Hyperscalers secure 42bp bond discount over software rivals
Hyperscalers have secured a 42 basis point borrowing cost advantage over software firms in U.S. bond markets following the launch of ChatGPT. A Banca d'Italia study finds debt investors rapidly priced infrastructure dominance against software disruption risks.
The emergence of the hyperium
Analyzing roughly 14,800 bond placements by U.S. non-financial corporations between January 2020 and February 2026, researchers Marco Albori and Andrea Zaghini find that debt markets quickly established an advantageous spread for AI infrastructure providers.
Following the November 2022 launch of ChatGPT, the five major hyperscalers—Alphabet, Amazon, Meta, Microsoft and Oracle—enjoyed a post-treatment yield discount of 29 basis points relative to all corporate issuers and 42 basis points relative to software firms.
To finance massive capital expenditure in data centres and chips, hyperscalers issued $108 billion in bonds during 2025, up from an annual average of $20 billion in 2023-2024, and raised over $56 billion in early 2026 alone.
SaaSpocalypse hits debt pricing
While infrastructure providers secured cheaper long-term debt, software companies faced widening credit spreads.
Investors grew concerned that generative AI would erode traditional software moat advantages, such as proprietary lock-in and recurring subscriptions.
When compared to the broader tech sector, software firms faced an 87 basis point yield premium post-shock.
The authors label the hyperscaler pricing edge the “hyperium,” noting that credit investors rewarded predictable infrastructure cash flows while penalising disruption vulnerability.
Subsidising tomorrow's monopolies
Bond markets are accelerating market concentration by subsidising dominant AI platforms at the expense of software issuers.
If monetization falters, this massive debt accumulation poses severe repricing risks for financial stability.
Regulators must monitor these spreads before credit dynamics permanently entrench tech monopolies.
Source: No. 1057 - Corporate bond pricing in the AI era
IN: