As of Sept. 13, Oracle said its remaining performance obligations (RPO) had reached about $664 billion after more than $30 billion of additional AI-cloud contracts in the fiscal first quarter ended Aug. 31. Oracle released the figures on Sept. 10. RPO is contracted future revenue not yet recognized as sales; it is a measure of booked demand, not a guarantee of near-term cash.
The conversion timetable matters. Oracle and its CFO have said roughly half of the $664 billion should be recognized within 36 months. That still leaves a large share to be delivered later, and delivery depends on Oracle adding power, GPUs and data-center capacity. Oracle’s cloud-infrastructure revenue rose 121% year over year to $7.4 billion in the quarter, but revenue growth does not turn the full RPO balance into cash immediately.
The financing mechanics are unusually important in these contracts. Reuters, citing Oracle’s commentary, reported that the vast majority of new orders were structured with customer prepayments or bring-your-own-hardware arrangements. Customer-supplied GPUs and advance cash reduce the chips Oracle has to buy upfront, easing its near-term cash burden while shifting more of the hardware funding to customers. Oracle’s prior investor materials also put prepaid and customer-supplied GPU portions of large AI contracts at roughly $75 billion when RPO was $638 billion in June.
That structure helps explain how Oracle can pursue the backlog without funding every GPU itself, but it does not make the build-out free. Oracle’s capital-spending plan points to roughly $90 billion to $95 billion, while first-quarter operating cash flow was about $23 billion and free cash flow remained negative at roughly $5 billion, according to company materials and reported results. The cash-flow picture was better than some fears, in part because customer prepayments supplied financing, but the business still has to spend ahead of revenue recognition.
Oracle also completed a $20 billion at-the-market equity sale in the quarter. Prior company commentary, as reported by secondary sources, put planned fiscal 2027 fundraising at about $40 billion, combining equity with debt or other financing. The customer-funded portions can lower chip cash needs; they do not remove the need to finance data centers, networking, power and the rest of the infrastructure. Oracle has said the contract structure did not change its capital-raising plan.
The competitive question is whether Oracle can convert the RPO at attractive economics while AWS, Microsoft Azure and Google Cloud are also racing to secure AI workloads and capacity. Large commitments from customers including OpenAI, xAI and Meta have been reported, but individual contract economics and delivery schedules are not fully public. Claims about the pace of conversion and GPU utilization should therefore be read as company or secondary-source disclosures, not as independently verified cash receipts.
The takeaway is narrower than a stock-price reaction: $664 billion is substantial booked demand, but its value depends on when it converts, how much customers prepay or supply themselves, and how Oracle funds the capacity required to deliver it. The open issue for investors is not whether the backlog is large, but how much becomes cash after capex, interest and the next round of debt or equity financing.