The number that stands out
Oracle closed its fiscal first quarter, the three months ended August 31, 2026, with $664 billion sitting in remaining performance obligations (RPO), the accounting term for revenue under contract that has not yet been billed. That figure grew by $209 billion over the past year, and $30 billion of it was booked in this quarter alone, according to results the company announced on September 10. For a company that reported $19.3 billion in total revenue for the quarter, a contracted backlog more than 34 times that size is an unusual gap, and it says more about where Oracle's business is heading than the quarterly revenue line does.
Where the growth is coming from
Oracle splits its cloud business into two pieces, and they are moving at very different speeds. Cloud infrastructure, the raw computing and storage capacity Oracle sells to run other companies' workloads, brought in $7.4 billion, up 121 percent from a year earlier. Cloud applications, Oracle's software-as-a-service products such as its enterprise resource planning suite, brought in $4.2 billion, up a comparatively modest 10 percent. Combined, total cloud revenue reached $11.6 billion, up 62 percent year over year, a figure consistent with the two segments' individual growth rates once weighted by their size.
Total company revenue of $19.3 billion was up 30 percent. Cloud revenue of $11.6 billion accounts for roughly 60 percent of that total, with the remaining $7.7 billion coming from license support, hardware and services, the non-cloud businesses Oracle does not break out growth rates for individually. Non-GAAP (a profit measure excluding certain one-time and non-cash items) earnings per share came in at $1.92, up 30 percent from a year ago and above the $1.73 that analysts had expected, a beat of roughly 11 percent. GAAP earnings per share, the unadjusted figure, was $1.56, up 55 percent. Non-GAAP operating margin was 42 percent; the GAAP figure was 35 percent.
What is actually filling the backlog
The growth is overwhelmingly tied to artificial intelligence workloads. Oracle said it delivered more than 300,000 graphics processing units (GPUs), the specialized chips used to train and run AI models, to cloud customers since the end of its previous quarter, and that its GPU fleet is running at 97.9 percent utilization, leaving little idle capacity. The company added 850 megawatts of data center capacity during the quarter to keep pace with demand. The $30 billion in new contracts signed this quarter, on top of the existing backlog, points to customers locking in AI computing capacity well ahead of when they will actually consume it.
Guidance going forward
Oracle raised its outlook alongside the results. For the full 2027 fiscal year, it now expects total revenue of at least $90 billion and non-GAAP earnings per share of $8.10. For the current quarter, it guided to total revenue growth of 30 to 34 percent, cloud revenue growth of 64 to 70 percent, and non-GAAP earnings per share between $1.83 and $1.91. Sustaining the top end of that cloud growth range would require Oracle to keep converting backlog into billed revenue at roughly the same pace it has managed this quarter, which is the practical test its next few earnings reports will answer.
What it means in practice
For enterprise IT buyers, the scale of Oracle's backlog is a signal that AI infrastructure capacity is being reserved well in advance industry-wide, which is worth factoring into any organization's own planning for GPU access or cloud AI budgets over the next year. For anyone tracking the cloud infrastructure market, Oracle's 121 percent infrastructure growth rate, far outpacing its 10 percent applications growth, is further evidence that the current cloud spending boom is concentrated in AI compute specifically rather than a broad-based rise in enterprise software spending. The gap between $19.3 billion in quarterly revenue and $664 billion in signed-but-unbilled contracts also means investors and customers alike are watching execution risk: Oracle has to actually build and deliver the data center capacity behind those contracts, not just sign them.
