AI Debt Binge Is Reordering Risk Hierarchy With Emerging Bonds

Heavy borrowing by US Big Tech to fund AI investment is reshaping corporate credit risk perceptions, with SK Hynix's 2031 bonds yielding only 9bps more than Amazon's—roughly one-third of the gap a year earlier. Credit-default hedging for SK Hynix is now cheaper than for many US technology giants, while emerging-market corporate debt risk premiums are converging with comparable US benchmarks for the first time. The shift signals investor concern that AI-related debt issuance is increasing leverage and credit risk among major US tech companies.

Analysis

The relevant signal is not near-term default risk; it is a repricing of capital-intensity. Hyperscalers funding AI infrastructure with incremental debt face a two-stage valuation headwind: higher interest expense is modest initially, but sustained capex-to-revenue elevation can reduce free-cash-flow conversion and justify lower equity and credit multiples over the next 6-18 months. AMZN is particularly exposed if AWS growth fails to accelerate enough to absorb depreciation, power, networking and accelerator commitments, because leverage metrics can worsen before reported revenue catches up.

SK Hynix's relative credit strength reflects its position at the memory bottleneck, but this is not a clean secular-credit trade: HBM pricing and utilization are unusually sensitive to a small number of AI customers and to eventual supply responses from Samsung Electronics and Micron. Over 1-3 months, continued hyperscaler bond issuance or weaker-than-expected AWS margins should keep pressure on US megacap-tech spreads; over 6-18 months, a memory downcycle or customer inventory correction could reverse SK Hynix's outperformance sharply.

The contrarian view is that Amazon's spread widening may be more a technical consequence of large new issuance than a durable deterioration in solvency. AWS remains a high-quality recurring cash-flow asset, and any capex moderation, asset monetization, or stronger retail cash generation would quickly improve leverage optics. The better expression is therefore relative credit protection rather than an outright bearish AMZN equity position, while requiring confirmation from net-debt/EBITDA, interest coverage and AWS operating-margin guidance.

Key Decisions for Investors

• Initiate a 3-6 month relative-credit position: buy 5-year AMZN CDS protection and sell matched-duration SK Hynix CDS protection, sized duration-neutral. Target a further 10-15bp widening in the AMZN/SK Hynix spread differential; exit if the differential tightens by 10bp or AWS margin guidance improves materially.
• For equity portfolios, modestly underweight AMZN versus MSFT through the next two earnings cycles: MSFT has greater ability to monetize AI capacity through enterprise software pricing, while AMZN needs AWS revenue acceleration to offset capex. Cover the relative short if AWS growth re-accelerates by at least 300bp year over year or AMZN signals materially lower 2027 capex.
• Watch HBM contract-price and capacity guidance from SK Hynix, Micron (MU) and Samsung Electronics. If HBM supply additions imply pricing pressure before 2027, close the SK Hynix credit-leg and consider a tactical long MU/short SK Hynix equity pair; this is an alert, not a current recommendation, absent verified contract and utilization data.
• Avoid unhedged SK Hynix bond exposure for USD portfolios: the incremental spread is unlikely to compensate for KRW volatility. Use CDS or FX-hedged local exposure rather than cash bonds unless the all-in hedged yield premium widens materially.