AI Debt Markets Take a Breather After Record-Breaking First Half
TREE NEWS reports: Global AI-related bond issuance slowed dramatically in September, with only about $23 billion in new deals priced — the second-lowest monthly total of the year — and zero issuance from AI borrowers in the U.S. investment-grade market. The sharp deceleration follows a frenetic first half that saw total AI-linked debt issuance reach $466 billion year-to-date, more than double the $216 billion issued in all of 2024.
Morgan Stanley characterized the pullback as a “pause, not a retreat,” projecting that issuance will pick up again in the fourth quarter, though it will not replicate the explosive pace seen in the first half of the year.
Why September Went Quiet
The slowdown stems from three overlapping factors. First, front-loading: hyperscalers including Google, Amazon, Meta, and Microsoft issued roughly $132 billion in investment-grade debt this year — a 25-fold increase year-over-year — with the bulk concentrated in the first six months. Including Oracle and SpaceX, the six largest hyperscalers have issued approximately $254 billion across all currencies. June alone saw $113 billion in AI-related issuance, a full-year peak, before volumes declined month-over-month.
Second, regulatory and political friction is becoming a tangible constraint. New data center construction faces permitting delays, power supply bottlenecks, and political resistance — what Morgan Stanley has dubbed the “three P’s”: people, power, and politics. These factors are transitioning from theoretical risks to concrete obstacles.
Third, sharply higher interest rates are forcing project-level terms to be renegotiated. Lower-rated issuers and project finance vehicles are particularly sensitive to funding costs, and the rapid rise in yields has made some deals uneconomical at previously agreed terms.
What It Means for Markets
The pause in AI debt issuance carries significant implications across asset classes:
- Investment-grade credit: With AI issuance at zero in September, supply pressure on IG spreads has eased temporarily. However, the market faces a larger test from macro conditions than from supply. Investment-grade bonds have posted a -3% total return year-to-date, with a -4% quarterly return — a drawdown deep enough to potentially trigger mutual fund and ETF redemptions.
- Equities: Hyperscaler capital expenditure guidance remains the key equity catalyst. Morgan Stanley forecasts combined 2027 capex for the six major hyperscalers at approximately $1.4 trillion — well above consensus estimates of $1-1.1 trillion. Third-quarter earnings, expected in October, could bring another round of upward capex revisions.
- Rates and credit spreads: The 10-year Treasury yield has climbed from 1% to over 5%, creating repeated repricing shocks. Morgan Stanley argues credit markets can absorb higher yields because nominal growth exceeds 6% and corporate earnings growth is even stronger, providing fundamental support.
- Crypto and risk assets: Higher-for-longer rates and tighter project financing conditions could slow the flow of capital into data center buildouts, indirectly affecting the broader AI infrastructure trade that has spilled over into crypto markets via GPU-related tokens and decentralized compute networks.
Q4 Outlook: Hyperscalers Return, But Differently
Morgan Stanley expects Q4 issuance to exceed September levels, driven by high-quality hyperscalers returning to U.S. investment-grade markets while continuing to tap non-dollar markets. Non-dollar hyperscaler issuance has already reached approximately $72 billion this year — nearly one-third of total global hyperscaler supply — with currencies expanding from just dollars and euros to include Canadian dollars, British pounds, Swiss francs, Australian dollars, and Japanese yen.
Data center project financing remains harder to predict. Morgan Stanley has lowered its 2025 ABS and CMBS issuance forecast to $25-30 billion, implying $5-10 billion of supply in Q4. The bank reiterated a preference for secured assets, which comprise roughly 30% of AI-related debt, noting that securitized data center bonds backed by operating assets are less vulnerable to construction risks such as permitting delays and power supply uncertainty.
Key Takeaways for Investors
- The pause is tactical, not structural. The drivers of AI debt issuance — massive capex plans, strong ROIC prospects, and favorable debt-vs-equity funding economics — remain intact. Hyperscaler leverage is just 1.3x (0.4x net), with a cash-to-debt ratio of 132% and a median credit rating of AA-.
- Macro, not supply, is the real risk. Morgan Stanley emphasizes that the macro environment — not issuance volumes — poses the biggest test for credit markets through year-end. With IG bonds already in drawdown territory, fund flow dynamics warrant close monitoring.
- Watch October earnings for capex revisions. Upward revisions to 2027 capex plans could reignite financing demand and validate the AI investment thesis.
- Secured assets offer relative safety. Data center ABS and CMBS backed by operating assets are better insulated from construction-phase risks than project-level debt.
- The easy spread compression trade is largely done. Morgan Stanley believes most of the hyperscaler spread compression has already played out, with issuance becoming more predictable and capex shifting toward shorter-duration assets like chips.
