Goldman Strategist Quantifies AI-Driven Earnings Bubble Risk
Capex surges and bond market turbulence signal capital competition between AI and governments.
1From anecdotal to data-driven
Six weeks after first flagging potential earnings bubble risks in tech stocks, Goldman's Peter Oppenheimer now presents hard data. AA-rated tech firms saw 65% YoY capex growth in Q2 - the tenth consecutive quarter above 35% growth. US convertible bond issuance hit $135B YTD, with AI-related firms accounting for 44%. These metrics substantiate his August thesis about capital competition pressures.
2The capital collision mechanism
Oppenheimer identifies a precise mechanism: AI infrastructure spending and government borrowing are directly competing for the same capital pool. Private firms raise debt for AI data centers while governments borrow more for infrastructure and defense. Simultaneously, energy-driven inflation pushes rates higher. This triple pressure is elevating global capital costs - the report's central thesis, now backed by credit issuance forecasts raised to $2.3T for 2023.
3Historical stress test applied
The analysis draws parallels to 2008 banks and 1990s dot-com bubbles where earnings (not valuations) collapsed first. Unlike credit-fueled bank fragility, tech currently shows robust profits and strong balance sheets. S&P 500 interest coverage ratios sit at 99th percentile historically, suggesting less systemic leverage risk. However, any profit slowdown amid high capital costs could pressure equity prices across the AI ecosystem.
4Apollo echoes savings shortage
Apollo's Torsten Slok independently identified similar dynamics five days earlier, declaring a regime shift from 'savings glut' to 'savings shortage.' His analysis shows hyperscalers' long-dated bond spreads widening, with 2026 paper trading below issue prices. This reflects capital competing for projects by demanding higher returns - particularly at the long end of the curve where data centers and infrastructure dominate funding needs.
5AI demand vs safety concerns
Recent market moves illustrate the tension. When Anthropic's CEO called for slower AI development on safety grounds, chipmakers fell 5-6% while hyperscalers rose. This reflects the asymmetry that cloud providers can pause capacity expansion while chip suppliers lack such flexibility. Nvidia maintains its $3-4T AI TAM forecast by 2030, but the episode shows how sentiment shifts could disrupt capex flows.
6Capex boom beneficiaries at risk
The report highlights vulnerability across the AI supply chain. While hyperscalers like Microsoft plan to triple data center capacity, 'pick and shovel' suppliers face greater earnings risk if capex slows. Goldman's analysis suggests the current capex boom - with AA tech firms sustaining 35%+ growth for 10 quarters - may face pressure from competing government borrowing needs and higher policy rates.
7Market divergence signals
The equal-weighted S&P 500 has outperformed its cap-weighted counterpart by the widest margin since 2009 - a sign investors question earnings concentration. Meta fell 10% despite beating estimates while Microsoft jumped 17% on strong earnings. Such dispersion suggests growing skepticism about whether a few tech giants can sustain market-leading earnings growth amid capital competition.
8Trigger in bond turbulence
Oppenheimer points to September's bond market moves as early evidence of his thesis playing out. With 30-year German/Japanese yields near zero in 2022, the sudden rise in long-term rates reflects capital competition concentrated at the curve's long end. This dynamic - where data centers and infrastructure create long-duration capital claims - may continue pressuring tech earnings as government borrowing accelerates.
Original article: Fortune | FORTUNE — Goldman Strategist Quantifies AI-Driven Earnings Bubble Risk