AI data center financing expansion is pushing credit supply higher, with investors favoring shorter duration and amortizing structures

Overview

This report consolidates views from Morgan Stanley investors, including macro accounts, core credit investors, corporate credit investors, and securitized credit specialist investors, on AI and data center financing. It notes that AI-related issuance is growing rapidly, with financing channels expanding from traditional U.S. dollars debt to non-USD debt, loans, equity, and private placement structures. Although investors still want exposure to AI and data center themes, concerns are rising about overbuilding, weak demand, technological obsolescence, power and supply-chain bottlenecks, local regulatory resistance, and credit quality dispersion.

Core views

The core view is that the fixed-income market can still provide financing for the AI data center investment cycle, but market pricing will increasingly differentiate by asset class, issuer quality, and transaction structure. In investment-grade markets, the supply pace of hyperscalers and ORCL, financing mix, non-USD issuance, equity financing, and credit quality are the key focus areas. The report maintains its estimate of AI-related investment-grade issuance at $3.5 billion to $4.0 billion and maintains an underweight stance on IG tech. In high-yield markets, investors focus more on construction risk, structural protections, refinancing after the first call date, and potential rating improvement after completion. In securitized credit markets, investors are watching whether ABS and CMBS can absorb a potential refinancing wave and whether leveraged finance can absorb part of the supply through scale and lower funding costs.

Analysis framework

The report combines investor interviews with a cross-credit-market framework, splitting AI data center financing into investment-grade credit, high-yield credit, and securitized credit, and comparing it along supply, spreads, financing mix, construction risk, tenant risk, asset risk, corporate risk, and macro theme risk.

Methodology notes

• credit risk frameworkData center build debt risk vector

  construction risk, tenant risk, asset risk, corporate risk, macro theme risk

  The report breaks the key risks of data center financing into whether projects can be delivered on time and within budget, whether tenants can pay rent on time, whether assets remain in demand and technically aligned at renewal, whether issuers can achieve returns through the AI investment cycle, and macro theme risk from supply-demand imbalance or bottleneck scarcity.

• market structure analysisInvestment-grade, high-yield, and securitized credit comparison

  differences in pricing the same data center theme risk across different markets

  Investment-grade investors focus more on hyperscaler supply, credit quality, and index eligibility; high-yield investors focus more on build-phase risk, structural protections, and refinancing; securitized markets focus more on stable asset cash flow, ABS/CMBS supply, and the ability to absorb refinancing.

• financing structure analysisfinancing channel diversification

  USD bonds, non-USD bonds, loans, equity, 144A, 4(a)(2) private placement formats

  The report argues that hyperscalers’ expansion of financing channels is not only to reduce cost of capital, but more importantly to broaden the investor base, maintain funding flexibility, and support a long-term capex cycle.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• investment-grade credit

  One of the main supply markets for AI data center financing, especially high-quality hyperscalers, ORCL, and other AI capex-related issuers.

  Strengths

  Large market capacity and broad investor base, and issuers can broaden funding sources through both USD and non-USD markets.

  Weaknesses

  Persistent elevated supply may pressure spread performance, and after ORCL's downgrade investors pay closer attention to fundamental stratification and credit quality.

  Comparison

  Compared with high-yield and securitized credit, the investment-grade market is more affected by issuance pace, index eligibility, the long-end curve, and broader tech credit quality.

  Risks

  Persistent supply, downgrades, off-balance-sheet leverage, returns falling short of expectations, and rising correlation of AI thematic risk with corporate credit risk.

• high-yield credit

  Primarily absorbs data center construction-phase financing and part of financing for some emerging data center developers.

  Strengths

  Higher yield can compensate for construction-phase risk, and 5NC2 structures create refinancing or call paths after completion.

  Weaknesses

  Issuers often have shorter operating histories, with more pronounced uncertainty around delays, cost overruns, and tenant quality.

  Comparison

  Compared with investment-grade markets, high-yield investors place greater emphasis on project-level risk, structural protections, and potential post-completion rating improvement.

  Risks

  Construction delays, cost overruns, inability to refinance at lower cost after the first call date, tenant defaults, and declines in project asset value.

• securitized credit, ABS, and CMBS

  Can provide cash-flow support financing for stable operating data center assets and may absorb refinancing after completion of high-yield construction debt.

  Strengths

  There are mature underwriting frameworks for assets that are completed and generating cash flow, and some structures include amortization protection via DSCR or LTV triggers.

  Weaknesses

  The market’s ability to absorb a large-scale refinancing wave remains unproven; CMBS generally lack amortization mechanisms.

  Comparison

  Compared with corporate credit, securitized credit focuses more on asset cash flow and collateral quality but is also exposed to securitizer risk.

  Risks

  Insufficient ABS/CMBS supply, failed refinancing, asset renewal risk, technological obsolescence, securitizer risk, and weak market liquidity.

• hyperscalers and ORCL credit

  Core financing entities in the AI capex cycle and central to changes in investment-grade supply and market sentiment.

  Strengths

  Most issuers in this group are large, with diversified funding channels and global investor reach.

  Weaknesses

  Very high capex intensity, higher effective leverage from off-balance-sheet commitments, and increasing credit quality differentiation.

  Comparison

  Compared with emerging data center developers, hyperscalers are more accepted by investors; compared with other investment-grade sectors, they face higher supply pressure and AI theme sensitivity.

  Risks

  Lagging AI returns, continued upward revisions to capex, ORCL-style rating pressure, potential short-term exhaustion of non-USD issuance capacity, and sustained spread widening.

Key data

• AI-related debt issuance in global credit marketsabout $336 billion YTDThe report says AI-related debt issuance in global credit markets has reached about $336 billion YTD so far.
• Investment-grade AI-related issuanceabout $222 billion YTD; about $257 billion when AVGO chip financing transactions are includedAs of July 10, AI-related debt issuance in the investment-grade market was about $222 billion.
• Morgan Stanley investment-grade AI issuance forecast$350 billion to $400 billionIt estimates that high-quality hyperscalers and ORCL will contribute $250 billion to $300 billion, with other AI-related debt around $100 billion.
• U.S. investment-grade total supply forecast$2.25 trillionThe report says this forecast is above market consensus and expects investment-grade spreads to underperform versus other fixed-income markets.
• High-yield and leveraged loan issuance estimate$50 billion for high-yield; $15 billion for leveraged loansHY investors broadly accept this issuance estimate, with attention focused more on risk, structural change, and refinancing paths.
• Non-USD issuance by hyperscalers and ORCL$62 billion YTDThe report notes these companies have issued in EUR, GBP, CHF, CAD, and JPY since 2025.
• Hyperscaler and ORCL 2027 capex growth estimateup 54%Revisions by Morgan Stanley equity strategists show capex growth still increasing sharply in 2027.
• Incremental data center capacity in 2027about 27GWChart commentary indicates high-quality hyperscalers are expected to add about 27GW of data center capacity in 2027.
• Data center build debt not yet in index universesabout $53 billionThe report says about $53 billion of data center build debt currently does not have index eligibility.
• High-quality hyperscaler investment-grade spread performance YTDwidening about 24bpThe report says the investment-grade index has narrowed about 2bp YTD, while high-quality hyperscalers have widened about 24bp.

Impact & implications

For investors, the AI data center theme is no longer only a growth financing opportunity but is increasingly becoming a core variable in credit supply, spread dispersion, and structural protections. In the near term, strong demand, offshore demand, and insurance capital may help absorb supply, but if rate cuts slow demand while supply remains elevated, investment-grade markets could see broader repricing. For issuers, financing channel diversification, amortization structures, collateral support, tenant quality, and index eligibility will influence bond performance and refinancing costs.

Risks

• AI data center construction may become excessive, with eventual supply exceeding demand.
• Improving compute efficiency, power efficiency, or token efficiency could reduce future data center capacity demand.
• Power interconnection, supply chains, permits, zoning, local restrictions, and NIMBYism may create construction bottlenecks.
• Construction projects may be delayed or suffer cost overruns.
• Data center assets may face renewal risk if location, workload type, natural disasters, cooling and power systems, or hardware upgrade capacity become insufficient.
• Declining tenant credit quality or failure to pay rent on time could weaken bond cash flows.
• Actual leverage for hyperscalers and ORCL may rise through procurement commitments, lease commitments, guarantees, and SPV liabilities.
• If AI-related investment-grade supply stays high, spreads may continue to widen and drag on the broader credit market.
• The lack of index eligibility for 144A-for-life bonds may affect liquidity, ownership demand, and spread levels.
• Securitized credit markets may not be able to fully absorb a future refinancing wave for data center financing.

What to watch

• Whether AI-related investment-grade issuance in the second half of 2026 remains at a pace similar to the first half.
• Whether hyperscaler and ORCL 2027 capex estimates continue to be revised higher.
• Whether after ORCL's downgrade the market shifts more clearly from supply discussion toward fundamental stratification.
• Whether newly issued AI-related bonds continue to trade wider than issue pricing.
• Whether non-USD bonds, loans, equity, and private market financing channels can continue to broaden the investor base.
• Whether data center construction debt with 144A restrictions is converted to unrestricted CUSIP and gains potential index eligibility.
• Whether HY data center construction debt can refinance at lower cost after the first call date.
• Whether data center construction debt receives rating upgrades after completion.
• Whether ABS and CMBS markets show signs of data center asset supply pressure or refinancing pressure.
• Whether political resistance to data center construction at local and federal levels, permitting constraints, and power interconnection bottlenecks intensify.