CRITICAL EVENT UPDATE · AI Infrastructure × Credit × Capital Markets
AI Debt Issuance Falls From $113 Billion to $23 Billion as Capital Enters a Selection Phase
Critical Event Update | About 80% Below the June Peak × Zero U.S. IG AI Bond Issuance in September × Financing-Constrained Expansion
Morgan Stanley data show global AI-related debt issuance falling from a $113 billion June peak to $23 billion in September, a decline of roughly 80%. U.S. investment-grade markets saw no new AI-related bond issuance in September. Heavy front-loaded funding is the primary explanation, but high rates, uncertain returns, social-license resistance and execution risks are raising the selection hurdle. This is not a closed funding market; it is a credit market differentiating strong balance sheets and verifiable cash flow from distant capacity plans.
Global AI-related debt issuance in September
June issuance peak
September decline from the June peak
U.S. investment-grade AI bond issuance in September
Debt issuance cools sharply | Funding remains available | Selection and structure scrutiny rise
01 · RESEARCH BRIEF
The one-minute brief
AI-related debt issuance fell to $23 billion in September, about one-fifth of June's $113 billion peak and nearly half August's level. Large U.S. technology companies borrowed about $306 billion from January through August, but U.S. investment-grade AI bond issuance fell to zero in September.[1] Morgan Stanley says earlier front-loaded funding is the main cause; investor caution around returns, structures, power and community opposition is also rising. Combined with Firmus withdrawing its $5 billion IPO, the evidence moves Financing-Constrained AI Expansion from a company event to a cross-market equity-and-credit record.
Audio transcript
AI-related debt issuance fell from one hundred thirteen billion dollars in June to twenty-three billion in September, while U.S. investment-grade AI bonds saw no September issuance. Funding has not closed, but capital is moving from broad availability to selection. Strong balance sheets and verifiable cash flow gain an advantage over leveraged, unbuilt capacity.
Known facts and open questions
- Confirmed
- Morgan Stanley's September and June issuance totals
- Confirmed
- U.S. investment-grade AI issuance fell to zero in September
- Primary explanation
- Heavy earlier funding reduced immediate issuance needs
- Unconfirmed
- A permanent funding shutdown or systemic default cycle
Issuance
September volume was about one-fifth of the June peak
Cost
High long yields raise project-return hurdles
Structure
Bonds, private placements, SPVs and vendor finance face more scrutiny
Execution
Power, permits, social license and construction determine access
Cash flow
Revenue, FCF and debt service replace announced capacity as proof
02 · FACTS → IMPACT → VIEW
Why does this change matter?
AI financing moves from broad expansion to selection and repricing
- What is confirmed
- ACIS expected high rates to split AI infrastructure into financeable, cash-generating projects and riskier projects dependent on distant capacity narratives.
- Why it matters
- AI debt issuance fell from $113 billion in June to $23 billion in September, while U.S. IG AI issuance stopped in September and Firmus failed to complete its public-equity raise.
- ACIS view
- Financing-Constrained AI Expansion strengthens across equity and credit markets, but a capital shutdown is not confirmed. Funding remains available with greater weight on balance sheet, structure, power execution and cash flow.
03 · EVIDENCE & ANALYSIS
Evidence and analysis
01｜What Happened
Financial Times, citing Morgan Stanley data, reported global AI-related debt issuance falling from $113 billion in June to $23 billion in September; September was nearly half August's volume. U.S. investment-grade AI issuance fell to zero in September after large technology companies borrowed roughly $306 billion from January through August.[1]
02｜Why It Matters Now
AI buildout still needs large amounts of external capital while the 10-year Treasury yield is above 5% and financing, energy and social-license costs are elevated. Lower issuance shows capital providers demanding clearer returns, security, customer contracts and execution evidence.
03｜Confirmed Facts vs Uncertainty
The issuance totals and zero September U.S. IG AI issuance are confirmed. Morgan Stanley says heavy earlier funding is the primary explanation; greater investor caution is a secondary factor. It is not confirmed that the credit window is permanently shut, all AI borrowers are constrained or a default cycle has begun.
04｜Transmission Mechanism
High long yields and project risk → higher spread and structure requirements → lower issuance and longer funding cycles → capital favors cash-generating, contracted projects → leveraged developers and unbuilt capacity face pressure → equipment orders, energization, revenue and FCF diverge further.
05｜Prior View → New Evidence → Updated View
Prior: AI had entered Financing-Constrained Expansion. New: issuance is roughly 80% below the June peak and resonates with Firmus's failed public-equity raise. Updated: funding is not closed, but the AI credit cycle is shifting from Broad Availability to Selection and Repricing.
06｜Cross-Asset / Cross-Industry Read-through
Cash-rich hyperscalers and investment-grade issuers gain relative advantage; neoclouds, developers and project SPVs face tougher structures; power, cooling, networking and chip suppliers need larger discounts for customer funding and cancellation risk; near-term bond supply pressure may ease, while the future financing pipeline can still lift technology spreads and term premium.
07｜What Does NOT Change
Lower September issuance is not an AI-demand collapse and does not overturn long-run compute, power and networking demand. Cumulative 2026 AI financing remains historically high, and strong-balance-sheet issuers retain market access.
08｜Risks / Alternative Scenarios
Upside: issuance resumes at tight spreads after a digestion period. Base: capital remains available with faster project segmentation. Downside: spreads, security and covenants deteriorate and delay construction. Alternative: the September decline is almost entirely front-loading and seasonality.
09｜Next Validation
24H: new AI bond soundings, technology spreads and fund flows. 7D: pricing, order books and security for pending financings. 30D: financing completion, construction, energization, supplier orders, interest coverage and free cash flow.
04 · INVESTMENT IMPLICATIONS
Industry and asset implications
Hyperscalers
Cash flow and investment-grade access become a relative moat.
Neoclouds and developers
Leveraged, unpowered and distant capacity faces a higher hurdle.
AI supply chain
Order quality must include customer funding and cancellation risk.
Credit and rates
Near-term supply eases, but the pipeline still supports term premium.
This is evidence of stronger capital selection, not a closed AI financing market.
05 · VALIDATION & RISKS
What to watch next
24H
New deals find orders at acceptable spreads
What would weaken the view: Investors demand sharp concessions or delays
7D
Pending financings disclose transparent structures and customer cover
What would weaken the view: More deals shrink, move private or add security
30D
Funding converts into construction, power and revenue
What would weaken the view: Projects slip, orders cancel or interest coverage weakens
What would change our view?
There are two major errors: treating lower issuance as a credit freeze, or treating record year-to-date funding as proof that every project earns sustainable returns. Cost, structure and Capacity-to-Cash are the decisive tests.
06 · FAQ
Key questions
Has AI debt financing closed?
No. September issuance fell sharply, but much funding was completed earlier and strong issuers still have access.
Why is this more important than the Firmus IPO withdrawal?
It expands a company-specific equity event into evidence across the broader AI credit market.
What would confirm a real credit contraction?
Persistent deterioration in new-deal spreads, order books, security and covenants that then delays projects, cancels supplier orders and weakens cash flow.
07 · TERMS & SOURCES
Terms, sources and related research
Key terms
- AI-related Debt
- Public bonds and private debt funding chips, data centers, cloud infrastructure and related AI investment.
- Selection and Repricing
- Capital remains available but is reallocated and repriced by issuer quality, structure and cash flow.
- Capacity-to-Cash
- The chain from announced capacity through permits, power, contracts, energization, revenue and free cash flow.
[1] Financial Times｜AI borrowing slows as investors grow wary of debt binge｜2026.10.10 ↗
[2] Reuters｜Tech companies tap debt and equity to fund AI and cloud expansion｜2026.10.07 ↗
[3] Goldman Sachs Research｜How AI Debt Is Reshaping Credit Markets｜2026.08.05 ↗
This public report relies on Financial Times reporting of Morgan Stanley data, a Reuters financing-event compilation and Goldman Sachs Research. The $23 billion and $113 billion figures reflect the data provider's AI-related debt classification, not all technology debt. Lower issuance is kept separate from a closed credit window, defaults or an AI-demand collapse.