Oracle's latest results reveal the scale—and tension—inside the AI infrastructure market. The company reported $664 billion in remaining performance obligations, a measure of contracted future revenue, after signing more than $30 billion in new AI cloud agreements during its first quarter.

The demand signal is extraordinary. So is the investment required to convert it into working capacity. AI cloud competition is increasingly a contest in financing, energy access, construction and delivery discipline.

The numbers behind the backlog

Reuters reported on 10 September 2026 that Oracle Cloud Infrastructure revenue reached $7.4 billion, up 121% from a year earlier. Quarterly capital expenditure reached $28.5 billion, while free cash flow was approximately negative $5.4 billion.

The Financial Times also examined the gap between Oracle's contract momentum and the funding burden required to fulfil it. Backlog is not cash already earned: capacity must be built, customers must consume services and revenue must be recognised over time.

Why supply has become the constraint

Training and operating large AI models requires chips, networking, land, power and cooling. Even a signed customer can wait if the provider cannot energise a facility or secure the required equipment. That makes execution speed a product feature.

The economics also depend on contract structure. Long commitments and customer prepayments can reduce financing pressure; short or uncertain utilisation makes large capital projects riskier. Providers must balance rapid expansion with technology obsolescence and concentration among a few large buyers.

The lesson for Saudi and GCC AI projects

Regional projects should model the full capacity stack before announcing scale: grid availability, construction timetable, equipment delivery, customer commitments and financing cost. Sovereign demand and abundant energy can help, but neither removes execution risk.

Buyers should negotiate portability, service levels and a clear ramp schedule. A lower compute price is not valuable if capacity arrives late or locks the workload into an inflexible architecture.

Karim's strategic takeaway

The AI infrastructure winners will not simply have the best model or largest order book. They will turn contracted demand into reliable, utilised capacity without losing financial control. For GCC operators, the strategic advantage is a bankable operating plan that connects power, capital and anchor customers—then measures delivery against it.