Five AI quirks and conundrums to watch this earnings season
QQQ•AI-related stocks have driven 75% of S&P 500 returns since ChatGPT launched, while spending on data centers and other AI-related capital expenditure by the six largest US hyperscalers is estimated to exceed $7 trillion from 2025 to 2030. Investors are watching borrowing costs, power constraints and whether AI revenue can justify the investment.
1. Stocks and debt diverge
AI-related shares have mostly risen even as some bonds issued to finance the buildout have weakened. CoreWeave’s October 2031 bond has fallen more than 20% since its April launch and trades below 80 cents on the dollar; credit default swap spreads have also recently reached records for AI borrowers including CoreWeave, Oracle, Amazon, Alphabet and Meta.
2. Spending and constraints
Analysts disagree sharply on CoreWeave’s outlook: HSBC forecast average returns through 2030 below its borrowing cost and set a $39 target, while Goldman Sachs upgraded the stock to “buy” and set a $106 target. Elsewhere, falling token prices have slowed spending growth even as usage rises, and Morgan Stanley analysts estimate US data center developers will face a 30% to 40% power shortfall through 2028.
3. A high revenue bar
McKinsey and S&P Global estimate the six largest US hyperscalers will spend more than $7 trillion on data centers and other AI-related capital expenditure between 2025 and 2030. Columbia University’s Stijn Van Nieuwerburgh estimates financing the AI buildout from last year to 2032 will cost $10.3 trillion, and says annual revenue of about $3.55 trillion by 2032 would be needed for a 10% return.
