Bank of America warns credit markets are questioning Magnificent Seven AI spending
Investors are increasingly scrutinising the economics of the AI infrastructure trade, with credit markets flashing caution despite strong equity performance among the sector’s largest players.
Bank of America strategist Michael Hartnett has identified a growing divergence between equity resilience and credit market caution regarding the artificial intelligence capital expenditure plans of the 'Magnificent Seven' technology firms. Hartnett points to rising U.S. investment-grade tech credit spreads and elevated credit-default swaps for Oracle as evidence that investors are questioning the economics of the AI infrastructure trade. The concern is driven by cheaper Chinese AI models and hardware, which challenge the assumption that superior AI requires expensive U.S. computing infrastructure.
Credit markets are showing increased caution, with U.S. investment-grade tech credit spreads rising and Oracle’s five-year credit-default swaps trading around 200 basis points, compared to nearly 53 basis points for a broader investment-grade CDS index. This tightening reflects a broader unease about the cost structure of the AI buildout, particularly as Chinese developer DeepSeek demonstrates that highly capable models can be developed using less-advanced hardware and significantly lower training costs.
The emergence of DeepSeek’s V3 and V4-Flash models has exposed potential flaws in the U.S. AI cost model. Its V3 model was developed using less-advanced Nvidia H800 chips with training costs under $6 million, and its new V4-Flash model costs just $0.14 per million input tokens. These developments challenge the prevailing narrative that better AI performance necessitates ever-larger amounts of expensive computing infrastructure, raising questions about the return on billions of dollars in annual capital spending by U.S. hyperscalers.
Despite these headwinds, major technology companies have not reduced their spending forecasts. Big Tech firms, including Microsoft, Meta, Oracle, Amazon, and Alphabet, have collectively written down nearly $1.09 trillion in future lease payments, much of it tied to data centres. Amazon recently raised its 2026 capital expenditure forecast to $220 billion, citing healthy AWS demand and ongoing capacity constraints, suggesting that management teams remain committed to the long-term growth story.
Hartnett suggests that the key indicator for the sector's health is the performance of the Roundhill Magnificent Seven ETF (MAGS), which holds around $70 billion in assets. The ETF has gained 4.82% over the past week and 4.38% over one month, outperforming the S&P 500, but has lagged the broader market over a six-month period with a 9.55% gain versus 12%. The fund carries a higher-than-average risk profile, with nearly 96% of its assets in its top 10 holdings and an annualised volatility of 22.2%, underscoring the concentrated exposure and potential for sharp price swings in the AI trade.
