In theory, a company named Oracle (ORCL +3.91%) should have a decent read on the future. The database giant is currently betting nearly $100 billion that its market read is correct. In practice, investors have spent the last year wondering whether Oracle's fate-predicting course was worth the tuition.
The stock closed at $137.79 on Sept. 29. That's up 3.9% for the day, which is nice, but it's not a smooth rocket ride. Last fall, Oracle's $300 billion OpenAI deal sent shares to a 52-week high of $322.54. Anyone who bought near that peak is currently looking at a 57.3% haircut.
Is Oracle stock a buy now? My answer is no, despite the sharp price cut. The reason comes down to a simple household rule that also applies to trillion-dollar tech ambitions: it's hard to get rich by spending more than you make.
Image source: The Motley Fool.
The world's most expensive renovation
Oracle is building AI data centers faster than its income statement can pay for them. In the first quarter of fiscal 2027 (which ended on Aug. 31), the company spent more than $28 billion on capital expenditures. Revenue for the same quarter was just over $19 billion. In other words, Oracle spent roughly $1.47 on property and equipment for every dollar of sales.
Free cash flow came in at-$5.4 billion. That followed a negative $23.7 billion for all of fiscal 2026. Over the trailing 12 months, Oracle has burned through $28.7 billion in free cash flow.
And the cash burn will continue. Management expects full-year capex of $90 billion to $95 billion, against revenue guidance of at least $90 billion. Oracle plans to spend about a full year's revenue on data centers, chiefly to handle AI workloads.
Three ways to pay for a $95 billion habit
Oracle is stretching its dollars as far as possible. The company has only $36.4 billion in cash reserves against $125.3 billion in total debt. At the same time, some of its largest AI customers are expected to bring their own AI accelerator chips to the data center while prepaying for the expected services. Oracle recorded $11.4 billion of these prepayments in Q1. As a result, capital expenses in fiscal 2027 may rise to $95 billion, but actual cash expenses should remain below $70 billion.
The distinction matters for Oracle's cash position, but the service obligations don't go away. Customers who pay in advance expect capacity in return.
Shareholders are filling the remaining gap. In Q1, Oracle sold $20 billion of new common stock. The dilution was limited, given Oracle's $417.7 billion market cap today, but the company is still raising additional capital at the expense of existing investors.
A $664 billion promise with a power problem
The bull case rests on remaining performance obligations (RPO), which reached $664 billion in Q1. That backlog is up $209 billion from a year ago and represents more than nine years of revenue at the current run rate.
But a backlog is a contract, not cash. OpenAI's $300 billion agreement alone would account for about 45% of the total. That key customer's ability to fund its commitments is already in question.
And Oracle may struggle to fulfill its obligations. You can build all the data centers you want, but they don't do much without electricity.
Oracle's Project Jupiter campus in southern New Mexico is supposed to run on a 2.5-gigawatt fuel-cell microgrid from Bloom Energy (BE +10.80%). That microgrid still needs a state construction permit, with a decision due by Nov. 23. The natural gas to fuel it depends on a pipeline called the Green Chile Project. (Yes, really. It's New Mexico, after all.)
Green Chile's targeted service date was pushed back from August 2026 to February 2027, and federal regulators haven't signed off on construction yet. That's bad news for Bloom and very uncomfortable for Oracle.
On Sept. 24, Oracle sent a force majeure notice to the campus developer's owner. That's legal paperwork that protects Oracle if the project misses its 2028 target. Oracle says the schedule hasn't changed and that these notices are routine. Maybe so. But when a company files "just in case" paperwork with dramatic language, investors should wonder about the potential risks.
NYSE: ORCL
Key Data Points
Credit where it's due
I get Oracle's appeal. Cloud infrastructure revenue more than doubled last quarter, rising 121% to $7.4 billion. The business isn't broken. And Oracle still sells database services, PeopleSoft contracts, and other old-school technologies. This business is shrinking, but it still provides significant cash-flow support for Oracle's AI ambitions.
But Oracle's finances are stretched, like a rubber band under increasing tension. Those bands can hold a surprising amount of force, but it's never fun when they snap.
Worth watching, not buying
I'm not rooting against Oracle. Co-founder Larry Ellison and other insiders own about 40.5% of the company, so the people steering the ship have plenty of skin in the game. The AI bet might work out beautifully.
But "might" is doing a lot of work in that hopeful sentence. I'd rather watch from the sidelines until Oracle shows it can turn that backlog into actual free cash flow. Among the five largest AI spenders, Oracle's stock carries the most financial risk in my view.
