Oracle’s fiscal 2026 second-quarter report showed the opportunity and the risk in its artificial-intelligence expansion at the same time. For the quarter ended November 30, 2025, revenue reached $16.1 billion, cloud revenue grew 34% to $8.0 billion, and Oracle Cloud Infrastructure (OCI) revenue rose 68% to $4.1 billion. Remaining performance obligations (RPO) reached $523 billion, up 438% year over year. Yet trailing-four-quarter free cash flow was negative $13.2 billion as capital spending reached $35.5 billion.
This retrospective preserves the five issues investors were watching before the December 10, 2025 release: whether OCI growth could continue, whether AI contracts could generate acceptable returns, whether the backlog was high quality, whether applications remained durable, and whether partnerships improved Oracle’s competitive position.
Quick facts: what Oracle reported
| Metric | Analyst expectation cited before the report | FY2026 Q2 actual |
|---|---|---|
| Total revenue | About $16.14 billion; approximately 13%–16% growth, according to KeyBanc and Bernstein | $16.1 billion, up 14% |
| OCI/IaaS revenue | About $3.88 billion; approximately 67%–75% growth | $4.1 billion, up 68% |
| RPO | About $523.3 billion, according to KeyBanc | $523 billion, up 438% year over year and $68 billion sequentially |
| Capital spending | Approximately $8 billion for the quarter, according to KeyBanc | Trailing-four-quarter capital expenditures of $35.5 billion; this is not a reported quarterly figure |
Oracle announced the results after market close on December 10, 2025, for the three months ended November 30, 2025. The company’s earnings call was scheduled for 4:00 p.m. Central Time. The official release is available at Oracle Investor Relations.
1. OCI growth was exceptional, but revenue timing was uneven
OCI/IaaS growth of 68% exceeded the most conservative pre-earnings estimates. That does not mean every dollar of contracted AI capacity was already producing revenue. GPU facilities can take months to permit, build, equip and bring online. Hardware delivery, customer deployment, utilization and accounting recognition occur at different times.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems#1 Best Overall
KeyBanc described AI-related revenue as potentially “extremely lumpy,” depending on when large deployments became operational. Morgan Stanley channel checks suggested that constraints on H100 GPU capacity could ease as bare-metal instances became generally available. A pre-earnings estimate also pointed to a possible sequential OCI GPU-IaaS increase of about $50 million; that was an analyst estimate, not reported revenue.
How to interpret the growth
- OCI includes AI training and inference, but also broader compute, storage, networking and platform workloads. It should not be treated as pure AI revenue.
- Ask how much growth came from training versus inference and conventional enterprise workloads.
- Compare contracted capacity with installed, live and actually consumed capacity.
- Watch whether growth broadens beyond a few very large model providers.
2. The AI opportunity carried a substantial cash cost
Oracle’s trailing-four-quarter operating cash flow was $22.3 billion, while capital expenditures were $35.5 billion, producing trailing-four-quarter free cash flow of negative $13.2 billion. Those periods match, but they should not be confused with the approximately $8 billion quarterly capital-expenditure estimate cited before the release.
The accounting presentation can also obscure the economic burden. Oracle may lease data-center capacity rather than own every facility, and some server components may be acquired shortly before a site becomes operational. Leasing can reduce reported capital expenditure in a period, but it creates recurring lease, operating and financing obligations. It does not eliminate the need to fund power, networking, equipment and commitments.
Questions that matter more than the headline growth rate
- When does management expect free cash flow to recover, and what milestones support that forecast?
- How much additional debt or equity financing is required for the build-out?
- Are AI infrastructure margins above or below Oracle’s company-wide margins?
- What return on invested capital does Oracle expect from contracted capacity?
- Are customer prepayments or customer-supplied GPUs reducing Oracle’s funding requirement?
Oracle later said that some large AI contracts could involve prepaid GPUs or customer-supplied GPUs, which can reduce the infrastructure Oracle must finance. The effect depends on contract terms and does not make every AI commitment equally profitable.
3. RPO was enormous, but backlog quality was the key question
RPO represents contracted revenue that has not yet been recognized. Oracle’s $523 billion figure was therefore not current sales, cash collected, profit or guaranteed near-term revenue. It rose 438% year over year and $68 billion sequentially, making conversion and collectibility more important than the absolute number.
The bridge from contract to cash
- A customer signs a contract, potentially with several tranches over a term of about five years.
- Oracle deploys facilities, GPUs and networking, or the customer supplies or prepays for some equipment.
- Capacity becomes available and the customer begins consuming it.
- Oracle recognizes revenue as performance obligations are satisfied.
- Collections, infrastructure costs, lease commitments and financing determine free cash flow.
Pre-earnings coverage attributed an approximately $300 billion Oracle contract to OpenAI. That figure should be treated as a reported contract value, not as current revenue or cash. Concentration creates separate risks: a customer may delay deployment, renegotiate, experience credit stress, change its model economics or use less capacity than expected.
What investors should seek in disclosures
- The share of RPO represented by the largest customers.
- Prepayment amounts, customer-provided hardware and minimum purchase commitments.
- Cancellation, termination and collateral provisions.
- The expected timing of revenue recognition.
- Whether Oracle must fund capacity before customer payments arrive.
4. Fusion, NetSuite and databases tested the quality of growth
AI infrastructure was not Oracle’s only growth engine. Applications and database services provide recurring demand that can diversify a business exposed to large, irregular GPU deployments.
Before the release, Morgan Stanley expected approximately 25% Fusion growth and approximately 21% NetSuite growth. Partner checks cited strong Fusion backlogs and win rates in financial services and healthcare, while demand among small and midsize NetSuite customers appeared less certain.
The reported results were more measured: Fusion Cloud ERP revenue was $1.1 billion, up 18%, and NetSuite Cloud ERP revenue was $1.0 billion, up 13%. SaaS revenue overall was $3.9 billion, up 11%.
What to watch in the applications business
- Fusion renewals, backlog and competitive win rates.
- NetSuite demand among small and midsize businesses.
- Whether cloud migration replaces declining on-premises license revenue.
- Cross-selling among databases, OCI and applications.
- Whether AI features generate incremental subscription revenue or are bundled into existing plans.
5. Partnerships could expand distribution—but also compress margins
Oracle’s relationships with Microsoft Azure, Amazon Web Services, Google Cloud, Nvidia, AMD, Cohere and other AI companies are intended to improve access to customers, GPUs and multicloud deployments. Database@Azure and related offerings can let customers use Oracle databases alongside another hyperscaler, reducing migration friction.
The commercial test is consumption, not the number of announcements. Investors should ask how much OCI growth is sourced through partners, whether multicloud database products produce incremental workloads, what revenue share Oracle gives partners, and whether a partner is also competing for the same AI infrastructure customer.
Where neocloud providers fit
Lambda, Vultr, Armada and Fireworks AI were cited as specialist providers in the pre-earnings discussion. Some may compete with Oracle for GPU workloads; others may lease Oracle capacity, aggregate demand or provide software. Their economics depend on GPU supply, networking, utilization, financing, power and switching costs. It is inaccurate to treat every specialist provider as a direct Oracle rival.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Best Value
How to judge a strong or risky quarter
| Stronger evidence | Riskier evidence |
|---|---|
| OCI growth meets expectations across multiple customers. | RPO rises while recognized revenue barely moves. |
| Management gives conversion, utilization and funding milestones. | Growth depends on a small number of contracts with little concentration detail. |
| Operating cash flow improves relative to capital spending. | CapEx, debt or lease obligations rise faster than cash generation. |
| Fusion, NetSuite and database demand remain healthy. | Applications slow while traditional license revenue declines. |
| Partnerships produce measurable consumption. | Partnerships are described only through announcements and potential. |
| AI margins and customer protections are disclosed clearly. | Management repeatedly delays free-cash-flow recovery. |
Common mistakes when reading Oracle’s report
- Calling RPO revenue or cash.
- Calling every OCI dollar AI revenue.
- Comparing a quarterly CapEx estimate with a trailing-four-quarter free-cash-flow figure.
- Assuming a signed contract proves customer profitability or Oracle’s collection.
- Treating leased data centers as a solution to capital intensity rather than a change in the cost profile.
- Assuming partner distribution is automatically margin-accretive.
What the results meant for investors
The FY2026 Q2 report largely confirmed the bull case for demand: Oracle delivered 68% OCI growth, 34% cloud growth and a $523 billion RPO. It also confirmed the central risk: the infrastructure required to serve that demand was consuming more cash than the business generated over the trailing four quarters.
For shareholders, the next question was not whether Oracle could announce large AI contracts. It was whether those contracts would become utilized capacity, recognized revenue, collected cash and durable returns without excessive customer concentration or fixed infrastructure obligations.
Separate cloud-buying considerations
Investment performance and cloud procurement are separate decisions. Businesses evaluating OCI can review Oracle Cloud, its cloud pricing, the Free Tier and Universal Credits. Oracle describes more than 20 always-free services and $300 in promotional credits for 30 days; production users should not treat a free tier as guaranteed capacity or enterprise support.
Universal Credits can use pay-as-you-go or annual-commitment models. Unused credits may expire at the end of the contract term, while excess usage is billed monthly at the contracted rate. Oracle’s cost estimator is for planning and evaluation, not an official quote.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
For Oracle Database workloads, Base Database Service pricing varies by OCPU/ECPU configuration, compute shape, storage, performance units, edition and included licensing versus BYOL. Oracle states that one x86 OCPU is equivalent to two vCPUs.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
