Artificial intelligence has driven the entirety of US economic growth's positive story in recent years. AI stocks now constitute approximately 33% of S&P 500 market cap. AI investment constituted 90%+ of US GDP growth in H1 2025. The top five hyperscalers — Microsoft, Amazon, Alphabet, Meta, and Oracle — are spending $602 billion in capital expenditure in 2026, a 36% year-over-year increase, with approximately 75% of that capex directed at AI infrastructure: data centers, chips, power, cooling.
The arithmetic of this capital allocation cycle has a structural problem. AI assets — GPUs, server infrastructure, data center equipment — depreciate at approximately 20% per year due to rapid obsolescence as successive model generations require newer hardware. At $602 billion in annual capex, the implied annual depreciation is approximately $120 billion in the first year and rising. In 2025, the combined profits of the top hyperscalers were approximately $100 billion. Their annual depreciation on AI assets already exceeds their combined profits. The capex is being funded by debt: hyperscalers raised $108 billion in debt in 2025, with $1.5 trillion projected over coming years. Technology equipment and software investment has reached 4.4% of GDP — near the peak of the dot-com bubble.
The most acute case study is CoreWeave, a GPU cloud provider that has become the structural spine of much of the AI buildout. CoreWeave carries $24.5 billion in total debt, with $7.5 billion in interest payments through the end of 2026. Sixty-two percent of its revenue comes from a single customer — Microsoft. Its five-year credit default swap spreads are above 640 basis points. The market-implied five-year default probability is approximately 42%. Oracle, a more established technology company that has pivoted heavily into AI infrastructure, has an 86% ratio of capex to sales and has been raising $45–50 billion in debt and equity. Its CDS spread has risen above 125 basis points — levels associated with distressed credits in 2009. Barclays projects that Meta's free cash flow will drop approximately 90% in 2026 as its AI capex commitments overwhelm operating cash generation.
The AI bubble is not just a financial market phenomenon. It is the primary transmission mechanism between a stock market correction and the real economy, because the wealth effect — the tendency of households to spend more when their investment portfolios rise — is concentrated in the top 10% of earners who hold the majority of equity. When AI stocks deflate, that spending floor disappears. The top 10% generate approximately 50% of all consumer spending. Their retrenchment alone is sufficient to tip a slowing economy into outright contraction.