Higher borrowing costs threaten the economics of AI infrastructure development, creating a new challenge for data centers. With the 10-year Treasury note yielding 5.22 percent after hitting multi-decade highs, financing costs are climbing rapidly. Morgan Stanley estimates $3 trillion will be spent on AI build-out through 2028, with roughly half financed through debt.
To attract financing, AI-related debt issuers must offer premiums over Treasury yields, meaning projects need to meet higher hurdle rates as rates rise. Meta and Blue Owl recently sold $27 billion in bonds at 6.58 percent for a Louisiana facility, while a Meta-backed El Paso data center carried a 7.53 percent yield in July. Smaller neoclouds face even more expensive financing, with Core Weave warning that interest payments could rise $30 million per quarter for each percentage point rate increase, with some bond yields reaching nearly 10 percent.
The challenge is circular: the flood of corporate bonds from AI hyperscalers increases competition for debt issuances, helping drive up Treasury yields further.
Source: New York Times Business · Summarized by HeadlinesBriefing