Banks get creative and look further afield as AI-fueled debt soars

AI-driven capital spending is fueling a record wave of debt issuance, with Amazon and Alphabet raising $60 billion in multiple currencies over the last 12 months and Amazon's €14.5 billion March bond sale marking the largest ever in the euro corporate bond market. Alphabet has also set borrowing records in yen, Canadian dollars, Swiss francs and sterling, while hyperscaler capex is projected at about $725 billion this year. The article suggests demand remains strong despite growing supply, but investors are watching for potential saturation as AI-related debt already approaches 15% of U.S. investment-grade issuance.

Analysis

The real signal is not just that hyperscalers can fund capex; it is that they are becoming quasi-sovereign borrowers with balance sheets large enough to reshape global duration demand. By pushing into non-dollar markets, they are effectively exporting U.S. credit supply into Europe and Asia, which should tighten spreads for top-tier IG outside the U.S. while also crowding out marginal corporate issuance in those currencies. That is a quiet negative for lower-quality single-A/BBB industrials that rely on the same global buyer base.

Second-order beneficiaries are the financing ecosystem around AI buildouts: bookrunners, secondary market makers, and data-center landlords with contracted cash flows. The lease-backed structure matters because it shifts the underwriting debate from AI usage growth to counterparty quality; if Amazon is the tenant, the paper trades more like a synthetic AAA asset than venture debt. That should keep high-yield demand alive for a few quarters, but it also means the market is re-pricing “AI infrastructure” from an equity narrative into a credit one.

The risk is not immediate demand exhaustion, but spread crowding plus refinancing fatigue over 6-18 months if issuance remains serial. Investors are likely underestimating how quickly market capacity can be absorbed when multiple mega-cap names and private data-center sponsors hit at once, especially if rates back up or FX volatility increases hedging costs. The contrarian view is that this is less a bubble in borrowing and more a sign that AI monetization is still too early to self-fund; that extends the runway for capex, but it also delays free-cash-flow inflection and raises the probability of a future capital-markets reset.

Key Decisions for Investors

• Long MS / BCS versus a basket of lower-quality global investment-grade credits for 3-6 months: the banks with leading syndicate share should keep monetizing oversized cross-currency deals while weaker IG names face relative spread pressure from supply crowding.
• Buy protection on broad U.S. IG credit via LQD puts or CDX IG downside hedges into any further AI-led issuance wave: risk/reward improves if hyperscaler supply accelerates above current run-rate and duration investors start to demand concession.
• Long AMZN and GOOGL on 3-12 month horizon, but pair with short a basket of data-center REITs or higher-leverage AI infra names: hyperscalers gain financing flexibility and cheaper capital, while levered developers face rising competition for debt capacity and tenant concentration risk.
• Tactically buy selected Euro/sterling IG financials and infrastructure names on issuance-led spread compression over the next 1-2 quarters: hyperscaler diversification into non-dollar markets should create a temporary technical bid for high-grade paper with benchmark liquidity.
• Avoid chasing HY data-center lease notes after the first 15-20 deals; entry-quality matters more than theme exposure. The asymmetric trade is in the underwriters and the highest-quality sponsors, not in mezzanine structures once the market becomes crowded.