Record bookings are not near-term revenue, and the spending needed to serve them lands first.
Oracle (ORCL) stock has fallen about 33% over roughly three months, while the S&P 500 edged higher, Amazon (AMZN) slipped 1.6% and Salesforce (CRM) rose 34%. The move is Oracle’s own, not the sector’s. What changed is narrower: what the AI build costs, who funds it, and how slowly the backlog behind it turns into revenue.
A $638 Billion Backlog On A Multi-Year Schedule
Revenue passed $67 billion for the first time in fiscal 2026, and remaining performance obligations closed the year at $638 billion, up 363%. The scale is real; the schedule is the catch. Management expects 12% of that backlog to become revenue within twelve months of the fiscal 2026 fourth quarter and another 34% over the two years after, with both rates expected to rise. The rest of the backlog sits further out.
One Year Of Building Costs More Than Fiscal 2026’s Entire Revenue
The spending does not wait. The guided net cash outlay for capital expenditures in fiscal 2027 is around $70 billion, more than fiscal 2026’s entire revenue, even against revenue growth guided at 34% in constant currency for fiscal 2027. Paying for it means raising about $40 billion in debt and equity, including a $20 billion at-the-market equity issuance, and gross margin is guided down again in fiscal 2027 as the data centers ramp. The question is whether the build can be funded without causing prolonged margin erosion.
Customers Have Already Contracted For Part Of The Hardware
That $70 billion is net of $20 billion to $25 billion of customer prepayments Oracle expects to collect, so reported capital spending runs higher than the cash Oracle fronts. Contracts that are either bring-your-own-hardware or prepaid now total $75 billion, at margins management says are no worse than the rest. The pressure lands on margin and share count, not on the order book.
The Older Businesses Kept Growing Through It
The businesses Oracle already had held up: cloud applications revenue grew 10% in the fiscal 2026 fourth quarter and the cloud database business grew 29%, with multi cloud inside it up 404%. Oracle AI Database@AWS is now available in 22 AWS Regions. None of that stopped the decline. A stock moving on one theme, whatever the rest of the mix is doing, is a different proposition from the Trefis High Quality Portfolio, which does not depend on the handful of largest technology names to produce its returns.
Megawatts, Not Bookings, Decide What Happens Next
The live question is not whether demand exists. It is whether capacity lands on schedule and infrastructure margin recovers, as management expects, once the data centers reach full contractual revenue. Oracle delivered more than 1.2 gigawatts across fiscal 2026 and guided fiscal 2027 first-quarter delivery approaching 1 gigawatt—which management characterized as nearly the previous four quarters combined. Fiscal Q1 2027 results on September 10 are the first read on whether that pace held. Shares sit at $150.85, between a 52-week low of $114.99 and a high of $324.63, and sorting a 33% three-month decline into opportunity or warning is what a dip-buyer’s framework is for.
Owning A Build-Out Means Owning Its Calendar
Oracle may well deliver the capacity and the infrastructure margin recovery it has guided to. Waiting on a construction calendar is a different exercise from a rules-based approach like the Trefis High Quality Portfolio, which spreads the judgment across many businesses instead of resting it on one build. That portfolio has a track record of outpacing the three major indices – the S&P 500, S&P Mid-cap, and Russell 2000.
