Oracle (ORCL): Backlog Hits 664bn Dollars as Cloud Revenue Jumps

Oracle ended its fiscal first quarter with remaining performance obligations of $664bn — up $209bn year on year, and well ahead of the roughly $630.6bn the market expected — while guiding full-year fiscal 2027 revenue to at least $90bn. Those two figures sit next to each other in the same release. The backlog is more than seven times the revenue it is meant to start converting into. Shares rose about 7% in extended trading on 10 September.

Every other number in the print is downstream of that ratio. The question Oracle has handed investors is not whether demand exists — $664bn settles that — but what has to be built, financed and collected before any of it becomes revenue, and who is on the other side of the contracts.

What actually happened

Reported after the US close on 10 September 2026, fiscal Q1 2027 revenue was $19.35bn, up around 30%, against a $19.14bn consensus. Adjusted earnings per share of $1.92 beat the $1.74 expected — a wide margin by Oracle's standards.

The cloud lines carried it. Total cloud revenue, combining infrastructure and applications, rose 62% to $11.6bn. Within that, Cloud Infrastructure grew 121% to $7.4bn and Cloud Applications grew 10%. Oracle said it delivered 850MW of additional datacentre capacity in the quarter — the physical counterpart to the infrastructure number, and a useful reminder that this revenue line is manufactured, not merely sold.

The backlog is the headline. RPO of $664bn was helped by more than $30bn of new AI cloud contracts signed during the quarter. For context on the trajectory: RPO stood at $638bn at the close of fiscal Q4 2026, having grown 363% year on year at that point. The step to $664bn is smaller in percentage terms precisely because the base has become enormous.

Guidance moved up. Oracle now expects fiscal 2027 adjusted earnings per share of $8.10 on at least $90bn of revenue, against an LSEG consensus of $8.07 and $89.76bn. That $8.10 is a raise from the $8.05 guided alongside the fiscal 2026 results in June. Next-quarter revenue growth was guided to 30-34%.

The number that was not in the headline

Capital expenditure is the other half of this business and it is the half that determines whether the backlog is an asset or an obligation. Oracle spent $55.7bn on capex in fiscal 2026, up from $21.2bn in fiscal 2025 — a rise of roughly 162% in a single year. For fiscal 2027 the guided gross figure is $90bn-$95bn, with $20bn-$25bn expected to be reimbursed by customers, leaving a net outlay of around $70bn from Oracle's own resources. The company has indicated plans to raise approximately $40bn through debt or equity financing to support it.

Set that against fiscal 2027 revenue guided to at least $90bn and the shape of the year becomes clear: gross capital spending is running at roughly the scale of total revenue.

Why the market reacted the way it did

A 7% after-hours rise on a double beat and a guidance raise is not surprising in isolation. What makes it notable is where the shares were standing when the numbers landed. Oracle went into this print having fallen heavily across 2026, well below the peak it reached in September 2025, on precisely the concerns this release had to answer. The move reads less as enthusiasm for the quarter than as relief that the conversion story did not break.

Three things in the release did the work.

Infrastructure growth accelerated rather than decelerated. Cloud Infrastructure at plus 121% is the line that has to keep compounding for the backlog to be credible. A deceleration here would have implied capacity or demand constraints; there was none visible.

The backlog beat a raised bar. Consensus had moved to around $630.6bn. Coming in at $664bn, with over $30bn of new AI contracts signed in a single quarter, rebuts the argument that the bookings surge was a one-off event concentrated in a handful of deals signed last year.

Earnings guidance went up rather than being held. This is the subtler signal. A company spending $90bn-$95bn gross on capacity faces rising depreciation, and rising depreciation is the mechanism by which an infrastructure build quietly erodes earnings. Raising the adjusted EPS guide to $8.10 says management believes it can absorb the depreciation step while the mix shifts towards lower-margin infrastructure. That claim will be tested every quarter for the next three years.

The Openbook read

Growth is the least ambiguous factor in the set and it is exceptional. Revenue up 30%, infrastructure up 121%, total cloud up 62%, next quarter guided to 30-34%. For a company of Oracle's age and scale these are not normal numbers. The one qualification worth holding: growth of this kind, sourced from a small number of very large contracts, has a different quality from growth sourced from thousands of renewals. It is faster and it is also more concentrated.

Profitability is the factor under the most genuine pressure, and it is where the mix shift bites. Infrastructure is a capital-intensive, lower-gross-margin business than the licence and applications revenue that historically defined Oracle. Cloud Infrastructure at $7.4bn of $11.6bn total cloud is already close to two-thirds of the mix, and it is growing at 121% against applications' 10% — so as that gap compounds, the blended gross margin faces structural downward pressure, and the depreciation from $55.7bn of fiscal 2026 capex is only now beginning to flow through the income statement in full. That the company still raised adjusted EPS guidance is the positive datum. The distinction between adjusted and reported profitability matters more than usual for a business in this phase, and the two will diverge as capacity is commissioned.

Solvency is the factor that has genuinely moved, and it is the one this whole story turns on. Net capital spending of roughly $70bn in a single year, against a plan to raise around $40bn in external financing, is a balance-sheet event, not a capex line item. The rating agencies have already acted: S&P Global Ratings lowered Oracle's long-term issuer credit rating to BBB- from BBB in July 2026, with a stable outlook. That is one notch above sub-investment grade. S&P singled out customer concentration as the vulnerability — specifically OpenAI, which by its analysis accounts for roughly half of Oracle's backlog and more than $300bn of RPO. The agency's concern was concrete rather than abstract: if the counterparty cannot meet its commitments, Oracle is left holding long-dated datacentre leases with no easy exit. OpenAI is privately held, has not generated a profit, and consumes cash at scale. The solvency score has to reflect that Oracle is funding hard assets with borrowed money against contracts with a counterparty whose own funding is not assured.

Momentum improves from a low base. A 7% extended-hours move on results, after a year of heavy underperformance, is a reversal signal rather than a trend. Momentum measures on Oracle will stay weak until the recovery is measured in months rather than a session, and the next scheduled test is the December quarter.

Reward/risk is unusually wide and unusually well-defined, which is rarer than it sounds. The upside case does not require heroic assumptions: the contracts exist, the capacity is being delivered at 850MW a quarter, and $90bn of fiscal 2027 revenue is guided rather than hoped for. The downside case does not require a demand collapse either. It requires only that one very large counterparty renegotiates, delays or fails — at which point Oracle owns the datacentres, owes the debt, and has lost roughly half the backlog. That asymmetry, rather than any valuation multiple, is what the five-factor profile is actually describing. Investors wanting to compare the capital-intensity and solvency profile against other large-cap infrastructure names can do so through the Openbook screener.

The read-across

For the hyperscalers, Oracle's print is confirmation that AI infrastructure demand has not cooled and that capacity, not appetite, is the binding constraint. Microsoft, Amazon and Alphabet are all committing capital on a similar logic, though each does so from a vastly stronger cash-generation position and none carries anything resembling Oracle's single-customer concentration. The correct read-across is to demand, not to balance sheets.

For the supply chain, 850MW of delivered capacity in one quarter and $90bn-$95bn of gross fiscal 2027 capex is an order book. Nvidia is the obvious beneficiary, but so is everything between the grid and the rack: power equipment, cooling, optical networking and construction. Investors holding the picks-and-shovels names should note the distinction in timing — Oracle's spend is being committed now, while much of the associated revenue at component vendors converts across 2027.

For anything OpenAI-adjacent, the concentration cuts both ways. Oracle is now among the most direct listed expressions of a view on OpenAI's ability to fund its commitments, alongside CoreWeave. A single funding event at that counterparty repricing several listed securities simultaneously is the sort of correlation that is invisible in a sector screen and very visible in a drawdown.

What to watch next

• The $40bn financing. Terms, size and mix between debt and equity are the most immediate disclosure. Pricing on any new issue tells you what the credit market thinks of the concentration risk S&P flagged — independently of what the equity market thinks.
• Fiscal Q2 results, due in December. The specific line to watch is not revenue but the RPO figure and the rate at which it converts. Backlog that grows while conversion stalls is a warning; backlog that holds while revenue accelerates is the thesis working.
• Capex against the $90bn-$95bn gross guide, and in particular whether the $20bn-$25bn of expected customer reimbursement actually arrives. That reimbursement is the difference between roughly $70bn net and roughly $95bn net, and it is a cash-flow item, not an accounting one.
• Gross margin disclosure as infrastructure extends its lead over applications in the mix. This is the Profitability factor's single test and it resolves quarter by quarter.
• Any further rating agency action, and any disclosure that changes the concentration picture. Contracts signed with counterparties other than the largest one would do more for the solvency read than another record RPO number.