Global AI-related debt issuance fell to $23 billion in September, a sharp decline from $113 billion in June, according to the Financial Times. This drop reflects a growing disconnect between the massive capital expenditures required for AI infrastructure and the economic returns currently being generated. Wall Street financiers are increasingly vocal about the sustainability of the sector's debt load, noting that the gap between costs and revenue is widening.
Gartner projects that AI providers will spend $7 trillion on infrastructure through 2030, yet expects related revenue to total no more than $3.6 trillion over the same period. If companies maintain their current investment pace, more than $4 trillion in spending could be subject to write-downs by the end of 2032. The research firm also predicts that more than 80% of specialized AI cloud providers, or neoclouds, will cease operating independently by 2030.
CoreWeave, one of the largest neoclouds with close ties to Nvidia, has experienced a multi-day selloff in its stock and debt. Yields on its debt pushed above 15%, and its stock price dropped 15.5% before rebounding after Nvidia CEO Jensen Huang spoke in support of the company. The episode highlights the fragility of financing structures underpinning the AI buildout.
Capital markets are showing signs of fatigue as a third of AI capital expenditure is now funded by debt rather than cash. Jamie Dimon, CEO of JPMorgan, confirmed that his bank is tightening credit to avoid concentration risk. Other infrastructure projects are also facing strain; approximately $18 billion in bank loans for an Oracle data center campus has slid into stressed territory. Additionally, Firmus, an Nvidia-backed AI cloud provider, shelved an initial public offering that had sought a valuation north of $30 billion.
The mismatch between immediate spending and delayed returns stems from the time required for businesses to integrate AI into their operations. Ioannis Antonoglou, co-founder of Reflection AI, told Semafor that companies may need a decade or two decades to fundamentally redesign their operations around the technology. This timeline creates a period where expensive infrastructure sits idle or underutilized relative to its cost base.
Some companies are mitigating costs by adopting open-weight models that are less compute-intensive. Anthropic provides a counterpoint to bearish narratives, having turned free cash flow positive in the third quarter and grown revenue by nearly 50% compared to the April-to-June period. However, even Anthropic's growth has slowed, reflecting broader patterns in the market. The increasing entanglement of the stock market and general economy with the AI boom means that if the sector's debt structure fails, broader economic consequences could follow.