Oracle's Cloud Business Is Doubling. The Cost of Delivering It Is the Question

Oracle reported its latest quarter and its cloud infrastructure revenue more than doubled, and the stock jumped. On the surface that is the whole story — an earnings beat on a demand boom, the kind of print that makes a growth story look self-evident. The numbers do read that way: revenue of about $19.3 billion ahead of expectations, adjusted earnings per share of $1.92 above the $1.74 analysts were looking for, and the engine doing the work is the business investors actually care about.

But Oracle is not a simple growth story anymore, and its own stock price is the proof. The shares still trade near $153 — roughly half of the $329 they touched within the last year — and the company that once carried a market value above $900 billion has shed hundreds of billions of that along the way. Here is the tension worth sitting with: a business whose fastest-growing revenue line more than doubles, and that still trades far below its high, is not being doubted for its demand. It is being priced for what meeting that demand costs.

The doubling is real, and it is the engine

Start with what is actually happening operationally, because it distinguishes this from a roadmap promise. In Oracle's fiscal fourth quarter of 2026, cloud infrastructure revenue — the raw compute and storage customers rent, known as IaaS — hit $5.8 billion, up 93% from a year earlier. A year before that the same line was $3.3 billion. In the quarter Oracle just reported, analysts expected that infrastructure business to land near $7.2 billion, or roughly double its year-ago level. This is a segment doubling off an already-doubled base, and it is recorded revenue, not a plan.

The distinction matters. Oracle has a second cloud line, cloud applications (SaaS — finished software like its ERP and NetSuite), which grew only around 10%. The doubling is entirely an infrastructure story. In the fourth quarter, cloud revenue as a whole (infrastructure plus applications) grew 47%, with infrastructure supplying nearly all of the acceleration while the software side barely moved. That is the shape of a company that has become, more than anything else, an AI infrastructure builder.

The backlog is both a promise and a bill

The metric that explains how a business can grow like this is Oracle's remaining performance obligations, or RPO — the dollar value of signed contracts not yet recorded as revenue. It is, in effect, an order book: money customers have already committed to pay. That backlog has gone from $455 billion a year ago to $553 billion to $638 billion by the end of the fourth quarter — roughly ten times Oracle's annual revenue.

This is the number that carries the whole debate, and it pulls in two directions at once. On one side it is the strongest possible evidence of real, committed demand: this revenue is booked, not hoped for. On the other side, every dollar in that backlog is also a delivery obligation. Customers signed up expecting capacity — data centers, power, and AI chips. The RPO surge is simultaneously a demand signal and a fixed bill of what Oracle has to build.

That dual nature is the entire investment question in one figure. A backlog that big does not arrive by itself; it arrives the way every data-center build does, with years of capital spending ahead of the revenue that pays for it.

What the growth costs shows up in the cash flow

Here the financial statements become the honest witness. Over the trailing twelve months Oracle has spent roughly $56 billion on capital expenditures while generating about $32 billion in operating cash flow — leaving free cash flow around negative $24 billion. Free cash flow is simply the cash a company is left with after paying to maintain and grow its business, and for Oracle today it is negative: it is spending more on capacity than the operation throws off.

The gap is being funded on the balance sheet. Oracle carries total debt of about $219 billion and roughly $98 billion in net debt, with leverage well above what a software company historically ran. Management has also described a sizable 2026 financing package to keep the buildout going, part of it through debt and part through equity-linked instruments that dilute existing shareholders. The company's path is coherent — convert the backlog into revenue, bring the capacity online — but the compound is powered, for now, by cash flow that does not cover the bill and a balance sheet that is taking on the load.

None of this makes the demand fake, and it is worth saying clearly: the more-than-doubling is achieved, operating, and reflected in the backlog. The mistake would be to read a 93%-or-better growth line as proof that the economics have caught up. They have not yet. The cost shows up in the cash flow statement long before it shows up in revenue.

What that leaves the reasonable investor to judge

After the drawdown, the stock no longer carries the nosebleed multiple it once did — it trades around 6.5 times trailing revenue rather than the double-digit figure at the peak. A big share of the delivery and leverage risk has arguably already been priced out of the share price. That makes Oracle less a "is this real" question than a sequencing question.

The judgment I keep coming back to is about pace, not direction. Demand is not the issue. The issue is whether Oracle converts that $638 billion backlog into revenue and margin faster than the debt and dilution it takes to build the capacity grow. If delivery outruns financing, the current price has a lot of room to catch up to the business. If the financing costs and dilution outrun conversion — especially if the AI demand that justified the buildout stalls — the share count and the interest bill compound against the equity.

For the beginner's frame, keep it simple: Oracle has proven it can grow an AI infrastructure business faster than almost anyone. What it has not yet proven is that it can deliver that growth without the balance sheet eating the gains. That is the specific, open question this earnings report leaves on the table — and it is a better lens for the stock today than the headline about a jump on a beat.