# US corporate AI debt surge tests investor limits as fatigue emerges > Source: > Published: 2026-08-21 14:15:33+00:00 Via newsweek.com # US corporate AI debt surge tests investor limits as fatigue emerges Hyperscalers have issued roughly $220 billion in AI-linked debt in 2026, up from $12.5 billion a year earlier, and bond investors are starting to demand bigger premiums to keep buying. US hyperscalers have issued approximately $220 billion in debt linked to AI spending as of mid-August 2026. For context, the same group raised just $12.5 billion during the equivalent period in 2025. That’s not a typo-free way of saying it grew a lot. It’s a 17x increase in roughly twelve months. ## The numbers behind the borrowing binge Amazon recently priced a $25 billion long-dated bond offering at approximately 120 basis points over Treasuries. That spread is significantly wider than what the company commanded in previous years, a clear signal that investors are demanding more compensation for absorbing yet another massive slug of tech paper. Tech investment-grade spreads overall now sit at 89 basis points over Treasuries. That’s 9 basis points wider than the broader investment-grade market. George Catrambone of DWS, the asset management arm of Deutsche Bank, pointed to fatigue as a defining feature of late-2026 offerings. New deals have required 10 to 15 basis points in concessions just to get investors to show up. ## How we got here Goldman Sachs has estimated that AI-related debt could reach roughly $489 billion by mid-2026, while Morgan Stanley’s global forecast runs as high as $570 billion. ## What this means for borrowers and buyers The widening of tech credit spreads relative to the broader IG market represents a real increase in the cost of capital for companies that have built their AI strategies around cheap debt. The 10 to 15 basis point concession that new deals require is essentially a tax on issuers for saturating the market. If that concession grows, it could cascade into secondary markets, pushing down prices on existing tech bonds and creating paper losses for holders. **Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).