Oracle's Stock Fell 57%. Its Contracted Revenue Rose $85 Billion in a Quarter.

Oracle's stock is down roughly 57% from its September 2025 peak and about a third over the past year, trading near $149. Wall Street's explanation is simple and partly true: Oracle is pouring tens of billions into AI data centers, free cash flow has gone negative, and the checks get bigger every quarter. That is why the stock no longer costs $345. The story leaves out the number on the other side of the ledger, and it is a large one.

At the end of its fiscal year on May 31, Oracle carried $638 billion of remaining performance obligations — the revenue it has signed contracts to deliver but has not yet earned. That backlog grew by $85 billion in a single quarter and rose 363% year over year. It is roughly nine and a half times the $67.4 billion of revenue Oracle recognized in all of fiscal 2026, and it now edges out Microsoft's backlog of about $627 billion. RPO is the strongest visibility a software company can show: these are signed obligations, not hope.

So run the comparison that matters. Oracle's entire stock market value is around $429 billion, and even its enterprise value in the low $500 billions sits below the $638 billion already under contract. The market is valuing the company at less than the revenue it has signatures for. That is gross numbers, not profit — Oracle still has to build the data centers to earn it, which is exactly the spending that scares investors. But the gap is either the definition of a bargain or a warning sign, and the entire investment question is which one it is.

Give the sellers their case first, because it is real. Capital expenditures jumped 162% in fiscal 2026. Free cash flow was negative $23.7 billion even though operating cash flow set a record at $32 billion. Oracle raised about $43 billion of debt and $5 billion of equity last fiscal year, and it plans roughly $40 billion more in the year ahead, including a $20 billion stock-sale program. That, plus a fiscal 2027 capex forecast above Wall Street's number, is why the stock sold off even though the report beat estimates, and went on to its worst week since 2001.

The key to the story is which problem that spending is. Companies burn cash either because demand died, or because they are choosing to spend ahead of demand they have already collected. Oracle's burn is the second kind, and the proof is unusually concrete. About $75 billion of the backlog represents prepaid or customer-supplied GPUs — customers paid Oracle, or supplied the hardware, in advance, so a meaningful slice of this build is pre-funded by the people buying it. And the spending is landing on the income statement, not just in press releases: cloud infrastructure revenue grew 93% in the fourth quarter, cloud revenue grew 47%, total revenue grew 21%, and operating cash flow rose 54% to a record. That is a scale-up converting signed demand into reported revenue, not a business in retreat.

The VA deal reads exactly this way. The Department of Veterans Affairs added nearly $17 billion to its Oracle Health electronic health record contract, lifting the ceiling from about $10 billion to nearly $27 billion and adding three option years that run through May 2031. The history matters: this is the system Oracle inherited with Cerner — launched in Spokane in 2020, linked to documented patient harm, and had its rollout halted by the prior administration in 2022. Lawmakers spent years talking about scrapping it. Instead, Washington just added $17 billion to it. For a segment investors had written off, the government's decision to double down is the opposite of the cancellation narrative.

Keep the deal in scale, and note what it is not. The $17 billion is spread across deployments through 2031 — on the order of a few billion dollars a year against the $90 billion Oracle has guided for fiscal 2027. Part of the raise just keeps pace with money the VA was about to run out of, and the added ceiling is not yet appropriated; the House's fiscal 2027 spending bill sets aside $3.4 billion for the EHR program. It is a real contracted signal, not a windfall. Its larger value is diversifying the backlog, because the single largest contract inside that $638 billion is Oracle's $300 billion, five-year cloud deal with OpenAI, which begins in 2027. The VA re-up is evidence that not every dollar of the backlog is one hyperscaler; it is contracted, governmental, and diversified demand the market stopped paying for.

Now the price, because "bargain" needs a number behind it. Oracle guided to about $8.05 of non-GAAP earnings per share for fiscal 2027 — up roughly 18% after stripping out fiscal 2026 one-time gains. At $148.87, that is about 18.5 times next year's guided earnings. Eighteen and a half times, for a business guiding total revenue up about a third, with a signed backlog at 9.5 times trailing sales. Salesforce trades near 21 times trailing earnings and SAP near 27 times; neither is growing anywhere near Oracle's contracted rate. Oracle is paying a discount for having the most signed growth in commercial software — a discount partly earned by real costs (margin dilution while the build runs, equity dilution from the ATM, and hyperscaler concentration) and partly the result of treating a temporary deployment as a structural burn.

The discipline is the same one that applies to any backlog bargain: it only closes if Oracle delivers the $8.05 while cloud infrastructure gross margins hold and capex finally peaks, and if OpenAI's contract stays intact into 2027. Slip on the first, and the discount was honestly earned. Renegotiate the OpenAI deal, and the $638 billion stops being armor and becomes unfinished bookkeeping. The VA news alone does not answer that. But at roughly 18.5 times next year's guided earnings, with a third of next year's revenue already under contract and converting at record rates, the market is pricing the failure case — and the contracts say that is the wrong case to pay for.