Report contents
Industry landscape
Where the company competes: databases fund it, cloud capacity decides growth
Market position: concentrated in software, crowded in infrastructure. Oracle sells three things that a CIO may buy as one bundle but that compete in separate markets: the Oracle Database and the middleware and tooling around it, enterprise applications (Fusion ERP, HCM and supply chain, plus NetSuite for mid-sized companies), and rented computing and storage through Oracle Cloud Infrastructure. The first two are subscription and license businesses with renewal annuities. The third is a capital-intensive utility where Oracle is the smallest of the four Western providers operating at scale.
The value chain explains why the two halves earn different returns. Oracle owns the database engine and the application code, sells mostly through its own salesforce to large enterprises, and supports those installations under multiyear contracts. In software, the margin sits with whoever owns the code and the installed base, because the buyer's realistic alternative is rewriting the applications that sit on top of the database. In cloud infrastructure, the margin sits with whoever fills a region's servers most completely: Oracle builds or leases the data centers, buys the processors, and rents capacity by the hour. Utilization and capital cost decide the economics there, not source code.
The core market is transaction processing, not "software"
Strip out the labels and Oracle's core market is enterprise transaction processing: relational databases running online transaction processing workloads for banks, insurers, retailers, telecoms and governments, together with the ERP, HR and supply-chain systems that write into them. That is the submarket where Oracle's pricing power and installed base actually live. Public cloud infrastructure is a second, separate arena, and it is the one driving the company's reported growth and its capital spending.
Geography differs by arena. Database and application licenses are sold globally, and Oracle reports revenue across the Americas, EMEA and Asia Pacific, with the Americas the largest profit pool. Cloud infrastructure is regional in a way software is not: data-residency rules and latency keep a European bank's core ledger in a European region, and a government workload usually stays inside the country. Regional density therefore matters. A provider with a dense local installed base of database customers can fill a new region more cheaply than a provider starting cold, which is the one place where Oracle's software franchise directly subsidizes its infrastructure business.
Market structure: a handful of database rivals, a three-firm cloud oligopoly
| Arena | What Oracle sells | Main rivals |
|---|---|---|
| Enterprise database (transaction processing) | Oracle Database, Exadata, Autonomous Database | Microsoft SQL Server, SAP HANA, IBM Db2, AWS database services, open-source PostgreSQL |
| Enterprise applications | Fusion ERP, HCM, SCM; NetSuite | SAP, Workday, Salesforce, Microsoft Dynamics 365, Infor |
| Public cloud infrastructure | OCI, OCI Database | AWS, Microsoft Azure, Google Cloud, Alibaba Cloud in China |
The concentration test cuts differently by row. In databases, four or five vendors matter for large transactional deployments, but open-source PostgreSQL and the cloud providers' own engines have made the low end of the market contestable, so entry barriers are high for mission-critical cores and moderate elsewhere. Enterprise applications are more crowded: SAP, Oracle, Workday and Microsoft each hold meaningful positions, with regional and vertical specialists filling niches. Cloud infrastructure looks like the most concentrated arena by revenue despite dozens of entrants, because unit costs fall with scale and three firms have already built the largest footprints. Oracle can enter that arena, and has, but it enters with a smaller base over which to spread the same fixed costs.
That structure sets up the central question for the moat. Oracle's profit is concentrated in support and renewal streams attached to software that customers find expensive to replace, and those streams are exposed at the edges to cloud-native alternatives. The infrastructure business adds revenue and strategic relevance, and it also adds the bargaining power of hardware suppliers and the capital demands of a utility. Whether Oracle earns attractive returns on that new capital is a separate question from whether its database franchise is durable, and the two should be judged on different evidence.
Competitive advantage
Does the company have a moat? Yes, but the protected part is no longer where the growth is
Overall rating: Fairly Strong. Oracle's most durable asset is an annual support bill on database software that customers renew because replacing it is dangerous. That bill was $19.8 billion in fiscal 2026 and has been roughly flat for five years, while total revenue rose from $42.4 billion in fiscal 2022 to $67.4 billion in fiscal 2026. Almost all of the growth came from renting computing capacity, a business with three larger competitors and no captive base. The moat is real, and it is a shrinking share of what Oracle sells.
Concentration explains why the annuity holds. Large transactional database deployments support a handful of vendors, and a buyer weighing a switch faces a multiyear migration, not a cheaper license. Returns confirm a barrier exists: Oracle earned an estimated 12.9% to 19.9% after-tax return on invested capital in each of the five fiscal years through May 2026, comfortably above its cost of funds as a BBB- issuer. Profits that survive a capex cycle, a leadership transition and a technology shift usually sit behind something. Profits that come from good execution alone get copied.
Market position: a stable installed base, a small position in rented capacity
Market position: Fairly Strong. Fiscal 2026 revenue splits five ways: software $24.5 billion, cloud infrastructure $18.1 billion, cloud applications $15.9 billion, services $5.7 billion and hardware $3.1 billion, the company reported in its annual filing. Support fees make up about four-fifths of the software line, and that slice behaves like a protected franchise. Cloud infrastructure grew 121% in the quarter ended August 31, 2026 on the company's figures, from a base that still leaves OCI the smallest of the four Western clouds. That growth is real and does not by itself indicate protection: a rented hour of computing is a commodity priced against three larger sellers.
Stability: Fairly Strong. No consistent five-year series of Oracle's share of transaction databases is published, so the cleaner test is the dollar level of support, which held flat across the period instead of shrinking. A captive base does not decline in dollars once credible cheaper alternatives mature, and open-source PostgreSQL and the cloud providers' managed databases did mature over those five years. Flat dollars are stronger evidence than a stable percentage, which can move with how an industry is reclassified.
Profitability: returns clear the cost of capital, with less room than in 2022
Core returns: Fairly Strong. ROE is unusable here: buybacks left equity so thin that the ratio reached 214% in fiscal 2024 and went negative in fiscals 2022 and 2023. Estimated ROIC is the better measure, with NOPAT taken as operating income × (1 − effective tax rate) and invested capital as equity + total debt − cash:
| Fiscal year | NOPAT (USD bn) | Invested capital (USD bn) | ROIC (%) | ROA (%) |
|---|---|---|---|---|
| FY2022 | 9.6 | 48.3 | 19.9 | 5.6 |
| FY2023 | 12.2 | 81.8 | 14.9 | 7.0 |
| FY2024 | 13.7 | 86.4 | 15.8 | 7.6 |
| FY2025 | 15.5 | 102.2 | 15.2 | 8.0 |
| FY2026 | 18.2 | 140.8 | 12.9 | 7.9 |
The level clears any reasonable cost of capital in every year, which is what a moat should produce. The direction does not. Invested capital nearly tripled from $48.3 billion to $140.8 billion while NOPAT less than doubled, and the return fell from 19.9% to 12.9%, its second consecutive annual decline. The five-year average of about 15.7% sits at the bottom of the 15%-25% band that usually separates a durable advantage from a good run of execution. ROA, which leverage does not distort, rose from 5.6% to roughly 8% and has held there, weighed down by $62.3 billion of goodwill and a fast-growing asset base. These figures are an upper bound on returns: they use the balance sheet at each year-end and exclude leases signed but not yet begun, which will add to invested capital before they add to NOPAT.
Margins and scale: Fairly Strong. Gross margin fell from 79.1% in fiscal 2022 to 65.2% in fiscal 2026 as lower-margin infrastructure revenue grew into the mix. Operating margin widened over the same span, from 25.8% to 30.9%, because revenue growth outran operating costs. The 25.4% net margin in fiscal 2026 flatters the operating economics: the November 2025 quarter posted a 38.2% net margin against a 29.5% operating margin, which points to a non-operating gain.
Scale is where the rating gets complicated. Oracle spreads R&D and a global salesforce across a large revenue base, which is a genuine cost advantage, but not a distinctive one against Microsoft or Amazon. In rented capacity it is the subscale participant, and that is where the capital is going.
Overall rating: Fairly Strong, with the annuity funding a thinner business
Two facts decide it. The support annuity is genuinely captive, holding its dollars through five years in which credible cheaper substitutes matured, and returns stayed above the cost of capital throughout. But profitability sits at the low end of the range that marks a durable advantage and has fallen for two straight years as capital-intensive revenue takes over. Captivity in the old business cannot carry a company whose capital and growth now sit in the new one.
Where the advantage comes from and how long it can last: switching costs carry it, capital dilutes it
Two mechanisms decide the outcome in enterprise transaction processing: whether the buyer can leave, and whether the seller's unit costs fall faster than a rival's. Regulation matters less than it does in banking or telecom, because data-residency rules decide where capacity sits, not who may build it. The single metric that tests the moat is the dollar level of support revenue. Hold it and the captive base is intact whatever the cloud does; let it slip while capex climbs and Oracle is trading a protected stream for an unprotected one.
Switching costs: Strong, and the annuity holds in dollars
A bank's ledger, an insurer's policy administration system and a retailer's order book all sit on a relational database, and a large share of those systems run on Oracle. Leaving means re-certifying or rewriting every application on top of it, retraining the staff who run it, and accepting cutover risk on a system that cannot be down for a week. The license fee is a small part of that bill. PostgreSQL is free and the big clouds sell managed versions of it cheaply, yet a rival's discount on the database itself is a rounding error against migration labor and downtime exposure. That asymmetry is why support renews almost automatically: the alternative to paying is an unsupported core, which most technology officers treat as uninsurable risk.
Habit plays a smaller part than risk. Large accounts often run a non-Oracle database beside Oracle for newer workloads, and that coexistence is the realistic path of erosion, not a wholesale switch. Habit does reinforce the base in one way: a large installed population sustains a deep pool of administrators, certified staff and third-party tooling, and that ecosystem raises the cost of leaving further. There is little true network effect, since one customer's database does not make another's more valuable.
The evidence that the base is captive rather than merely large is that its dollars did not shrink over the five years in which open-source and cloud-native databases matured into serious alternatives. Oracle also kept widening operating margin while carrying a cloud build, which a vendor losing pricing power in its largest renewal pool would struggle to do.
What the annuity is worth, estimated: support was 29% of fiscal 2026 revenue. Applying the 54.5% margin the company reported for its cloud and software segment that year gives roughly $10.8 billion of margin from the support line, and that is a floor, because renewals carry little of the selling cost that new licenses do. What would break it: two consecutive fiscal years of declining support dollars, which would say the installed base is leaving rather than merely that new licenses are slowing.
Scale: Moderate, and the fixed costs that matter now sit in the cloud
In software, Oracle's scale is real but unremarkable. R&D and a global salesforce spread across $67.4 billion of revenue, but Microsoft and Amazon are larger, so relative scale favors Oracle in none of its markets. In rented capacity the arithmetic differs: buildings, power contracts and chips for a region are fixed costs, and the provider that fills them keeps more of each hour sold. Minimum efficient scale — the volume at which unit costs stop falling — matters because a region that sits half full still pays for its land and its power. Oracle spreads those commitments over fewer billable hours than three larger rivals, so it reaches that threshold later.
Its one genuine scale-adjacent edge is seeding: existing database customers give a new region anchor tenants, which lowers the utilization risk that makes new cloud regions fail. The edge is limited. The accelerators come from Nvidia, the buildings are contracted out, and any larger rival can order more of both.
The cost of competing at subscale is visible in the cash flow. Capex was $55.7 billion in fiscal 2026 against operating cash flow of $32.0 billion, free cash flow was negative $23.7 billion, total debt reached $129.5 billion, and S&P Global Ratings cut the credit rating to BBB- in July 2026. Property, plant and equipment rose from $43.5 billion at the end of fiscal 2025 to $129.7 billion a year later and $161.8 billion at August 31, 2026. Shareholders' equity rose $24.3 billion in that August quarter while net income was $4.8 billion, which implies a large share sale on top of retained earnings.
The incremental return decides the case. From fiscal 2025 to fiscal 2026, invested capital rose $38.5 billion while NOPAT rose $2.6 billion, a return of about 6.8% on the new capital, estimated from the statements and below an 8%-9% cost of capital for a BBB- issuer. One year is a small sample and much of that capacity was not yet billing, so the evidence so far only shows the direction. If the incremental figure does not climb toward the low teens as contracted capacity starts paying, the cloud is consuming the annuity rather than extending it.
Durability: substitutes thin the inflow before they touch the base
The threat is not a rival winning accounts Oracle already has. It is new workloads never arriving. Applications built in the last five years more often start on a cloud provider's managed database than on Oracle's, and each system built elsewhere is a renewal that never enters the base. The annuity can therefore hold for a decade while the pool that would replace it shrinks, which shows up first as flat dollars and later as a gentle decline. The same logic applies to the AI build: if AI workloads standardize on cheaper accelerators or on the cloud providers' own tooling, Oracle's rented capacity competes on power contracts and price, not on anything a customer cannot buy elsewhere.
Two things that resemble protection are not. Brand recognition has not kept new applications on Oracle's database and does not fill data centers. Access to capital is not scarce: Microsoft, Amazon and Google borrow on better terms and hold larger balance sheets.
Customer captivity alone is the second-strongest kind of barrier, behind captivity reinforced by scale. Oracle has the first and, in the business absorbing its capital, not the second. That is why the moat rates Fairly Strong rather than Strong, and why the support line is the number to track: it is the only part of the business with a demonstrated barrier, and for now it is flat. Whether the database base can seed enough cloud demand to earn a return on the capital being committed is an open question, and it turns on execution and capital allocation, not on the strength of the franchise.
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Comparison and summaryHow Oracle Corporation compares with its closest rivals, and our moat rating
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on Oracle Corporation
Oracle Corporation
Business, moat and key questions
Open company page