buildout. This surge in bond issuance is now beginning to show signs of investor fatigue. There are also growing concerns around circular financing arrangements, with NVIDIA providing financing to OpenAI to fund chip purchases, a significant proportion of which would ultimately flow back to NVIDIA. Taken together, these developments have raised questions about both the economics and sustainability of the current AI investment cycle.
The recent weakness in technology therefore appears less a reflection of deteriorating fundamentals (semiconductors remain strong with revenues, earnings and profit margins at record highs) and more a reassessment of the durability and economics of the AI investment cycle. Economic activity is robust, earnings growth remains broad based, and market leadership has widened beyond a narrow group of mega-cap technology companies. This shift is best reflected in the broadening trend across equity markets.
While concerns around the economics of the AI buildout are undoubtedly real, so too is the scale of capital flowing through the supply chain. The key question is not whether demand for AI infrastructure is substantial, but whether that demand proves durable enough to justify the extraordinary level of investment taking place today. The answer will depend on whether AI companies can meet their revenue forecasts and honour their compute spending commitments to the hyperscalers that are funding the buildout.