Oracle’s AI Capex Is Becoming Customer-Financed
Oracle generated a record $23.1 billion of operating cash flow last quarter, up 184% year over year. At the same time, it spent $28.5 billion on capital expenditures, leaving free cash flow negative by $5.4 billion.¹
That sounds like an AI infrastructure company beginning to grow into its capex. But one line in the cash-flow statement changes the interpretation.
Oracle recorded $11.363 billion of cash inflow from customer prepayments with a “significant financing component.”¹
That amount equals 49.2% of Oracle’s reported operating cash flow for the quarter. It also equals 39.9% of gross capital expenditure.
The important question is therefore not simply whether Oracle is generating enough cash to finance its AI build. It is whose cash is financing it.
Oracle’s cloud infrastructure business is scaling quickly. Q1 infrastructure-as-a-service revenue reached $7.4 billion, up a company-reported 121% year over year. Oracle added 850 MW of data-center capacity during the quarter and said it had delivered more than 300,000 GPUs to AI cloud customers since the end of Q4. Remaining performance obligations reached $664 billion, up $209 billion from a year earlier. Oracle also booked more than $30 billion of additional AI cloud contracts during Q1.¹
The architecture increasingly looks like this:
customer contract → customer prepayment → GPU procurement → data-center capacity → utilization → cloud revenue
That is different from the traditional version:
provider capital → infrastructure → customer demand → revenue → cash recovery
The difference is financing risk.
A customer that prepays for hardware is effectively moving cash toward the front of the infrastructure cycle. Oracle receives capital before all of the associated cloud revenue has been recognized. The customer obtains future compute capacity. Oracle reduces the amount of incremental external capital required to procure the equipment necessary to serve that contract.
Oracle has been explicit about this structure.
At the end of fiscal 2026, it said the prepaid and customer-supplied hardware portions of its large AI contracts totaled $75 billion. Oracle said this “substantially reduces” the amount of capital it must raise to build its AI data centers.²
The latest quarter extends that model. Oracle said the structure of more than $30 billion of new AI cloud contracts signed in Q1 would create no incremental impact on its existing capital-raising plans.¹
This matters because the broader AI infrastructure boom is becoming increasingly capital-intensive.
The Bank for International Settlements warned this week that capital expenditure by major AI firms is increasingly outrunning internally generated cash flows, pushing more of the investment boom toward debt and private credit.³
Oracle illustrates another route. The financing burden does not have to remain entirely with the infrastructure provider. Large customers can absorb part of it directly through prepayments or by supplying GPUs themselves.
That does not mean Oracle’s financing problem has disappeared. It still completed a $20 billion common-stock sale during Q1. Oracle had previously said it expected to raise approximately $40 billion through debt and equity financing in fiscal 2027.¹ ² And despite the customer financing, Q1 free cash flow remained negative $5.4 billion.
The customer-funded model therefore does not eliminate capital intensity. It changes its distribution.
That distinction also changes how Oracle’s operating cash flow should be read. The $23.1 billion figure is real GAAP operating cash flow. But economically, almost half of it came from a line item explicitly carrying a financing component through customer pre-payments rather than solely from cash generated by selling services already recognized as revenue.
The stronger signal may therefore be neither free cash flow nor capex alone. It may be customer-funded capex.
If AI infrastructure contracts increasingly arrive with prepayments, customer-supplied GPUs, take-or-pay commitments or other forms of upfront capital support, hyperscalers and neoclouds may be able to expand faster than their own balance sheets would otherwise permit.
But another risk appears in its place. Capital risk becomes increasingly linked to the financial strength, concentration and long-term demand commitments of the customers providing that funding.
The next question for AI infrastructure may therefore not simply be: How much capex can the provider finance?
It may be: How much of the capex can the customer be persuaded to finance first?
References
1. Oracle, Oracle Announces Q1 Results Driven by Triple Digit Growth in Cloud Infrastructure Revenues, September 10, 2026. Oracle Q1 FY2027 results
2. Oracle, Oracle Announces Record Q4 and FY 2026 Results Driven by Cloud Infrastructure & Cloud Applications, June 10, 2026. Oracle FY2026 results
3. Pablo Hernández de Cos, Bank for International Settlements, Artificial intelligence, growth and financial stability: challenges for central banks, September 10, 2026. BIS speech on AI investment and financial stability
4. Reuters, Oracle tops estimates as AI demand tempers cash-burn fears, September 10, 2026. Reuters coverage of Oracle Q1 results
Comments