AI debt financing and credit-market capacity: AI debt supply remains financeable, but investors are demanding more compensation and differentiating sharply by issuer and deal

Overview

This UBS Global Strategy monitor assesses whether debt markets can absorb rising AI-related financing needs across investment-grade bonds, high-yield bonds, leveraged loans, CDS and datacenter ABS. UBS concludes that market capacity remains intact, but supply is clearing at more demanding prices and with increasingly wide dispersion across issuers, maturities and transaction structures.

Core views

UBS frames the central issue as the ability of debt markets to finance an expanding AI capital-expenditure cycle in the second half of 2026. Its conclusion is that capacity is not the binding constraint: investors are still buying AI-linked supply, but require higher compensation and more favorable terms. The monitor tracks issuance, CDS, IG cash-credit curves, datacenter leveraged-finance performance and datacenter ABS to judge how that capacity is clearing. US investment-grade issuance continues to demonstrate capacity despite higher Treasury yields. The average weekly US IG issuance run rate in September was about $41bn, versus about $38bn in August, and year-to-date US IG technology issuance increased by roughly $20bn since the previous update. However, September technology supply was lighter than expected and lacked notable mega-deals, which UBS attributes to elevated long-end rate volatility. Supply has broadened beyond hyperscalers into semiconductors, hardware, networking and datacenters; UBS expects this diversification to continue into 2027 as financing needs spread across more AI-sensitive subsectors. Credit performance has been more resilient than equity performance at the aggregate level, but issuer-level dispersion is substantial. Hyperscaler CDS spreads were broadly stable since August while hyperscaler equities were down about 1.4% over the same period. Month-to-date, spreads tightened across hyperscalers and semiconductors but widened for HY datacenters. Within hyperscalers, Oracle and Meta outperformed while Amazon and Alphabet underperformed month-to-date. Within semiconductors, Nvidia and Broadcom widened relative to the August update but rallied from their widest levels after earnings. UBS views this divergence as evidence that AI credit is becoming a security-selection market shaped by earnings, monetization perceptions and expected debt-financing needs. IG cash spreads have generally held firm, supported by lighter September supply and earnings that reinforced the case for AI monetization. Semiconductor curves bull-flattened since the August update, hyperscaler curves tightened broadly across maturities, and power/cooling and networking were relatively stable. Yet duration remains UBS's principal IG vulnerability. Hyperscaler bonds issued this year traded at roughly 41bp of 10s30s curve steepness, little changed from the prior month, and UBS notes that the back end has not begun to outperform. Meta's curves steepened further, while Alphabet's August issuance traded at similar steepness to its February deal. UBS expects additional financing needs and persistent rate volatility to steepen IG 10s30s technology curves further. New-issue pricing shows that demand remains available but at a wider clearing level. For comparable maturities, repeat-issuer new-issue spreads were about 26% wider for Dell and about 40% wider for Alphabet than their earlier 2026 transactions. Even so, August deal subscription levels remained a healthy 3-5x, indicating that investors will absorb supply when spreads adequately compensate them. The report therefore expects broader AI-linked issuance to bring continued selectivity and larger new-issue concessions. The weakest and most differentiated area is leveraged datacenter finance. HY datacenter and neocloud performance varies materially by tenant quality, project stage, permitting, lease structure and tranche seniority. ZENARC issued at a 9.0% yield and its spread widened nearly 140bp after issuance; its bonds traded down to 92 cents following a permit denial. By contrast, projects closer to completion held in more firmly. Leveraged-loan datacenter deals underperformed over the past month as higher rates increased floating-rate interest burdens for capital-intensive borrowers. UBS expects higher-for-longer rates to pressure leveraged datacenter operators that remain at earlier development stages. Datacenter ABS performance also became more mixed. Non-AAA tranches underperformed in spread over the past month, with BBB tranches particularly weaker, as higher rates, thinner late-summer liquidity and expectations for more issuance weighed on performance. UBS's overall message is that AI financing remains accessible, but the terms of financing, maturity exposure and deal-level underwriting are becoming increasingly important determinants of credit performance.

Analysis framework

UBS monitors AI-linked financing across issuance volumes, new-deal pricing, CDS spreads, cash-credit curves, bond-price performance, order-book demand and datacenter ABS metrics. It compares performance since its 10 August update across hyperscalers, semiconductors, networking, power/cooling and datacenter credits, then links spread and curve changes to rate volatility, debt-financing needs, earnings, monetization expectations and project-level underwriting factors.

Methodology notes

• Fixed Income and CreditSpread analysis

  Credit-spread and new-issue-concession analysis

  The report uses changes in CDS and cash-bond spreads, issuance spreads and deal pricing to assess the compensation investors demand for AI-linked debt.

• Fixed Income and CreditYield curve analysis

  10s30s technology credit-curve analysis

  UBS compares 10-year and 30-year spread relationships to identify long-duration pressure and assess whether the back end of hyperscaler curves is being rewarded for additional risk.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI financing monitored across hyperscalers, semiconductors, hardware, power/cooling and datacenters

  The report follows how AI capital needs and financing conditions extend from large platform companies into supporting technology and physical-infrastructure subsectors.

Asset mapping & comparison

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

• Meta Platforms (META)

  Hyperscaler included in UBS's CDS and IG cash-credit-curve analysis.

  Strengths

  CDS outperformed month-to-date; hyperscaler curves tightened broadly.

  Weaknesses

  Meta's credit curve steepened further and its performance was described as more idiosyncratic.

  Comparison

  Outperformed Amazon and Alphabet in month-to-date CDS performance; compared with other hyperscalers including Oracle and Microsoft.

  Risks

  Long-end curve exposure remains vulnerable to rate volatility and additional AI-financing supply.

• Alphabet (GOOGL)

  Hyperscaler and repeat issuer analyzed for CDS and new-issue pricing.

  Strengths

  August issuance traded at similar curve steepness to its February deal; investor demand remained available at wider spreads.

  Weaknesses

  Underperformed month-to-date among hyperscalers; comparable-maturity new-issue spreads were about 40% wider than earlier 2026 transactions.

  Comparison

  Lagged Oracle and Meta month-to-date; compared with Dell on repeat-issuer concessions.

  Risks

  Higher debt-financing needs and long-end rate volatility could maintain pressure on longer-dated bonds.

• AMD (AMD)

  Semiconductor issuer included in UBS's AI-linked CDS and IG issuance monitoring.

  Strengths

  Its recent IG bonds showed lower latest spreads than issuance across several maturities.

  Weaknesses

  Semiconductor performance remained differentiated at the issuer level.

  Comparison

  Part of the semiconductor cohort with Broadcom, Intel and Nvidia.

  Risks

  Semiconductor credit spreads were wider month-on-month despite retracing from peaks.

• ZENARC

  HY datacenter issuer used as an example of transaction-specific underperformance.

  Weaknesses

  Issued at a 9.0% yield; spreads widened nearly 140bp since issuance and the bond traded down to 92 cents.

  Comparison

  One of the larger underperformers in UBS's datacenter sample, while projects nearer completion held firmer.

  Risks

  A permit denial illustrates project-stage and permitting risks in HY datacenter financing.

Key data

• September average weekly US IG issuance~$41bnVersus roughly $38bn in August despite higher Treasury yields.
• Increase in US IG technology issuance since prior update~$20bnYear-to-date increase; other credit-market issuance was relatively stable.
• Hyperscaler equity performance since August~ -1.4%While hyperscaler CDS spreads were broadly stable.
• Hyperscaler 10s30s curve steepness~41bpFor bonds issued year-to-date; little changed from the prior month.
• Repeat-issuer spread widening~26% for Dell; ~40% for AlphabetComparable-maturity new-issue spreads versus earlier 2026 transactions.
• August deal subscription levels3-5xUBS cites healthy demand despite wider new-issue pricing.
• ZENARC issuance and subsequent spread move9.0% yield; nearly +140bpIts spread widened nearly 140bp since issuance; the bond traded to 92 cents after a permit denial.

Impact & implications

UBS believes the AI debt boom can continue because investors are still funding new supply, but financing conditions are becoming more issuer-specific and maturity-sensitive. The report sees long-duration IG technology debt and floating-rate, early-stage leveraged datacenter borrowers as the main pressure points, while stronger issuers and more advanced projects appear better positioned to retain investor support.

Risks

• Higher Treasury yields and long-end rate volatility could further pressure long-duration technology credit.
• Broader AI-linked issuance could require larger new-issue concessions and steepen technology credit curves.
• Floating-rate structures and higher-for-longer rates could weigh on leveraged, capital-intensive datacenter operators.
• Datacenter credit outcomes may deteriorate where tenant quality, permitting, development stage, lease structure or tranche seniority are weaker.

What to watch

• The pace and subsector composition of AI-linked debt issuance, particularly whether supply broadens beyond hyperscalers.
• New-issue concessions and order-book subscription levels as indicators of debt-market capacity.
• 10s30s technology curve steepness and long-end sensitivity to Treasury yields and rate volatility.
• Issuer-level CDS and cash-spread dispersion following earnings and changes in AI monetization expectations.
• Datacenter project execution, permitting outcomes, tenant quality, lease structures and floating-rate interest burdens.
• Performance of non-AAA datacenter ABS tranches as future issuance expectations evolve.