Oracle reported remaining performance obligations of $523 billion at the end of the most recent quarter, reflecting the total value of signed contracts not yet recognized as revenue. Of that backlog, approximately $300 billion is tied to a single cloud infrastructure agreement with OpenAI, making the start-up the dominant factor in Oracle’s RPO growth this year. With over 700 AI customers on its books but one contract accounting for more than half of unrecognized revenue, investors face significant customer concentration risk that hinges on OpenAI’s ability to fulfill its payment commitments.

During its recent earnings call, co-CEO Clay Magouyrk detailed Oracle’s ability to reassign AI computing capacity between customers in a matter of hours. Thanks to the company’s uniform, batch-agnostic GPU architecture, Oracle regularly shifts thousands of GPUs daily to meet changing customer requirements. This flexibility underpins a high utilization rate—critical to sustaining the company’s 65 percent gross margin—by ensuring that depreciation costs on servers and GPUs are spread across active workloads rather than idle capacity.

While fast reallocation addresses short-term customer defaults, it offers little protection if the AI industry overbuilds capacity. Goldman Sachs projects data-center occupancy rates will peak in late 2026 before moderating as new facilities come online. In that scenario, excess GPU and server capacity would be hard to absorb, leaving Oracle with underutilized assets and elevated debt levels used to finance its data-center expansion. Should demand fall short of optimistic forecasts, Oracle’s dependence on OpenAI and its own balance-sheet exposure could turn a once-touted strength into a major financial liability.