Global companies raised $23 billion in AI-related debt in September 2026, according to Morgan Stanley data covering public bonds and private placements. That was less than half August’s amount and nearly 80% below June’s record $113 billion. The slowdown points to more scrutiny of AI financing, but does not by itself show that investment has stopped or that credit markets are in broad distress.
2
5
18
September’s fall came after June’s record
AI-related borrowing had declined steadily from its June high, when companies raised a record $113 billion to fund infrastructure such as data centers and chips. September’s $23 billion was less than half the previous month’s total; the cited reports do not give an exact August figure.
2
5
One secondary report puts AI-linked borrowing at about $466 billion from January through September 2026, compared with $101 billion over the same period in 2025. That would mean substantially higher year-to-date borrowing despite September’s slowdown. The figures are not independently confirmed by the Morgan Stanley data excerpts available here, so treat the comparison as provisional.
9
A separate report says Morgan Stanley forecast nearly $570 billion in global AI-related debt issuance for 2026. That is a full-year forecast, not a confirmed year-to-date total.
1
4
US investment-grade issuance is a narrower measure
In the US investment-grade market, blue-chip technology companies borrowed about $306 billion between January and August, but AI-related bond issuance stopped in September, according to reporting on the Morgan Stanley data. The $306 billion figure covers broader borrowing by those companies; it is not an AI-only total.
2
5
What the slowdown says about investor appetite
Investor concern about whether large AI investments will generate adequate returns is one reported pressure on financing. The Bank for International Settlements also warned of increasing caution about future AI profitability as leverage at major US technology companies rose.
18
That caution is more evident among riskier borrowers: Reuters reported that lenders were demanding more compensation to fund lower-rated AI companies whose future earnings remained largely unproven.
17 This points to a market where access and pricing can differ by borrower—not a uniform shutdown in financing.
The BIS assessment also found no overall signs of market stress. Taken together, the evidence supports a reading of more selective lending and investor scrutiny, rather than proof of a system-wide credit crisis.
18
The figures leave important questions open
The available sources do not establish how much earlier fundraising or political opposition to data-center projects contributed to September’s decline. Nor do they provide a verified, apples-to-apples comparison of AI-only US investment-grade issuance for January–August versus September. Those limits matter: the monthly fall is clear, but the reasons and the full year-to-date picture are less certain.
The most defensible conclusion is that AI-related borrowing slowed sharply after a record June, while investors increasingly weighed returns and borrower leverage. The slowdown is significant, but it is not evidence on its own that the AI buildout has stopped.
2
17
18