Oracle's fiscal fourth quarter did not weaken the AI infrastructure case. It made the case more demanding.
The numbers, in a less capital-heavy business, would have been read as clean proof of strong growth. Revenue rose 21% to $19.18 billion, above the $19.10 billion consensus. Adjusted EPS was $2.03 stripping out one-time gains, or $2.11 including them. Either way, both beat the $1.96 Wall Street expected. Net income climbed to $4.2 billion.
But the stock fell about 10% after hours. Investors did not treat the print as an earnings beat. They treated it as a financing update.
That is the central debate now: Has Oracle won the AI demand race? Most investors say yes. Can it convert that demand into strong free cash flow without changing its risk profile? That is the new question.
What the Quarter Proved
Oracle is no longer a small player in AI infrastructure.
Oracle Cloud Infrastructure, or OCI, revenue rose 93% to $5.8 billion. Total cloud revenue rose 47% to $9.9 billion. Cloud applications grew 10% to $4.1 billion.
The headline number, though, was remaining performance obligations, or RPO. That is the value of signed contracts not yet booked as revenue. RPO hit $638 billion, up 363% from a year ago and $85 billion higher than at the end of Q3.
For the first time, CFO Hilary Maxson gave a timetable. Oracle expects to recognize 12% of that RPO over the next 12 months. Another 34% should land in months 13 to 36. That clears up one big question. The visibility is real.
But it shifts the debate. Demand is no longer the issue. Returns are. Can these workloads earn enough after financing costs, depreciation, power constraints and execution risk?
Co-CEO Clay Magouyrk pushed hard on demand. He said Oracle signed $67 billion of AI infrastructure contracts in the quarter. Most were prepaid or used customer-supplied hardware. Global GPU usage ran at 97.5%. At the Abilene, Texas site, 42% of capacity is delivered. Another 35% should come online in 90 days. Oracle plans to bring nearly one gigawatt of capacity online in Q1 alone, almost the same as the entire prior fiscal year.
The OpenAI Question
The OpenAI relationship gives Oracle real AI credibility. It also creates real risk.
Bank of America analysts said more than 50% of Oracle's RPO comes from a single multi-year OCI deal with OpenAI. That is a striking number. It means Oracle's AI growth story is heavily tied to one customer's spending plans, funding path and capacity needs.
The "Stargate" site in Texas, built with OpenAI, should be more than three-quarters done within 90 days. That is real execution. But it also means Oracle's near-term success leans on how well OpenAI scales.
What Changed From Q3
Q3 reassured investors. Customer prepayments and bring-your-own-GPU deals seemed like helpful side features that would reduce Oracle's funding burden.
Q4 made clear they are now central to the capital model.
Oracle said prepaid and customer-supplied hardware deals now total $75 billion. Magouyrk said this "substantially reduces" the capital Oracle has to raise.
Maxson added that net cash capex in fiscal 2027 will be about $70 billion. Reported capex will be higher by $20 billion to $25 billion due to prepayments and timing. Capex for the year just ended was $55.7 billion, up 162% from $21 billion the year before. Depreciation nearly doubled to $7.62 billion.
The funding picture is still big. Oracle plans to raise another $40 billion in fiscal 2027 through debt and equity. That includes a previously announced $20 billion stock sale. Last year, the company raised $43 billion in debt and $5 billion in stock. Free cash flow was negative $23.7 billion in fiscal 2026, even with $32 billion in operating cash flow.
Oracle now has about $117 billion in high-grade debt outstanding. That makes it the largest non-financial issuer in the Bloomberg US high-grade index. Debt and lease liabilities ended the quarter at $156 billion, up 50% year over year. Another $261 billion in lease commitments has not yet started.
The customer prepayment structure is real risk reduction. But only relative to a more capital-heavy version of the same plan. It does not erase the funding question.
What Management Is Promising
Maxson reaffirmed fiscal 2027 revenue guidance of $90 billion. She raised non-GAAP EPS guidance to $8.05. Q1 guidance was strong: revenue growth of 27% to 29%, cloud revenue growth of 58% to 64%, and EPS of $1.72 to $1.76.
The bigger guidance, though, was qualitative.
Maxson said fiscal 2027 gross margin will "step down" as data centers ramp. Mix is shifting toward infrastructure, which carries lower margins than software. But she said infrastructure margins should improve quickly as new sites reach full contracted revenue. Once OCI reaches steady state, she said return on invested capital should hit the high 20s in percent terms.
Magouyrk added that Oracle avoids fixed-price exposure when future costs are uncertain. Contracts include pass-through mechanisms for supply chain cost increases.
That commentary was more detailed than in prior quarters. It was also more defensive. Oracle is now explaining contract structures, capex timing, component cost risk, prepayment accounting and project-level returns. Those are the variables that now drive the stock.
Software Is No Longer a Side Issue
Co-CEO Mike Sicilia made the case that AI strengthens Oracle's software business. He cited more than 1,000 AI agents across application suites, more than 300 Fusion go-lives in the quarter, and cloud database revenue up 29%. Multicloud revenue grew 404%.
"In the mission critical system space, which is where we play at Oracle, people have quickly moved on" from AI doubts, Sicilia said.
The numbers tell a more mixed story. Software revenue fell 2% in Q4 and 1% for the full year. Cloud applications grew only 10%, far slower than OCI. Both legacy software and cloud software came in below expectations.
This matters beyond Oracle. The whole software sector is facing investor concern that AI will disrupt the user-based subscription model. The iShares Expanded Tech-Software ETF is down 13% this year while the S&P 500 is up 8%. Oracle's Q4 did nothing to push back on that narrative.
Operating efficiency helped offset some of the pressure. Oracle spent $1.8 billion on restructuring this year, nearly five times last year's level. Layoffs that began in March helped lift Q4 adjusted operating margin to 45% from 43% the prior quarter.
Why the Stock Fell
The decline was not a rejection of AI demand. It was a repricing of execution and financing risk.
Bulls got more backlog, faster OCI growth, clearer RPO conversion timing and higher EPS guidance. Skeptics got higher capex, deeper negative free cash flow, gross margin pressure, more equity issuance, rising debt, and heavy concentration in a small set of AI customers.
That split explains the reaction. Oracle beat the quarter. It also made clear that the next phase of value creation depends less on software-style operating leverage and more on infrastructure-style project execution.
Investor Takeaway
Oracle's Q4 strengthened the AI infrastructure thesis. It also reframed the stock.
The company now looks less like a high-margin software compounder with a cloud option. It looks more like a hybrid of enterprise software, hyperscale infrastructure and project-financed AI capacity. By fiscal 2030, Oracle expects OCI revenue to reach $166 billion, or about three-quarters of total sales. That is the transformation playing out in real time.
That hybrid model can work well if OCI delivers the promised margins and high-20s returns. It can also fail if capex inflation, customer concentration, funding costs, delivery delays or AI demand swings impair returns.
The quarter did not settle the debate. It raised the stakes.
Investors now appear to be underwriting Oracle as an AI compute platform with unusually strong contracted demand. But the balance sheet and cash flow profile make execution risk central to valuation. The burden of proof has shifted.
Oracle no longer needs to prove AI demand exists. It needs to prove that demand can be turned into durable, high-return infrastructure economics without breaking the software-margin identity that historically defined the company.