Oracle's AI Demand Is Real. So Is the Cost of Paying For It.

Oracle just reported the kind of cloud quarter most companies would frame and hang on a wall. Cloud-infrastructure revenue jumped 121% to $7.4 billion, total revenue rose 30% to $19.35 billion, and management raised its full-year outlook to at least $90 billion. The stock rose about 7% after the close yesterday. Then, today, it gave the gain back and dropped around 5%. This is the second straight report in which Oracle beat expectations on AI growth and fell on the bill for it.

The AI Cycle Decoder's first question is always the same: is this really a demand story, or a financing story wearing demand's clothes? The answer here is both — and which side you weight is the entire investment question.

The demand is genuinely there

Start with what the market is not disputing. Oracle's cloud infrastructure, the GPU-heavy business at the center of its AI bet, more than doubled to $7.4 billion in the fiscal first quarter, and total cloud revenue grew 62% to $11.6 billion. Adjusted earnings of $1.92 beat the $1.74 the Street expected, and revenue came in about $200 million ahead. Management did something software vendors rarely do with a straight face now: it guided the full year to at least $90 billion of revenue and $8.10 in adjusted earnings per share, above consensus.

The forward order book is the loudest signal of all. Oracle's remaining performance obligations — the contracted revenue it has not yet recognized — reached $664 billion, up $209 billion from a year earlier and well above the roughly $631 billion analysts had modeled. The company booked more than $30 billion of new AI cloud contracts in the quarter. That is not a company making AI claims. That is a company with decades of committed demand on its books.

Nobody should look at a business growing revenue 30%, with the AI line doubling, and call demand the problem.

The market is pricing the bill, not the growth

The problem is what that growth costs to deliver, and the evidence is in the cash flow. Oracle spent $28.5 billion on property, equipment and data centers in a single quarter — up from $8.5 billion a year earlier — and it intends to spend $90 billion to $95 billion across fiscal 2027. The quarter's free cash flow came in negative, at about $5.4 billion, and the full-year plan implies roughly a $45 billion free-cash-flow deficit.

That is the crux, and it reframes the beat. Earnings per share can look fine while the company's cash goes negative, because the spend is capitalized on the balance sheet and depreciated slowly rather than expensed immediately. The gap between a growing backlog and the cash it consumes today is exactly where the real economics live.

To fund it, Oracle is using leverage, not just its own profits. It carries on the order of $125 billion in debt, and it is selling equity to pay for capacity — a $20 billion at-the-market share sale during the fiscal first quarter, following a $40 billion raise it announced in June. Dilution is not fatal for a company whose market is expanding this fast, but this is not a costless choice. Every piece of new capacity is being paid for with debt and new shares, and the repayment has to come from the revenue that capacity eventually generates.

The backlog is a promise, not yet a payment

There is a quieter number under the headline backlog worth weighting. RPO reached $664 billion, but it added only roughly $26 billion in the latest quarter after adding about $85 billion in the prior one. The backlog is increasingly a stock of commitments made in earlier quarters — the massive OpenAI-linked deals signed in the last two years — rather than fresh additions each quarter. As Oracle recognizes that contracted revenue, the backlog only grows as fast as new bookings can be won.

This is the dual signal at the heart of the whole setup. A $664 billion backlog is proof of durable demand. It is also a load of obligations Oracle must fund, deliver and support on massive scale, with the clock running and the economics still unproven at the cash level. Demand and leverage are the same contract, seen from two sides.

What separates the story from the verdict

Oracle has converted its AI bet into something real, and that growth is the part of the thesis that has survived. But the market is no longer paying Oracle for the demand — it is paying for the delivery risk, and it has been for a while. The stock is down roughly half over the past year and about a fifth year-to-date even as revenue accelerated, because investors keep asking the one question that matters: when does this backlog turn into cash flow that covers a $90 billion-plus annual build?

A selloff on a genuinely strong report is often a buying opportunity rather than a verdict. This one is different because it has a real driver — the cash burn and the dilution that fund it — not just a mood. The durable case rests on when that conversion happens: whether delivered cloud revenue and margins grow fast enough to close a roughly $45 billion free-cash-flow hole before the leverage gets more expensive and the share count keeps climbing. That is the fact that separates a company compounding its way through a dark-horse buildout from one funding its growth a step ahead of the revenue. The demand is proven. The payout is the open question.