Morning Edition · Sunday, October 18, 2026Published at 1:07 AM EDT · New York
The Nasdaq Composite fell 2.1% on Friday to 26,140.03, extending a sharp weekly decline, as the cost of insuring debt of AI-linked companies against default rose sharply.
The Financial Times reported that the $10 trillion American buildout of artificial intelligence infrastructure is confronting the physical and financial costs of construction, and that parts of the expansion are slowing. The clearest evidence this week came from credit markets. A Jefferies report cited by the Economic Times found that Oracle and CoreWeave face heightened credit risk as financing costs rise.
The price of protection against default shows the shift. Semafor reported that insuring $1 million of CoreWeave bonds over five years now costs about $100,000 a year, up from under $40,000 a year ago, and that yields on its bonds topped 15%. The same contracts on Oracle (credit default swaps, which pay out if a borrower defaults) rose to 288 basis points from 260 a week earlier. Oracle shares closed down nearly 9% on Wednesday.
Alphabet, Amazon, Meta and Oracle have issued more than $228 billion in bonds this year, more than double their total for 2025. That debt load now faces the prospect of higher US interest rates following September's inflation data. On Friday the S&P 500 fell 1.1% to 7,659.44 and the Nasdaq Composite fell 2.1%.
European industrial suppliers are taking different approaches to this risk. Schneider Electric and Siemens are moving further into software, while ABB is betting on the physical equipment that data centers need.
Part of a tracked trend
AI Build-Out Meets the Cost of Credit
As borrowing costs stay high, debt-financed AI infrastructure projects keep repricing in credit markets first, so leveraged builders retrench while cash-rich platforms consolidate control of computing capacity.
What this means
Long-dated, debt-financed projects were planned for a world of cheap credit. When financing costs rise, the most leveraged builders suffer first, through refinancing costs and falling bond prices, well before demand for computing weakens. The highly indebted second-tier providers and their lenders are exposed. Hyperscalers that fund most spending from their own cash flow are less exposed, and so are equipment suppliers paid on delivery. The common view is that AI demand guarantees the returns. The more relevant precedent is the 2000-2002 telecom bust, when Global Crossing and WorldCom went bankrupt under their debts, while the fiber they laid survived and was bought cheaply by others. The desk's read is that over the coming months the stress stays concentrated in leveraged data-center developers, which will delay projects, sell assets or raise equity on worse terms, while the largest platforms keep spending. This depends on US yields staying near current levels. The read would be wrong if the Fed signals no hike and the high-yield market reopens to these borrowers at lower spreads.
What to watch
Observations to monitor, not financial advice.
Synthesized from: Financial Times · Economic Times · Financial Times Technology
Start a discussion in Townsquare.
More from this edition
Comments
1Oct 18, 5:07 AM
CoreWeave protection at about $100,000 per $1 million is roughly 1,000 basis points a year, about 3.5 times Oracle's 288, so markets price these two AI borrowers very differently.