There is a number in enterprise software that has never existed before, and there is a second number sitting right next to it. Oracle ended fiscal 2026 with remaining performance obligations of $638 billion, up 363% year on year, and free cash flow of negative $23.7 billion.
Both are consequences of the same decision. Oracle signed enormous multi-year AI cloud contracts, and delivering them requires building data centres now against revenue that arrives later, which is why record operating cash flow of about $32 billion was entirely consumed by construction.
Oracle reports first quarter fiscal 2027 results after the US close on 10 September 2026, early on 11 September in India. The question is not whether demand exists. It is whether demand is becoming revenue fast enough to justify the spending.
What a backlog actually promises
A remaining performance obligation is contracted work not yet delivered. The measure shows demand and visibility, and it says nothing at all about timing, margin, or whether the customer can pay.
The composition here is unusual enough to matter. Most of the recent increase came from large scale AI contracts where the customer prepaid Oracle for GPUs or supplied the hardware directly, and those prepaid and customer-supplied portions total about $75 billion. That is a real commitment, and it is also a reminder that a chunk of the backlog is hardware pass-through rather than recurring software economics.
Why the market is nervous about a record order book
Because the last set of results produced a sell-off on exactly this point. Oracle's fourth quarter revenue hit a record and the shares still fell over 7% after hours as capital expenditure of $55.7 billion alarmed investors, which tells you where the market's attention has moved.
The arithmetic behind that reaction is simple. A company converting roughly 12% of its backlog into revenue in a year while spending $70 billion to build the capacity is financing tomorrow's revenue with today's balance sheet, and that works only while credit stays cheap and demand stays firm.
A backlog is a promise about the future. Capital expenditure is a bill in the present.
Growth on the delivered side has been genuine, with Oracle Cloud Infrastructure growing 93% in the latest reported quarter and cloud revenue guided to rise 58% to 64%. The debate is about the gap between that growth and the size of the commitments behind it.
Why an Indian investor should care
The connection runs through two channels. Indian IT services firms including TCS, Infosys and HCLTech sell to many of the same enterprises funding AI deployments, so the pace of AI infrastructure spending shows up in their deal pipelines with a two to three quarter lag.
The second is physical. India's own data centre build-out is part of the same global cycle, and the capital, power and land decisions being made by hyperscalers set the pricing environment domestic developers operate in, a link our India data centre boom piece covers alongside the power constraint in AI's building spree.
There is also the sentiment channel. AI capex numbers set the tone for global risk appetite, and Indian IT indices have moved on Nvidia and hyperscaler guidance more than once in 2026, a pattern our AI stocks India piece traces.
What investors should watch
The first is conversion, not bookings. The useful figure in tonight's release is how much revenue was actually recognised from the backlog, since another record RPO number without conversion would confirm the market's worry rather than settle it.
The second is capex guidance. A raise above the $70 billion FY2027 path pushes free cash flow further negative and increases the financing requirement, while a trim would be read as demand cooling. Neither is comfortable, which is what makes it the sharpest line in the release.
The third is customer concentration. Very large AI contracts mean a small number of counterparties, and the credit quality of those counterparties is now part of Oracle's own risk profile.
The fourth is what it says about the cycle. Whether AI capex is investment or overbuild is the question our AI bubble 2026 analysis works through, and single company results are the only real evidence anyone has.
Risks to monitor
The second risk is financing. Sustained negative free cash flow means the build is funded by debt and prepayments, and the cost of that debt is rising alongside a US 10-year Treasury near 4.85% and a Federal Reserve that markets put at roughly 63 to 65% odds of hiking on 16 September.
The third is timing. Depreciation on tens of billions of dollars of infrastructure starts hitting earnings well before the full revenue arrives, which compresses reported margins in exactly the period investors are watching most closely. This is general information, not investment advice.
Enterprise software spent thirty years being valued on how predictable it was. Oracle has spent one year turning itself into something closer to an infrastructure developer, and the market has not decided yet whether to pay for it on the old multiple or the new risk.