Oracle says its OpenAI buildout is funded. What will it earn?
Oracle says the money for its OpenAI contract is in place. Its company-wide cash flows show why funding and earning a return are different tests.
5 min read
Oracle says it has found the money. Now it has to earn the return.
That is the useful distinction behind its unusually blunt statement on funding its OpenAI contract. On 20 November, the company said the financing it needs is in place, including vendor financing, and further private discussions with suppliers are optional. It also reiterated its expectation of positive company-wide free cash flow in the second half of its 2028 financial year, with a target of more than $30bn in 2029.
For shareholders, those are two different promises. The first concerns having enough cash and financing to build. The second concerns what is left after running and expanding the business. A convincing answer to the first deserves attention. It does not automatically answer the second.
The interesting Oracle question is therefore moving beyond whether someone will finance the data centres. It is whether the income those facilities earn will justify the cost of the hardware, buildings, power and capital required to deliver it.
A very large cheque, with a very large bill beside it
Oracle's first-quarter results showed a business expanding quickly: revenue reached $19.3bn, with cloud infrastructure revenue up 121% to $7.4bn. That is substantial trading activity, rather than a promise that customers might eventually appear.
The company-wide cash-flow statement shows the scale of the spending. In the three months to 31 August 2026, operating cash flow was $23.103bn. Capital expenditure was $28.499bn. Subtract one from the other and Oracle's reported free cash flow was negative $5.396bn. These figures cover the whole business; they do not isolate the cash flows of the OpenAI contract.
The comparable quarter a year earlier had a much smaller gap: $8.140bn came in through operations against $8.502bn of capital expenditure. Both numbers have grown dramatically, but capital spending has pulled further ahead.
Oracle’s capital spending pulled further ahead of operating cash flow
Oracle group, first fiscal quarter; US$bn
Across Oracle, operating cash flow rose from $8.140bn to $23.103bn and capital expenditure from $8.502bn to $28.499bn, comparing the quarters ended August 2025 and August 2026.
Cash generated and capital spending
US$bn
- Q1 FY2026
- Q1 FY2027
Company-wide figures for quarters ended 31 August. Capital expenditure is shown as a positive spending amount. The 2026 operating cash flow includes $11.363bn of increased deferred revenue from customer prepayments with a significant financing component.
Source: Oracle Q1 FY2027 financial tables, pages 8-9
View data
| Category | Q1 FY2026 | Q1 FY2027 |
|---|---|---|
| Operating cash flow | $8,140,000,000 | $23,103,000,000 |
| Capital expenditure | $8,502,000,000 | $28,499,000,000 |
There is another important detail in the financial tables: the latest operating cash flow included $11.363bn of increased deferred revenue from customer prepayments with a significant financing component.
That is real cash. It is also money received before the corresponding revenue is earned, with services still to be delivered. Treating all of it as cash produced by this quarter's completed work would give the wrong impression.
Nor should it simply be thrown out as worthless. A customer willing to pay early is helping finance capacity and showing commitment. The question is what Oracle has promised in return, when it must deliver, and how much cash those obligations will consume.
The financing mix changes who carries the risk
At its October investor day, Oracle described vendor financing, customer-provided hardware and customer prepayments as ways to reduce upfront cash needs. It expected peak capital expenditure in financial years 2027 and 2028, followed by a step down in 2029.
Those arrangements can make a huge project easier to fund. They do not make the equipment free.
A supplier who finances hardware is extending payment terms or credit. A customer bringing its own hardware changes what Oracle needs to buy and what service it earns money from. A prepayment moves cash earlier. Each can improve the timing of a project without proving that the underlying investment earns an attractive return.
Oracle's own capital-spending reconciliation makes the distinction concrete. It reports a separate $17.966bn net cash outlay for capital expenditure. Starting with the $28.499bn of capital expenditure, that measure deducts $11.363bn of financing-related customer prepayments and adds $830m of net short-term financing repayments.
This measures the funding requirement. The reported free-cash-flow calculation still uses the $28.499bn capital-expenditure figure. Substituting the lower number while keeping the prepayments inside operating cash flow would count the same benefit twice.
The financing announcement should reduce uncertainty about the current plan. The operating accounts will have to demonstrate whether that plan produces enough cash for everyone with a claim on it.
This is why looking beyond the headline debt figure matters. A comfortable cash balance and a funded investment programme can coexist with substantial future commitments. Payment dates, contract terms and the cost of refinancing tell us more than a reassuring adjective.
A backlog is valuable, but it is not a bank balance
Oracle reported $664bn of remaining performance obligations in its first-quarter release. This is contracted revenue still to be recognised, including work customers may have paid for in advance. The total does not tell us how much cash is still to arrive.
Its quarterly filing puts a timescale on that backlog: Oracle expected to recognise about 13% as revenue in the twelve months after 31 August 2026. Most of the delivery schedule stretches further out. The headline total also says little about the margin Oracle will earn on that work.
There is a chain between signing a contract and earning a satisfactory return. Capacity must be delivered, the service supplied, costs controlled and money collected under the agreed terms. Even a profitable contract can consume cash before it produces it.
For an investor, the useful follow-up is therefore the pace of conversion. Does additional capacity become revenue promptly? Does that revenue bring cash on terms that support the next stage of construction? How much spending is needed to maintain the service once the initial buildings are complete?
These questions also belong in the wider debate about financing AI infrastructure. Making an asset easier to finance can accelerate deployment. It cannot, on its own, establish what the asset will earn over its useful life.
The strongest case for Oracle is a timing argument
The bullish reading deserves a fair hearing. Building capacity requires money before that capacity can earn revenue. Negative free cash flow during construction is not, by itself, evidence of a bad investment. Refusing every project that depresses near-term cash generation would rule out plenty of worthwhile expansion.
Oracle's case is that the expensive construction phase gives way to a larger business that can fund its own growth. Its published timetable gives shareholders something concrete to judge. If spending moderates while the new capacity keeps earning, the cash-flow picture can improve considerably.
The tougher question is how much of the eventual cash generation depends on spending falling. A result driven by profitable growth and a result driven by delaying necessary investment can look similar for a quarter. Over several years, they are quite different propositions.
There is also a difference between a good project and a good share purchase. Even a successful buildout may already be reflected in what investors pay. Financing certainty can reduce one risk without making every valuation sensible.
What the next set of accounts needs to show
I would put the operating cash flow, customer prepayments and capital expenditure beside one another each quarter. That makes it harder to mistake a change in funding timing for an improvement in the economics of the business.
Then I would follow how quickly the new capacity earns revenue, and whether the promised reduction in spending arrives without weakening delivery. These are more useful tests than deciding that every funding announcement proves the boom is either secure or doomed.
Oracle's statement answers an important objection to its plans. The money to build, it says, is there. The accounts now have to show what building buys.
Comments
No comments yet - start the conversation.