Oracle's IaaS grew 121% but the stock is down 50%: is a re-rating priced in or still blocked?

Oracle just reported one of its strongest quarters ever, and the stock still fell. For the quarter ended August 31, revenue rose 30% to $19.3 billion, cloud infrastructure revenue grew 121%, and the backlog hit a record $664 billion. That is the kind of quarter a beaten-down AI name is supposed to rally on. Instead the shares are down about half from their September peak, and extended the move lower after the print. The gap between the numbers and the tape is not noise. It is the whole question: is the market ignoring an improving story, or rationally pricing a risk that the headline growth cannot fix?

The backlog is real. That is not the issue

Start with what is actually working, because it is genuinely working. Remaining performance obligations — the value of contracts Oracle is still obligated to deliver on — reached $664 billion, up $209 billion in a year, and the company booked more than $30 billion of new AI cloud contracts in the single quarter. OpenAI alone signed on for roughly $300 billion of computing over about five years, and Oracle's cloud applications and infrastructure together now pull in $11.6 billion a quarter, 60% of total revenue.

This is not a beaten-down-stock recap selling hope. There is a genuine, contractually documented growth engine here. IaaS growth of 121% on a base that large is meaningful. The operating story has clearly improved, not broken.

The discount is carrying a cash-flow problem

So why is the market refusing to pay up? Because none of that backlog is cash yet, and the cash-flow path is moving the wrong way. This is the one thing my discipline treats as the hard proof of a story: real free cash flow. Oracle does not have it, and it is getting thinner, not fatter.

In the quarter, Oracle spent $28.5 billion on capital expenditures, up from $8.5 billion a year earlier. Free cash flow came in at negative $5.4 billion — a swing of roughly $5 billion in the wrong direction. Across all of fiscal 2026 the company already blew past its own capex plan, spending $55.7 billion against a $50 billion target. The pattern is an overrun, then an even bigger plan, then another overrun.

The money for that buildout has to come from somewhere, and that is the second part of the problem. Oracle sold $20 billion of new stock this quarter alone and carries on the order of $125 billion of debt, with more funding planned. Operating cash flow did hit a record $23 billion, but a good share of that is customer prepayments rather than repeatable cash generation, and capex still swallowed it. In fairness, management discloses a net cash outlay for capex of about $18 billion after customer prepayments and short-term financing — the cash burn is not quite as bad as the GAAP number implies. But the direction is unmistakable: this buildout is funded by dilution and debt today, betting that the backlog converts into cash later.

Then there is concentration. The denominator of that $664 billion is heavily one customer. OpenAI has been the engine of the backlog, and OpenAI's own expansion plans are not settled — it has already pulled back from expanding Oracle's flagship Stargate data center site, chasing newer chips instead. A backlog concentrated on a customer whose plans are in motion is not the same thing as durable, diversified cash flow.

A fair price, not a re-rating opportunity

Put the pieces together and the human error to avoid is treating the big number as the answer. The $664 billion backlog is real growth, but the market's roughly 50% haircut is pricing the part that actually controls the outcome: whether that backlog turns into free cash. Right now it is not, and the multiple is closer to fair than to a gift.

At the current stock price around $153, against Oracle's own fiscal 2027 non-GAAP earnings guidance of $8.10 a share, the forward multiple is roughly 19 times earnings. That is not expensive for a 120%-growing cloud segment — and it is not cheap for a company whose headline cash-flow metric is negative and widening. The depressed multiple here is not an opportunity by itself; it can reflect the permanent deterioration the market is worried about, and the worry is concrete, not emotional.

This is the honest version of my normal setup, and I can be wrong again. My usual setup is a beaten-down company where expectations have reset while profitability is quietly improving underneath. Oracle has the reset. It does not yet have the improving free cash flow. The market is pricing the old risk profile — the financing and execution risk — and the cash-flow path has not turned clean enough to justify paying more.

What would flip it

A re-rating here is not impossible; it is just not earned yet. The mechanical way the multiple compresses is for the backlog to convert: capital spending flattening toward the cash that operating generation pulls in, until free cash flow inflects from negative toward positive. That is the one variable the case turns on, and it is observable every quarter in the gap between operating cash flow and capex. Cash flow must show up, not just revenue.

The break condition is equally specific. If the quarterly cash burn widens further, or conversion keeps undershooting even as the backlog grows, the discount is deserved and the thesis fails. And a rally that arrives without that conversion evidence would be a tape move, not a thesis — the setup would be a trap, not an opportunity. Until the cash-flow path turns, the honest read is that the re-rating is blocked, not priced in.