Oracle booked more than $30 billion of additional AI cloud contracts in its fiscal first quarter, taking its remaining performance obligations to $664 billion against analyst estimates of $639.89 billion. The shares rose about 5.5 percent premarket on Friday and roughly eight percent in extended trading after the release, adding around $24 billion of market value. Total revenue was $19.3 billion, up thirty percent, with cloud infrastructure revenue up 121 percent to $7.4 billion. The company nudged its fiscal 2027 adjusted earnings forecast to $8.10 a share from $8.05.
The backlog is the headline and the backlog is the thing to be careful about, and both of those statements are true at once.
Remaining performance obligations means contracted revenue that has not yet been recognised. It is work customers have committed to buying and Oracle has committed to delivering. It is a queue. It is not money, it is not margin, and it is not in the building. The number went up $209 billion year on year, which is a genuinely extraordinary rate of order capture and tells you the demand question is answered.
The question that is not answered is conversion. J.P. Morgan’s analysts laid it out plainly: whether the backlog growth is sustainable, whether it converts into revenue given data centre delays, and whether more capital raises are needed to build the capacity that the conversion requires. Oracle’s shares were down more than twenty-one percent this year before this, against an S&P 500 up nearly eleven, precisely because the market has been worried about debt-funded building running ahead of the cash to service it. Lale Akoner of eToro put the whole problem in one line: Oracle’s difficulty has never been finding customers, it has been proving the build-out will eventually throw off enough cash to justify the cost.
So, a cafe.
Imagine walking into a small shop and finding four hundred dockets on the order rail. Every one is a real customer who has really paid. The owner is delighted, and should be, because that is the hardest part of the business solved. Now look at the counter. One machine. Two people. A grinder that does one dose at a time.
The dockets are a claim on that machine’s future output. They are only worth what the machine can produce, when it can produce it, and every hour the rail grows faster than the machine runs, the value of each docket quietly falls. To fix that you buy a second machine, which costs money you do not have yet, which you borrow against the rail. The rail is now collateral for the equipment that exists to serve the rail.
This is Oracle’s exact position, at a scale of hundreds of billions, and it is why $664 billion is simultaneously the most impressive and the most demanding number in enterprise technology. Every dollar of it is an obligation in both directions.
There is one figure in the release that cuts through the argument better than the backlog does. Cloud infrastructure revenue grew 121 percent. That is recognised revenue, not a promise, and it is the part of the rail that has actually been made and handed across the counter. Triple-digit growth in delivered capacity is the evidence that the machine is getting faster, and it is what the stock moved on as much as the order book.
The stock trades at 16.86 times forward earnings against Microsoft at 23.84 and Amazon at 22.58, which tells you the market is still pricing in the conversion risk rather than assuming it away.
Four hundred dockets. One machine, running hot.
Source: Reuters, “Oracle shares rise as AI cloud backlog beats estimates”
