AI infrastructure spending remains surprisingly up as hyperscalers signal robust growth, providing the basis for positive rating actions in Q3/25, according to S&P Global Rating.
Enterprise IT spending persists despite macroeconomic uncertainties as enterprises push productivity gains while capping job growth, S&P Global Ratings analyst Christian Frank said in “AI Spending Balloons Despite Bubble Concerns–Taking U.S. Tech Ratings On The Ride.”
The acceleration of hyperscaler spending in pursuit of “planet scale” AI, as Microsoft CEO Satya Nadella put it, has repeatedly surprised to the upside the past two years. This coordinated escalation reflects that the limits on growth for hyperscalers are capacity rather than demand constraints, and management teams view underinvestment as the greater strategic risk than overcapacity.
S&P took one high-profile negative rating action on Oracle since the start of Q3/25. Three upgrades and two positive outlook actions came in the investment-grade category. Within the speculative-grade category, rating actions were mostly for reasons that don’t point to any particular industry trend: five upgrades and eight downgrades, including two defaults (Astra Acquisition and Atlas Midco), with five positive outlook actions and three negative ones.
Beyond AI-specific workloads, hyperscalers benefit from rebounding enterprise migrations of production applications and core database workloads to the cloud. They are also moving infrastructure to support AI use cases that require public or hybrid cloud tools. Improved availability of offerings such as Oracle’s database across other cloud platforms is reducing friction for enterprises with legacy deployments.
Circular financing dynamics are emerging, with interconnected investments between OpenAI, Nvidia, CoreWeave, Microsoft and Oracle potentially creating artificial demand. OpenAI is the linchpin, making deals for hundreds of billions of dollars with hyperscalers, while revenue reportedly remains in the tens of billions of dollars. CEO Sam Altman hinted that OpenAI could achieve $100 billion of revenue in 2027, so it may yet live up to those commitments. It will have to deliver on unprecedented revenue targets to ensure a steady flow of new capital.
Nevertheless, credit fundamentals for rated hyperscalers remain strong. The four largest maintain cash-rich balance sheets with low net leverage despite high gross debt.
