Goldman Sachs raises its 2027 hyperscaler debt-issuance forecast to $420bn and remains cautious on AI-related credit exposure.

Overview

This Global Credit Trader update examines how rising AI infrastructure spending is reshaping global corporate-credit issuance, market composition, and relative-value considerations. Goldman Sachs raises its forecast for direct hyperscaler borrowing and argues that the extended buildout warrants caution toward AI-linked credit even as the broader USD IG market remains resilient.

Core views

Goldman Sachs refreshes its top-down AI financing framework after the latest earnings season. Using Global Investment Research equity CapEx estimates, it still assumes 35% of 2027 hyperscaler CapEx will be debt financed, but modest upward revisions to CapEx now imply $420bn of direct global hyperscaler gross debt issuance in 2027, versus its prior $400bn estimate. This excludes data-center and chip project-finance issuance through special-purpose vehicles, for which Goldman Sachs separately forecasts $300bn in 2027. The covered hyperscalers are Alphabet, Amazon, Meta, Microsoft, and Oracle; the firm expects elevated AI CapEx through the end of 2027 because of an ongoing supply-demand imbalance in the buildout. The projected $420bn of 2027 direct issuance would be more than 60% above Goldman Sachs' $250bn full-year 2026 estimate. Year to date, hyperscalers have issued $229bn across currencies. Goldman Sachs expects 65% to 75% of 2027 hyperscaler gross issuance to price in the USD IG bond market, conditional on market conditions and investor receptivity given issuer-concentration concerns. It also believes the group has meaningful capacity to expand debt before reaching the size of today’s largest issuers. The report places hyperscaler borrowing within a broader financing cycle: it tracks $578bn of global gross AI-related debt issuance year to date, including hyperscalers, project-finance issuers, and other AI-linked companies. Of this total, Goldman Sachs estimates $490bn in global IG markets and $88bn in global leveraged-finance markets. Hyperscalers comprise only 40% of overall AI-related issuance, demonstrating that AI financing reaches well beyond the largest technology companies. For AI-adjacent sectors such as Energy & Utilities, Oil & Gas, Machinery, Chemicals, and Metals & Steel, the estimates count 30% of gross issuance as AI-related; undrawn facilities, revolvers, and pending financings are excluded. Financing is expected to broaden across currencies. Goldman Sachs expects the EUR share of AI-related issuance to rise because its year-to-date contribution has lagged other global IG markets, while other non-USD IG markets are much smaller than EUR IG and therefore have less capacity to absorb incremental issuance. The firm also sees convertible bonds taking a larger role in financing the AI buildout. In index terms, AI-linked credit is already a material and growing share of USD markets: 12% of USD IG notional and 6% of USD HY, compared with 3% and 1%, respectively, in EUR IG and HY. Goldman Sachs sees room for these shares to expand as the multi-year issuance cycle continues; for comparison, Banking accounts for 22% of USD IG and 31% of EUR IG market value. The report finds a persistent performance bifurcation between AI-related spreads and the rest of the market. The broader USD IG market has remained relatively resilient even as AI-related spreads widened. Goldman Sachs attributes that resilience to solid economic activity, constructive credit fundamentals, attractive all-in yields that sustain demand from yield-oriented investors, and diversification from large sectors less tied to AI, including Banks, Energy, and Healthcare. Although AI-related spreads—particularly hyperscaler spreads—have recently found local relief, the firm remains cautious about adding meaningful exposure. Near-term funding needs may be partly completed, but the scale and duration of the buildout leave issuance risks skewed to the upside if benign sentiment enables further supply. Goldman Sachs considers hardware credits less insulated from pressure as compute shipments and related financing grow. It also views data-center and compute project-finance issuance as particularly difficult to assess, while expecting that activity to continue. The report calls for selectivity in that segment because project-specific risks must be underwritten against a technical backdrop unlikely to ease soon. Across its broader relative-value dashboard, Goldman Sachs has a slight preference for USD credit over EUR credit, citing a more challenging growth-inflation-policy mix in the euro area, while recognizing that cross-border issuance is blurring regional distinctions. It is neutral between IG and HY in both USD and EUR after previously favoring IG, as valuation, rate sensitivity, and supply technicals make a clear IG overweight less compelling. Within USD IG it favors BBBs over AAs and As for additional spread and potential balance-sheet deterioration at the highest rating tiers. In USD HY, it favors Bs, is neutral BBs, and underweights CCCs; BBs face a challenging AI-related supply technical, while CCCs require careful issuer selection. It prefers HY bonds over leveraged loans, in part because of loans’ substantial software exposure and unresolved 2028 refinancing needs. For floating-rate investors, it favors CLO mezzanine tranches over leveraged loans, citing historically low impairment for BB and BBB CLO tranches and carry that can cushion moderate-to-elevated default outcomes.

Analysis framework

The report starts with a top-down estimate of hyperscaler debt needs by applying a debt-financing share to equity-research CapEx forecasts. It then compares projected issuance with current issuance, market capacity, currency-market size, and index composition. Goldman Sachs evaluates market implications through spread performance, supply technicals, credit-market diversification, issuer concentration, and relative-value comparisons across ratings, regions, HY bonds, leveraged loans, and CLOs.

Methodology notes

• Industry AnalysisSupply-demand framework

  Top-down AI financing and issuance analysis based on hyperscaler CapEx, debt-financing assumptions, and credit-market absorption.

  Goldman Sachs applies a 35% debt-financing assumption to projected 2027 CapEx, then considers issuance supply, investor demand, market capacity, and currency-market depth to assess credit implications.

• Fixed Income and CreditSpread analysis

  Comparison of AI-related and non-AI credit spreads, as well as relative spread compensation across regions and rating cohorts.

  The report uses spread behavior and spread differentials to identify divergence within the AI ecosystem and to frame relative-value preferences.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI buildout financing is assessed across hyperscalers, hardware, data-center project finance, and AI-adjacent sectors.

  The report links higher AI CapEx and compute shipments to borrowing needs and then to differentiated credit pressure across the financing ecosystem.

Asset mapping & comparison

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

• Alphabet, Amazon, Meta, Microsoft, and Oracle

  The report defines these issuers as the hyperscaler group driving its direct 2027 debt-issuance forecast.

  Strengths

  Goldman Sachs sees meaningful capacity for the group to grow debt before reaching the size of today’s largest issuers.

  Weaknesses

  Their expanding funding needs contribute to issuer-concentration concerns in USD IG.

  Comparison

  The group is compared with the three largest issuers in the Bloomberg USD IG Corporate Index.

  Risks

  A multi-year AI buildout could generate more issuance than currently expected.

• AI-related hardware credits

  Part of the wider AI financing ecosystem exposed to rising compute shipments and related financing.

  Weaknesses

  Goldman Sachs believes hardware names are less insulated from credit pressure as the buildout expands.

  Comparison

  Viewed as less insulated than other AI-linked areas.

  Risks

  Growing compute financing and supply pressure.

• Data-center and compute project-finance debt

  Finances AI infrastructure outside direct hyperscaler borrowing.

  Strengths

  Goldman Sachs expects issuance activity to continue.

  Weaknesses

  The scale of issuance is difficult to assess.

  Comparison

  Excluded from the direct hyperscaler forecast and assessed separately.

  Risks

  Project-related risks and a technical backdrop unlikely to abate near term.

Key data

• 2027 direct global hyperscaler IG issuance forecast$420bnRaised from roughly $400bn following modest upward revisions to CapEx forecasts.
• Debt-financed share of 2027 hyperscaler CapEx35%Core assumption in Goldman Sachs' top-down issuance framework.
• 2026 direct global hyperscaler IG issuance estimate$250bnThe 2027 forecast is more than 60% higher.
• Hyperscaler debt issued year to date$229bnIssued across currencies during 2026 to date.
• Global AI-related gross debt issuance year to date$578bnIncludes hyperscalers, project-finance issuers, and other AI-related entities.
• AI-related global IG and leveraged-finance supply$490bn IG; $88bn leveraged financeGoldman Sachs estimates of year-to-date AI-related issuance.
• Hyperscaler share of AI-related issuance40%Shows that AI-related financing extends beyond hyperscalers.
• AI-related index notional share12% USD IG; 6% USD HY; 3% EUR IG; 1% EUR HYGoldman Sachs expects these shares to rise as the issuance cycle extends.
• Expected USD IG share of 2027 hyperscaler issuance65% to 75%Subject to market conditions and investor receptivity.

Impact & implications

Goldman Sachs expects AI financing to become a larger structural source of global credit supply, with USD IG absorbing most hyperscaler issuance and EUR markets taking a greater share over time. The report sees broader USD IG resilience but argues that incremental AI-related exposure requires caution because issuance, concentration, hardware financing, and project-specific risks may continue to pressure spreads.

Risks

• Hyperscaler issuance could exceed estimates if benign market sentiment enables more debt financing during the multi-year buildout.
• Issuer concentration may constrain investor receptivity to the projected USD IG supply.
• Hardware credits may face greater pressure as compute shipments and associated financing increase.
• Data-center and compute project-finance debt carries project-specific underwriting risks and uncertain issuance volumes.

What to watch

• Revisions to hyperscaler CapEx forecasts and the implied debt-financing requirement through 2027.
• The share of hyperscaler issuance priced in USD IG versus EUR and other markets.
• AI-related spread performance relative to the broader USD IG market.
• Growth in AI-related debt as a share of USD and EUR IG and HY indices.
• Data-center and compute project-finance issuance and related market technicals.