A backlog figure is supposed to be the dull footnote of a quarterly report, the line item accountants care about and everyone else skips. For Oracle, it has become the headline. When the company reports results, the first thing traders scan is not earnings per share or even cloud revenue growth, but remaining performance obligations: the contracted revenue customers have signed up for that Oracle has not yet recognized.
The stock now moves on that number more than on almost anything else.
The reason is straightforward. Oracle’s backlog swelled into the hundreds of billions of dollars as AI developers signed multi-year commitments for computing capacity, and investors have adopted it as a stand-in for how much of the AI infrastructure buildout Oracle will capture. A company once known for databases and enterprise software licenses is being valued like a utility-scale infrastructure builder. That makes the backlog less an accounting detail than a referendum on whether the story is real.

What the Backlog Actually Measures
Remaining performance obligations, or RPO, capture revenue that is under contract but not yet earned. If a customer signs a five-year agreement for cloud capacity, the unrecognized portion lands in the backlog and gets drawn down as Oracle delivers the service. Software companies have long reported this figure because subscription revenue is predictable, and a rising backlog usually signals healthy demand. What has changed is the scale and the type of customer. A few enormous contracts can now move the total by amounts that used to represent an entire year of company revenue.
Oracle’s pitch to AI developers is speed and flexibility. It has been willing to build dedicated clusters of GPUs, wire them into high-performance networking, and tailor capacity to the demands of a single customer, an approach larger rivals have sometimes been slower to match. Reports around the Stargate project and an enormous commitment from OpenAI put Oracle at the center of the largest data center build in the industry. Whatever the final terms turn out to be, the strategy has worked as a sales engine: the company has signed deals that would have sounded implausible for a database vendor a few years ago.
Wall Street loves the backlog because it offers something rare, which is visibility. Current revenue tells you what a company did last quarter, while a backlog tells you what it has already been promised for years afterward. Valuation models run on growth expectations, and a contracted number lets an analyst build a forecast that looks like arithmetic instead of speculation. It also lets a stock trade on revenue that does not yet exist. That is exactly why the figure draws so much scrutiny, and why a disappointing print can hurt as much as a strong one can lift.

The Gap Between Signed and Delivered
A backlog is a promise, not a payment. To turn those contracts into revenue, Oracle has to build data centers, secure enormous amounts of electricity, and buy chips and networking gear well before the first dollar of rental income arrives. That spending shows up as capital expenditures that dwarf anything the company’s software business ever required, and it has pushed free cash flow into uncomfortable territory while the company leans on debt markets to fund the construction. Margins matter here too. Renting out GPU capacity is a different business from selling high-margin software, and the profit on each dollar of backlog depends on costs that are still being worked out.
Concentration is the second pressure point. When a handful of customers account for a large share of the backlog, the number is only as sturdy as their ability to keep paying. Many of the biggest AI developers are funding their commitments with capital raised from investors, not from profits, so their spending depends on funding conditions that can change quickly. The full terms of these contracts are not public, so outsiders cannot easily judge how firm the commitments are, how they ramp over time, or what happens if a customer wants to renegotiate. A backlog built on a few giant agreements looks very different from one spread across thousands of mid-sized buyers, even if the headline figure is identical.
The smarter way to read the number is alongside three others: how quickly backlog converts into reported cloud infrastructure revenue, how much capital spending the company guides to, and what happens to gross margin as new capacity comes online. If conversion keeps pace with construction, the backlog looks like a runway. If revenue lags while borrowing climbs, it starts to look like a bill coming due. A rising RPO figure answers only the first question, which is whether customers want to sign. It says very little about whether Oracle can deliver at a price that leaves something for shareholders.

A backlog is a promise to pay, and Oracle has collected an enormous stack of them, but promises are only as good as the checkbooks behind them, and the buildings meant to fulfill them are not finished yet.
Frequently Asked Questions
What is Oracle’s remaining performance obligations (RPO) figure?
RPO is contracted revenue that customers have committed to but Oracle has not yet recognized. It is reported each quarter and acts as a gauge of future cloud revenue.
Why does Wall Street focus on Oracle’s backlog?
It offers visibility into years of expected revenue tied to AI infrastructure demand, so investors can model growth before it appears in reported results.
What are the main risks behind Oracle’s backlog?
The backlog depends on Oracle building data center capacity on time and on a small group of large customers continuing to pay. Heavy capital spending and borrowing add pressure.






