Throughout 2025, Oracle significantly expanded its AI infrastructure footprint, driven by generative AI demand. In the quarter ended August, cloud infrastructure revenue rose by 68% year-over-year to an annualized run rate of $16.5 billion. Simultaneously, remaining performance obligations (RPO) jumped from $113 billion at the end of 2024 to $455 billion by September, then to $523 billion by November. This half-trillion dollar backlog reflects multi-year contracts, with one customer—OpenAI—accounting for roughly $300 billion of the total. Management forecasts that AI-specific commitments will eventually deliver gross margins of 30%–40%, compared with historical database software margins above 60%.

To fund its data center expansion, Oracle issued incremental debt, increasing gross borrowings from $96 billion at the end of 2024 to nearly $130 billion by year-end. Free cash flow swung to a negative $10 billion in the most recent quarter as capital expenditures outpaced operating cash inflows. In December, credit default swap spreads on Oracle’s five-year debt widened to 1.41%, the highest level since 2009, signaling investor concern over elevated leverage and customer concentration. Analysts note that if OpenAI’s commitments underperform or margins fall below projections, Oracle’s coverage ratios could weaken materially.

Oracle’s reliance on OpenAI exposes it to competitive shifts in the AI landscape. In late 2025, Alphabet’s Gemini 3 model, built on proprietary TPU chips, forced OpenAI to declare a “Code Red” to regain model leadership. Concurrently, rivals such as Anthropic and xAI secured multi-billion-dollar partnerships that could divert future capex away from Oracle’s cloud. Should frontier model performance plateau or alternative infrastructure providers gain share, Oracle could face underutilized data centers and pressure on its multi-year RPO to translate into realized revenue.