Hyperscalers drive massive AI spending and debt issuance

Major technology hyperscalers, including Amazon, Alphabet, Meta, Microsoft, Oracle, and Nvidia, are significantly increasing capital expenditures and debt issuance to fund artificial intelligence infrastructure. UBS projects that these companies could spend as much as $4.1 trillion on AI infrastructure between 2026 and 2028.

JPMorgan has noted that hyperscaler debt issuance is beginning to compete with US Treasury bonds for investor attention. Through late August 2026, these companies collectively issued approximately $320 billion in new long-term investment-grade debt, representing roughly 68-70% of total new long-term US Treasury bond issuance during the same period. This surge in corporate borrowing may push Treasury yields higher as the government competes for fixed-income capital.

As part of the infrastructure build-out, companies are addressing massive electricity constraints. Oracle, for instance, is designing multi-gigawatt data center campuses intended to be powered by small modular reactors (SMRs). NuScale Power is a notable player in this sector, holding the only SMR design approval from the Nuclear Regulatory Commission.