$664bn of Backlog, $11.4bn of Customer Prepayments (10 September 2026): Oracle's Q1 FY2027 Result — and Who Actually Funded the Capex

Oracle beat its own guide on every line that was guided. Revenue came in at $19.3 billion, up 30% against a 27% to 29% guide; non-GAAP EPS at $1.92 against $1.72 to $1.76; cloud infrastructure up 121% to $7.4 billion; and the backlog the whole stock trades on rose to $664 billion, above the $630.6 billion the market expected. The shares had fallen 5.38% into the print and rose 4.13% after it. But the number that answers the question this post was written to ask is buried in the cash flow statement: Oracle spent $28.5 billion of capital expenditure in three months and its total borrowings went down $4.2 billion. Customers and shareholders paid for the quarter. The bond market did not.

What actually happened

Oracle reported after the US close on Thursday 10 September 2026, covering the quarter that ended 31 August. Every guided line cleared its guide.

Consensus, per the LSEG and StreetAccount figures reported by CNBC, had been $19.14 billion of revenue, $1.74 of adjusted EPS, $11.51 billion of cloud revenue, $7.09 billion of cloud infrastructure and $630.6 billion of RPO. Oracle cleared all five. Guidance moved up with the result: Q2 revenue growth of 30% to 34%, non-GAAP EPS of $1.85 to $1.93, and a full-year outlook of at least $90 billion of revenue with $8.10 of non-GAAP EPS.

Two operating disclosures sit underneath those numbers. Oracle said it delivered 850 megawatts of additional datacentre capacity in the quarter, and more than 300,000 GPUs to AI cloud customers since the end of Q4 — almost triple the capacity delivered in the prior quarter. Capacity delivery, not contract signing, is what converts backlog into revenue, and this is the first quarter in which that delivery number is large enough to matter to the income statement.

The market reaction split across the close. The shares fell 5.38% during Thursday's regular session to $152.94 on 43.0 million shares, then rose 4.13% in extended trading to $159.26. That still leaves the stock roughly 51% below the $328.33 close of 10 September 2025 — the session that created the backlog narrative in the first place.

The backlog grew, and the headline percentage stopped working exactly as expected

Remaining performance obligations reached $664 billion, up $209 billion from $455 billion a year earlier and up $26 billion from the $638 billion reported at 31 May 2026. Oracle booked more than $30 billion of additional AI cloud contracts during the quarter.

Set the balance against the run rate and the scale is easier to hold. At $19.3 billion a quarter, $664 billion is about 8.6 years of revenue already under contract. That is the asset. The other side of the same sentence is that the capacity to serve it has to exist before any of it is recognised.

Who paid for the quarter

This is the part the headlines did not carry, and it is the most consequential thing in the release.

Capital expenditure in the quarter was $28,499 million, against $8,502 million a year earlier. Property, plant and equipment net of depreciation rose from $99,957 million at 31 May to $127,845 million at 31 August — nearly $28 billion of hard assets added in thirteen weeks. Reported operating cash flow was $23,103 million, up 184%, which produced reported free cash flow of about -$5.4 billion against -$362 million a year ago.

Read one line down in the cash flow statement filed with the 8-K and the composition changes the meaning. Inside operating cash flow sits an $11,363 million increase in deferred revenues from customer prepayments with a significant financing component — a line that was nil in the year-ago quarter. A separate $3,997 million increase in other deferred revenues sits alongside it. On the balance sheet, other non-current liabilities rose from $16,178 million to $28,183 million, and current deferred revenues from $9,916 million to $14,686 million.

In plain terms: customers paid Oracle in advance, in size, for capacity that has not yet been delivered, and the accounting for those prepayments recognises that they contain an embedded financing element. Strip that $11.4 billion out and operating cash flow for the quarter is roughly $11.7 billion, and free cash flow is roughly -$16.8 billion. Neither figure is wrong; they answer different questions. The reported number tells you what came through the door. The adjusted number tells you how much of the buildout the business generated on its own.

The rest of the gap was equity. Oracle completed the sale of $20 billion of common stock through an at-the-market programme, booking $19,909 million of net proceeds, and stockholders' equity rose from $43,056 million to $67,196 million. Add $19.9 billion of stock to $11.4 billion of prepayments and you have $31.3 billion of new funding against $28.5 billion of capex — which is why, in the heaviest capex quarter in the company's history, notes payable and other borrowings fell from $129,541 million to $125,337 million.

Chief executive Clay Magouyrk put the same point in one sentence on the call, quoted by CNBC: the company closed more than $30 billion of additional AI contracts in Q1 without requiring additional capital from Oracle. The release said the same thing in filing language — that based on the structuring of those new contracts, there is no incremental impact on its plans to raise capital.

What this does and does not fix

It changes who bears the risk in the near term. It does not change how much risk exists.

Funding capex with customer prepayments moves timing risk from Oracle's balance sheet to the customer's: the counterparty has already parted with the cash, and Oracle holds it as an obligation to be discharged in capacity rather than in dollars. Funding with equity moves the cost to existing holders through dilution rather than to bondholders through interest. Both reduce the sensitivity of the story to the long end of the Treasury curve, which is exactly where the term-premium move of August was reaching the shares.

What it does not fix is the depreciation schedule. Depreciation in the quarter was $3,156 million against $1,351 million a year earlier, up 134%, and that is the early part of a run that lengthens with every megawatt energised. Nor does it fix the interest already contracted: interest expense was $1,428 million in the quarter against $923 million, up 55%, consuming 8% of revenue against 6% a year ago. A quarter without new borrowing does not reduce the coupon on $125 billion of existing debt.

The concentration question is also unchanged, because an earnings release does not address it. What is disclosed, in the fiscal 2026 Form 10-K, is that no single customer accounted for 10% or more of total revenues in fiscal 2026, 2025 or 2024. What is inferred, from press reporting, is that a large share of the contracted backlog traces to a small number of very large AI agreements. Those statements describe different things — recognised revenue in the past, contracted obligations in the future — and only the first is a filing disclosure. The structure is the same one that sits underneath the residual-value guarantees in Nvidia's filings, and the opposite of the one that makes Broadcom's 1.8% capex intensity the useful contrast.

The number that still has not been published

This preview named one figure as the most checkable thing in the release: any revision to the RPO conversion schedule. It was not revised, because it was not in the release.

The fiscal 2026 10-K stated that of $638 billion of RPO at 31 May 2026, approximately 12% was expected to convert within twelve months, 34% in months 13 to 36, 34% in months 37 to 60, and the remainder thereafter. The Q1 earnings release restated none of that, and as of 11 September 2026 the SEC had received only the 8-K carrying the results — the Form 10-Q, where the updated schedule lives, had not been filed.

The right-hand column is arithmetic, not disclosure. It is useful only as a placeholder: roughly $80 billion of near-year conversion against a guide of at least $90 billion of revenue implies most of fiscal 2027 is already contracted, and roughly $452 billion still lands beyond year three. Whether the contracts signed this quarter pulled that mix forward or pushed it further out is the one thing the 10-Q settles, and it is worth reading when it appears.

What would change the picture from here

The sequential RPO line. $26 billion of net addition against $85 billion in Q4 and roughly $317 billion a year ago. That is the series that carries information; the year-on-year percentage is now structurally meaningless.

Whether the prepayments repeat. An $11.4 billion prepayment line appearing from nothing is the single largest change in the funding picture. If it recurs, the buildout is substantially customer-financed and the credit question softens. If it was a one-quarter structuring of specific contracts, reported operating cash flow falls back toward the roughly $11.7 billion the business generated without it, and the funding gap returns to the debt and equity markets.

Capacity delivery against contract signing. 850 megawatts and more than 300,000 GPUs delivered is the conversion engine. Chief financial officer Hilary Maxson told reporters that full-year capital spending guidance is unchanged, and said nothing known today suggests Oracle's New Mexico site or others are delayed relative to the schedules in the fiscal 2027 outlook — a response to reporting that a natural gas pipeline serving that site was running behind. Oracle is still working to obtain an air permit there.

The credit market's read, not the equity market's. Oracle's five-year credit default swap became the market's liquid expression of a view on the whole AI capital cycle during 2026. Barclays credit analyst Andrew Keches, quoted by CNBC in July, wrote that its appeal "has extended beyond company-specific fundamentals, reflecting its role as a liquid hedge on AI capex, OpenAI execution and broader data-center spending narratives." A quarter in which total borrowings fell is the first datapoint this year that cuts against that use of it.

None of that resolves how the shares trade from here, and it is not meant to. It is a map of where new information enters a story whose revenue line was published three months in advance and whose real variable was always the financing. How the eight-currency factor read is constructed is set out on the about page, and what the dollar's interest-rate factor is currently picking up from the long end is tracked on the USD currency page.

Educational macro context only — not investment advice.