Morgan Stanley: September AI Bond Issuance Drop Is a Pause, Q4 Rebound Expected
nashnova research
Global AI-related bond issuance dropped to roughly $23 billion in September — a near-year low — with U.S. investment-grade AI issuance hitting zero; Morgan Stanley calls it a pause, not a retreat, and expects a Q4 rebound.
How much AI debt has been issued this year, and why did September go cold?
Year-to-date global AI bond issuance reached $466 billion through September, more than double last year's full-year total of $216 billion. June peaked at $113 billion in a single month; each month since has been lower.
September came in at roughly $23 billion — the second-lowest month of 2025. U.S. investment-grade AI issuance was exactly zero.
This means → the first-half pace was itself abnormally high. September's drop is not a collapse but a breather after front-loaded supply.
What caused the cooldown?
Front-loaded issuance: Google, Amazon, Meta, and Microsoft issued roughly $132 billion in investment-grade debt this year — about 25 times last year's figure — with most concentrated in H1. In plain terms = they already raised what they needed.
Construction bottlenecks: new data centers face permit delays, power-supply shortages, and political resistance. Morgan Stanley sums it up as "the three Ps" — Permits, Power, and Politics — now shifting from latent risk to binding constraint.
Higher rates squeeze projects: yields have risen sharply, forcing project-level deals to be renegotiated. This means → top-rated hyperscalers barely feel it, but lower-rated issuers and project financings are far more rate-sensitive and have pulled back.
What drives the expected Q4 rebound?
Morgan Stanley expects hyperscalers to return to the U.S. investment-grade market while continuing to tap non-dollar markets. Non-dollar hyperscaler issuance has already reached roughly $72 billion this year — nearly a third of the global total — with currencies expanding from dollars and euros to include CAD, GBP, CHF, AUD, and JPY.
The key catalyst is 2027 capex expectations. Morgan Stanley forecasts the six largest hyperscalers will spend a combined ~$1.4 trillion in 2027, well above the consensus range of $1.0–1.1 trillion. Q3 earnings reports — expected from October — could trigger another round of capex upgrades and directly drive new funding needs.
This means → the rebound is not about sentiment improving; it is about real construction plans that still need financing.
Why aren't these giants worried about high rates?
Morgan Stanley sees high-quality hyperscalers as "largely insensitive" to higher rates — their return on invested capital outlook is strong, and debt financing remains cheaper than equity.
In plain terms = the return on building data centers far exceeds borrowing costs, so they keep issuing even as rates rise.
The numbers back this up: aggregate leverage is just 1.3× (net leverage 0.4×), cash-to-debt ratio is 132%, and the median credit rating is AA-. Balance-sheet room is ample.
What is the biggest risk into year-end?
Morgan Stanley stresses that for the investment-grade credit market, "the macro environment — not supply glut — is the biggest test before year-end." IG bond total returns are down 3% year-to-date and 4% for the quarter, a drawdown that could trigger large-scale mutual-fund and ETF redemptions.
The bank nonetheless believes credit markets can absorb higher yields: nominal growth is running above 6%, corporate earnings growth is even stronger, and the rate rise has fundamental support. Its economists expect the Fed to hike only twice more — more dovish than the nearly four hikes the market is pricing in.
This means → the real risk is not AI bonds flooding the market, but a broader macro deterioration dragging down all credit assets.
Which types of AI debt should investors watch?
Morgan Stanley reiterates its preference for secured assets, which account for roughly 30% of AI-related debt. Stabilized ABS (asset-backed securities) and CMBS (commercial mortgage-backed securities) face less construction risk.
Chip financing — deals collateralized by the chips themselves — allows faster cash-flow generation and amortization, but publicly traded products remain limited.
This reflects Morgan Stanley's stance of "bullish on the sector, selective on assets" — much of the hyperscaler spread-compression trade may already be priced in, and whether Q3 earnings bring another capex upgrade will be the key checkpoint.
市场有风险，内容仅供研究参考，不构成投资建议。