EL7.AI
BondsMedium•16 October 2026•
4 min read

AI Debt Issuance Falls to $23 Billion in September

Key Facts

1AI debt issuance fell to $23 billion in September 2026, compared with $113 billion in June.
2Companies raised about $466 billion in AI debt during 2026 as of the report date, compared with $101 billion in the corresponding period.
3Gartner estimates that about $8.5 trillion in cumulative revenue is needed through 2030 to cover data center investment, financing costs and the assumed return.
4Gartner expects cumulative revenue attributable to AI data centers to reach no more than $3.6 trillion.
5The yield on CoreWeave bonds issued in April traded near 15% on Thursday, according to Reuters.

Global AI-related debt issuance fell to $23 billion in September 2026, according to an Axios report dated October 16 citing the Financial Times. Comparable issuance totaled $113 billion in June, before new financing declined in September. The drop came as the Bank of England warned of the risks of rising leverage in financing the sector. Companies had raised about $466 billion in AI-related debt during the year as of an Investing.com report published on October 12, citing Morgan Stanley. Borrowing in the corresponding period a year earlier totaled $101 billion, putting the monthly decline in the context of a sharp annual expansion.

In an estimate of the financial commitments facing the sector, Gartner said owners of AI data centers need cumulative revenue of $8.5 trillion through 2030. Those requirements include covering announced infrastructure capital expenditure and associated financing costs. They also include achieving a return on invested capital that the firm's model assumes will be 12%. By contrast, Gartner expects cumulative enterprise and consumer AI revenue attributable to data centers to reach no more than $3.6 trillion over the same period. The firm's estimate puts expected revenue below the level needed to cover those costs and achieve the assumed investment return.

The expected expansion in financing encompasses both debt issuance and private credit, according to estimates cited by the Bank of England in its September record. The bank cited JPMorgan analysts' forecast that AI-related capital expenditure financed through debt issuance would total about $4.1 trillion between 2026 and 2030. It also cited Morgan Stanley's estimate that private credit would finance $700 billion in data center capital expenditure between 2026 and 2028. In the sterling market, issuance by major AI infrastructure companies accounted for 47% of corporate bond issuance from the start of the year through the September record. The Bank of England highlighted that share despite the small size of those issues compared with the dollar and euro markets.

The Bank of England warned that rising leverage, opacity and circular financing arrangements could complicate the assessment of risks associated with AI financing. It said those factors "could amplify losses if expectations are disappointed," linking the scale of potential losses to whether the assumptions underpinning investments materialize. In its July record, the bank had identified the scale of future demand for computing as one of the critical assumptions in assessing the sector's debt risks. Alongside it, the bank highlighted the timely availability of electricity to bring data centers into operation. The factors it identified also included the depreciation rate of data center facilities and AI chips as part of its assessment of projects' financial risks.

In the banking sector, Morgan Stanley Chief Financial Officer Sharon Yeshaya told Reuters the bank was "highly selective" about the types of risk it takes on. Her comments came in Reuters coverage carried by Investing.com on October 14 about banks' financing and adoption of AI. Citi Chief Executive Jane Fraser offered a different perspective on concerns about the continued flow of investment into the sector. She said the bank saw no slowdown in investment across the broader AI ecosystem. She added: "There simply isn't enough computing capacity," expressing the continued need for computing despite financing concerns.

In the bond market, Reuters reported that the yield on CoreWeave bonds issued in April was trading near 15% on Thursday, according to its report published on October 15. That level reflected the pressures facing financing for heavily indebted computing companies. On new financing, Investing.com cited Bloomberg on October 2 as saying that the banking consortium linked to Broadcom had begun assembling a $60 billion package. The financing targets Anthropic and other AI companies, according to the report. The test investors are watching, according to Reuters coverage dated October 15, is Anthropic's initial public offering and whether its revenue growth can ease concerns about accelerating AI spending.