Oracle Corporation (NYSE: ORCL) — A Database Fortress Financing an AI Utility

Report date: 2026-08-09

Sector / industry: Information Technology · Enterprise Software and Cloud Infrastructure

Fiscal year end: May 31 · CIK: 0001341439

Price: $147.02 at 2026-08-07 close · Rebuilt market capitalization: approximately $423.4B · Conventional EV: approximately $526.6B · Lease-adjusted EV: approximately $564.5B Latest reported period: Q4 and FY2026, ended 2026-05-31

Standing disclaimer — Sections 1–15 carry no investment recommendation and no price target. Valuation is discussed as embedded expectations and scenarios. The single exception is the clearly fenced Claude’s Take above Section 1.

⚡ Claude’s Take

Claude’s subjective opinion; not investment advice. The analytical body below carries no position.

Verdict: HOLD / DO NOT AVERAGE THE FIRST BOUNCE; a speculative entry zone is roughly $120–$135, or about 15–17x Oracle’s $8.05 FY2027 non-GAAP EPS guide. That guide excludes material costs and is not a substitute for free cash flow, so the band demands a larger discount than a normal software multiple. Framing: a falling-knife capital-cycle special situation attached to a high-quality database franchise.

The 31% decline since June has repaired the obvious multiple excess, but the evidence deteriorated faster than the price alone suggests. FY2026 FCF was negative $23.7B, cloud-and-software margin fell four points, FY2027 still needs roughly $40B of debt/equity financing, and S&P cut Oracle to BBB-. Customer-prepaid or customer-supplied hardware meaningfully reduces funding and obsolescence risk, while IaaS +93% and $638B of RPO prove demand. They do not prove after-replacement returns or remove $260B of unrecognized lease commitments. At $147, 18.3x guided non-GAAP EPS looks superficially reasonable; 26.8x trailing GAAP EBIT and negative FCF do not.

The factor tape argues against treating the first rebound as confirmation. The stock remains below its 50- and 200-day averages, is down 42% over twelve months, and carries 52% specific volatility; Momentum loading has collapsed near zero while Growth exposure remains high. This is not a crowded short, and the database moat makes it a poor reflexive short. It is also not yet a conventional value stock. Conviction: medium. Bullish flip: two consecutive quarters of RPO conversion with cloud/software margin bottoming and a disclosed after-replacement project return above the cost of capital. Bearish flip: another ratings downgrade or material customer delay while Oracle continues equity-funded capex. Tag: “The contract is funded; the return is not.”

Changes since 2026-06-09

The June 9 HOLD / AVOID-HERE conclusion was directionally right, but its $140–$170 accumulation zone is no longer valid without a wider risk discount. The stock entered that range by realizing a stated bear trigger, not by cleanly preserving the thesis.

• Price reset: $211.82 to $147.02, down 30.6%. The own-history composite valuation percentile fell from 76th to 55.6th; P/S remains elevated at the 75.9th percentile.
• Demand strengthened: Q4 IaaS grew 93%, RPO rose $85B sequentially to $638B, and Oracle disclosed $75B of customer-prepaid or customer-supplied hardware portions.
• Duration did not improve: only 12% of RPO remains scheduled within twelve months. The larger balance is still heavily back-loaded.
• Economics worsened: FY2026 FCF was negative $23.7B; cloud/software segment margin fell from 63% to 59%; FY2027 gross margin is guided lower.
• Financing risk crystallized: S&P downgraded Oracle to BBB-, directly hitting the prior bear trigger. FY2027 requires roughly $40B of planned funding, including a $20B ATM.
• Core moat held: software support grew 1%, database cloud 29%, Multicloud AI Database 404%, and SaaS deferred revenue 16%. The evidence supports database captivity, not a general OCI scale moat.

📈 Stock Price Action — Five-Year Event Map

Oracle rose 74.9% from an adjusted $84.04 on 2021-08-09 to $147.02 on 2026-08-07, but that endpoint conceals a full boom and bust. The five-year closing low was $57.98 in September 2022; the five-year and 52-week high was $324.63 on 2025-09-10; the stock then fell 64.6% to $114.99 in July 2026 before rebounding 27.9%. It remains 54.7% below the high. (AZI price history)

The 2021 move followed Oracle’s Q2 FY2022 filing. The 2023 and 2024 gaps align with the company’s Q1 FY2024, Q2 FY2024, Q3 FY2024 and Q4 FY2024 results. The 2025 peak followed Q1 FY2026’s RPO disclosure; the latest reversal followed the FY2026 FCF and funding print plus S&P’s July downgrade. Price moves are facts; overlapping causal attribution remains interpretation.

Factor evidence describes an early rebound inside a broken trend. At August 7 the stock was 8.1% above its 21-day EMA but 1.4% below its 50-day and 15.4% below its 200-day; the 50-day remained below the 200-day. Three-month de-annualized return was approximately -24.7%, six-month +3.7%, and twelve-month -42.0%. FactorsToday’s all-factor model explained 43.4% of variance, with Market +0.94, Growth +0.90, Software +0.70, Technology +0.64, Quality -0.28 and Momentum only +0.06. The five-year return was positive but the Sharpe ratio was only 0.24 and maximum drawdown 64.6%. (FactorsToday)

1. Executive Summary

Oracle is no longer economically describable as a database company with a small cloud option. It is a database-and-applications annuity financing one of the industry’s most aggressive AI-infrastructure expansions. FY2026 revenue rose 17% to $67.4B; cloud infrastructure grew 77% to $18.1B; Q4 IaaS grew 93%; and remaining performance obligations reached $638B. The demand evidence is stronger than it was in June. The economic evidence is not: cloud-and-software segment margin fell from 63% to 59%, reported free cash flow was negative $23.7B, and S&P downgraded Oracle to BBB-, one notch above speculative grade. (FY2026 Form 10-K; FY2026 results)

The company contains two different competitive positions. Oracle AI Database, software support, Fusion, NetSuite, and industry applications benefit from customer captivity: migration risk, data gravity, implementation cost, retraining, process redesign, and compliance recertification. The $19.8B software-support stream still grew 1% even as licenses fell 9%, and customers migrating from support to Oracle Cloud contributed $5.4B of annualized cloud revenue over three years. Multicloud distribution lets customers keep Oracle-managed data while using AWS, Azure, or Google Cloud, extending the moat into rival infrastructure. By contrast, general-purpose OCI and accelerated-compute capacity do not yet possess a demonstrated scale moat. Third-party research placed Oracle at roughly 4% of cloud-infrastructure services in Q1 2026 versus 28% for AWS, 21% for Azure, and 14% for Google. One quarter of share gain and extreme growth during scarcity do not satisfy Greenwald’s tests of stable share and persistently superior returns.

The $638B RPO is commercially meaningful but financially easy to misuse. Only 12%, approximately $76.6B, is scheduled for recognition within twelve months; 34% falls in months 13–36, another 34% in months 37–60, and 20% thereafter. Oracle said most large Q3/Q4 additions included customer prepayments or customer-supplied GPUs, with those portions totaling $75B. That materially improves commitment and funding quality. It does not mean Oracle received $75B in cash: the audited filing separately identifies $4.6B of customer prepayments carrying a significant financing component, while the larger figure combines cash and customer-owned hardware. The backlog therefore lowers demand risk, but it leaves construction, power, operating, renewal, margin, and financing risk with Oracle in varying degrees.

The capital cycle is the central analytical frame. Oracle, Amazon, Microsoft, Alphabet, frontier laboratories, and neoclouds are adding capacity simultaneously. Current scarcity supports high utilization and rapid revenue growth; Oracle says GPU utilization is 97.5%. Supply arrives with long lags, however, while accelerator generations turn quickly and sophisticated buyers can multi-source. Oracle’s FY2026 net PP&E more than doubled, capex reached $55.7B, and the company plans approximately $70B of FY2027 net cash capex plus customer-prepayment and timing effects. It raised $43B of debt and $5B of equity in FY2026 and plans roughly $40B of additional FY2027 financing, including a $20B at-the-market equity program. The question is not whether the backlog converts into revenue; it is whether the asset cohort earns an after-replacement return above the cost of debt and equity after scarcity fades.

Management’s new answer is a claimed high-20s steady-state project return. The calculation adds depreciation back to after-tax operating margin and divides by gross project investment. It is useful as a project cash-yield hypothesis but is not conventional ROIC: it does not transparently charge replacement capex, fully allocated networking and shared datacenter assets, ramp losses, working capital, or financing. The filed results presently show the opposite direction at group level—gross margin fell, cloud/software direct costs rose much faster than revenue, free cash flow turned deeply negative, and leverage increased. Proof requires cohort returns after depreciation and replacement capital, not a utilization statistic during shortage conditions.

At $147.02, valuation is no longer the June report’s dominant objection. The shares have fallen 30.6% from $211.82 and 54.7% from their five-year peak. Conventional EV is approximately $526.6B and lease-adjusted EV $564.5B. On the latter, Oracle trades at 8.4x FY2026 sales, 27.4x GAAP operating income and 18.9x operating income plus D&A. The stock is 25.1x trailing GAAP EPS and 18.3x Oracle’s $8.05 FY2027 non-GAAP EPS guide. Price/sales still sits around the 76th percentile of Oracle’s own history, while P/E is near the 48th; P/B is not decision-useful because accounting equity is thin and tangible book is negative. The de-rating narrowed valuation risk but did not create obvious cash-flow support, because conventional FCF remains negative and FY2027 financing includes meaningful dilution.

Executive verdict: Oracle’s durable database captivity is funding a real but unproven infrastructure expansion. Demand and contract funding strengthened; margin, free-cash-flow, and credit evidence worsened. The institutional conclusion is position-free: the backlog increasingly validates revenue; it still does not validate shareholder returns.

2. Business Overview

2.1 Revenue architecture

Oracle reports Cloud and License as one operating segment, plus Hardware and Services. The reported segment is too broad for economic analysis because it combines high-margin support and software with capital-intensive infrastructure. The more useful cut is by offering:

Cloud represented 51% of revenue in FY2026, up from 43% in FY2025 and 37% in FY2024. Within cloud, infrastructure rose from 35% of cloud revenue in FY2024 to 42% in FY2025 and 53% in FY2026. That mix shift is the strategic transformation in one series: Oracle is gaining recurring consumption revenue, but each incremental infrastructure dollar brings more physical capital and lower gross economics than support or applications. (FY2026 Form 10-K)

Cloud applications include Fusion ERP, HCM and SCM; NetSuite for mid-market customers; and industry products such as Oracle Health, financial-services and utility applications. These products become embedded in approvals, payroll, finance, procurement, supply chains, clinical workflows and audit trails. Subscription revenue is recurring, implementation is costly, and changing vendors can require a multi-year re-platforming. The counterweight is competition: SAP remains formidable in large-enterprise ERP, Workday in HCM and finance, Salesforce in customer workflows, and Microsoft in bundled productivity and business applications.

Cloud infrastructure includes compute, storage, networking, database services, accelerated computing, sovereign and dedicated regions, and Oracle-operated database services inside rival hyperscalers. Two activities sit under one label but have different quality. Database services monetize Oracle’s installed-base captivity and proprietary engine. Large AI clusters monetize scarce power, chips, construction speed and operational execution. The former is a direct extension of a moat; the latter must earn one through durable cost, service, or contracting advantages.

Software support remains the cash anchor. Substantially all customers that purchase a license elect support, according to the filing. Support funds updates, security, regulatory compatibility and access to technical assistance, while customers avoid the cost and operational risk of replacing mission-critical databases. The $19.8B revenue stream grew despite falling license sales and migration toward cloud. That is better evidence of retention than an unsupported percentage estimate.

Hardware centers on Oracle Engineered Systems, including Exadata, and servers/storage used in integrated deployments. Hardware can reinforce database performance and on-premise or Cloud@Customer differentiation, but it is not a standalone growth engine. Services help implement and optimize Oracle technology. The company says services represented 8% of FY2026 revenue and carry lower margins than cloud/software and hardware. Services can improve customer outcomes and switching costs while diluting reported margin.

2.2 Customer value propositions

The value proposition varies sharply by cohort:

• Installed-base database customers avoid application rewrites, migration failure, downtime, retraining and recertification. Oracle can now deliver the database on OCI, on-premise, Cloud@Customer, or inside AWS, Azure and Google Cloud. The mechanism is customer captivity, not low price.
• ERP/HCM/SCM customers buy an integrated system of record, quarterly updates, a common data model and embedded automation. The mechanism is process and implementation switching cost, shared with other scaled suites.
• Frontier-model and large AI customers buy dense clusters, scarce power, fast delivery, networking and flexible funding. The mechanism today is scarcity and execution. Portability and buyer concentration make durable captivity uncertain.
• Regulated, sovereign and healthcare customers buy data residency, control, isolated regions and specialized workflows. Oracle’s deployment flexibility is a credible niche advantage, although Oracle Health’s acquisition-level returns remain undisclosed.
• Net-new cloud-native developers buy price/performance and service breadth. Here Oracle faces its weakest position: competitors possess larger ecosystems, more services and broader developer mindshare.

Oracle’s product strategy attempts to connect those cohorts. A database customer can move to Oracle-managed cloud without abandoning its engine; an applications customer can consume OCI and database services; an infrastructure customer can adopt Oracle data and applications. The strongest observed motion is support-to-cloud migration: $5.4B of annualized cloud revenue over three years came from customers migrating annual support contracts. The least proven motion is the reverse halo—whether AI-infrastructure wins generate material new Fusion, NetSuite or Health workloads.

2.3 Revenue quality, concentration and recognition

Cloud applications and support are recurring and relatively predictable. Infrastructure is also recurring or committed in many cases, but capacity, billing, delivery and customer composition make it different from a seat subscription. Oracle’s RPO includes deferred revenue, billed but unrecognized amounts, and future amounts to be invoiced. The filing also notes that seasonality, large-contract timing, foreign exchange and variable-consideration elections affect the balance. RPO is contracted demand, not cash and not current revenue.

Oracle does not disclose customer concentration within the $638B RPO. Management said four customers signed contracts exceeding $8B in Q4 and emphasized diversification, but the filing does not name customers or publish the largest-customer share. Prior press reports tied a very large portion of the earlier backlog to OpenAI; that is relevant context, not a substitute for current audited concentration disclosure. A decision-useful report therefore treats concentration as high but unquantified.

The customer-prepayment and bring-your-own-hardware structures are economically constructive. A customer supplying GPUs transfers accelerator purchase and residual-value risk away from Oracle; prepayment reduces financing needs and raises commitment. Yet Oracle continues to fund land, power, buildings, networking, general compute and operations. Its own risk factors warn that lease terms and pricing may not match customer contracts and that customer nonperformance can leave it committed to space, capital and financing. Funding structure reduces a risk; it does not transform the activity into SaaS.

2.4 Oracle Health

Oracle paid $28.3B cash for Cerner in June 2022, then assumed the integration, restructuring and leverage consequences just as the AI build began. The strategic idea is coherent: combine clinical records and workflow with Oracle database, cloud and AI, while selling into a highly regulated customer base with extreme switching costs. The financial proof is not. Oracle no longer provides enough Oracle Health revenue, margin, retention or cloud-migration disclosure to calculate acquisition-level returns. Management expects double-digit Oracle Health growth in FY2027, but a forecast is not evidence. The correct status is strategically plausible, economically unproven.

2.5 Business-overview verdict

Oracle owns a durable recurring software core and a rapidly expanding infrastructure business. The core supplies customer captivity and cash-earnings resilience; the infrastructure supplies growth and backlog but consumes capital. Consolidated labels obscure that difference. The company should be analyzed as a barbell, not awarded the software franchise’s historical economics on every new revenue dollar.

3. Industry Dynamics

3.1 Three profit pools, three structures

Oracle competes in database management, enterprise applications and cloud/AI infrastructure. All three benefit from secular data and compute growth, but their supply economics differ.

Database management. Gartner projects worldwide DBMS spending of approximately $161B in 2026, growing 18.4%, and $255B by 2029. The market is attractive for incumbents because databases sit beneath mission-critical applications and migration is operationally risky. It is also fragmenting: managed PostgreSQL, cloud-native databases, Snowflake, Databricks, hyperscaler services and open-source tools win many net-new workloads. Oracle’s serviceable market is therefore the enterprise operational/relational and hybrid pool where its installed base, Exadata, sovereignty and multicloud delivery matter—not every dollar of analytics and open-source spending. Public evidence does not support a precise current Oracle market share. (Gartner DBMS forecast)

Enterprise applications. ERP, HCM, SCM and industry systems are sticky, concentrated markets. Implementation and process redesign create customer captivity; recurring cloud delivery supports visibility; integrated data and compliance favor scale. Yet customers can choose among several credible vendors and use specialist products at the edges. Current peer prints show Oracle applications growing more slowly than some direct competitors: Oracle cloud applications grew 10% in Q4, SAP’s Q1 cloud ERP suite grew 30% constant currency, and Workday subscription revenue grew 14.3%. Definitions and periods differ, but the contrast challenges the claim that Oracle’s full-stack AI strategy already creates superior applications growth. (SAP Q1 2026; Workday Q1 FY2027)

Cloud and AI infrastructure. Entry barriers are enormous: power, land, construction, chips, networking, software orchestration, security certifications, global operations and balance-sheet capacity. Those barriers support an oligopoly, but Oracle is the small No. 4 rather than a coequal incumbent. Synergy estimates reported by industry publications put Q1 2026 shares near AWS 28%, Azure 21%, Google 14% and Oracle 4%. Q2 cloud-infrastructure spending reached approximately $143B, an annualized $572B pool, with the top three still controlling roughly 63%. (CRN/Synergy Q1; ITPro/Synergy Q2)

3.2 The capital-cycle problem

Marathon’s capital-cycle framework asks what happens to supply, not merely whether demand grows. High current returns and scarcity attract investment; capacity arrives after a lag; pricing and utilization then normalize. Cloud/AI sits deep in that boom/lag sequence. AWS grew 37% in Q2 and produced a 39.4% segment margin. Azure exceeded $100B and grew 43% in Microsoft’s Q4. Google Cloud approached a $99B quarterly annualized run rate and accelerated sharply. Oracle IaaS grew 93% in Q4. Every major provider therefore has demand evidence, and every major provider is expanding.

Current peer disclosures show the size of the supply response. Amazon raised expected 2026 cash capex to about $220B; Alphabet guides $195B–$205B; Microsoft spent $115.9B cash capex in FY2026 and added 88 datacenters; Oracle plans about $70B of FY2027 net cash capex plus $20B–$25B of customer-prepayment and timing effects. Neoclouds and frontier laboratories add more capacity. Demand can remain strong while returns fall if capacity grows faster, scarcity pricing weakens, or accelerator replacement cycles compress asset lives.

Oracle is more exposed than the top three in two ways. First, it lacks their scale and balance-sheet flexibility. Second, a concentrated set of sophisticated AI buyers can multi-source and negotiate funding, pricing and hardware ownership. Oracle’s mitigants are faster construction, secured power, high-density cluster operation, database adjacency, sovereign deployments and flexible contracts. Those are real execution advantages. They become a durable scale or captivity moat only if Oracle retains share, customers renew and returns remain above the cost of capital after shortage conditions end.

3.3 Competitive conduct and regulation

The cloud contest resembles entry and preemption more than a stable utility market. AWS, Azure and Google defend broad platforms; Oracle enters through database adjacency, large AI clusters and regulated deployments; neoclouds enter through accelerator specialization. Rapid market growth permits every provider to post high growth without taking much installed share. That makes current growth a weak moat test. The revealing period will be 2027–2029, when ordered capacity is operating and early AI contracts approach renewal or repricing.

Regulation creates a two-sided effect. The European Commission’s June 2026 preliminary position that AWS and Azure should be Digital Markets Act gatekeepers cited entrenched users, lock-in, switching costs, ecosystems and investment scale. The UK Competition and Markets Authority also focused cloud competition concerns on AWS and Microsoft. Near term, interoperability, switching and egress remedies can lower Oracle’s customer-acquisition barrier and validate multicloud. Longer term, the same rules attack captivity in databases and platform services. Oracle benefits as cloud challenger and faces risk as database incumbent. (European Commission; UK CMA)

The EU Data Act has applied since September 2025 and requires cloud providers to remove switching and multicloud obstacles; switching and egress charges must be eliminated from January 2027. EU AI Act transparency duties took effect in August 2026 for interactive and generative systems. Compliance favors vendors with scale and governance capabilities, while interoperability reduces some closed-platform advantage. Oracle’s sovereign, regulated and healthcare products can gain from complexity, but its embedded agents must absorb the same compliance burden. (EU Data Act; EU AI transparency guidance)

3.4 Industry verdict

Database and enterprise applications remain structurally attractive for incumbents: concentrated profit pools, recurring revenue and high switching costs. Cloud infrastructure is oligopolistic but late-cycle. Oracle occupies the favorable side of the first structure and the risky edge of the second. Industry growth is not in doubt; normalized incremental returns are.

4. Competitive Position

4.1 Greenwald moat map

Greenwald reduces durable competitive advantage to customer captivity, economies of scale and limited special cases such as protected technology. Oracle’s position is best understood product by product:

The strongest evidence is the software-support annuity. A no-moat software product would not retain a near-$20B support stream while licenses decline and cloud alternatives proliferate. The economic outcome that would deteriorate without captivity is clear: customers would migrate away, support would contract rapidly, pricing would fall and database cloud would not grow. Instead, Oracle is following its data into rival clouds. This passes the moat mechanism test.

Multicloud is strategically more important than its undisclosed revenue base suggests. Oracle Database@Azure, @Google Cloud and @AWS turns a potential disintermediation channel into distribution. A customer can use a preferred hyperscaler’s procurement, analytics and AI while retaining Oracle database performance, compatibility and support. That lowers the friction of staying with Oracle and may expand database consumption. It is primarily defensive: it proves Oracle can retain captive workloads, not that it is winning the majority of net-new databases.

4.2 Scale test for OCI

Oracle’s infrastructure growth is extraordinary, and its approximate share increase from 3% to 4% is real. Greenwald’s scale test nonetheless asks whether a firm holds a cost advantage within a bounded market and whether share remains stable over time. Oracle is roughly one-seventh AWS’s share and one-fifth Azure’s. Its competitors spread silicon design, networking, developer tools, security and global operations across much larger revenue bases. Oracle may hold local density advantages in very large clusters and can use customer-supplied hardware, but public disclosures do not establish a lower all-in cost curve.

Management’s high-20s steady-state project-return claim is the best direct counter-evidence. CFO Hilary Maxson described an after-tax operating-margin-plus-depreciation return on gross project investment and said prepaid or bring-your-own-hardware structures produce similar or better economics. This implies Oracle believes pricing, cost pass-through, utilization and customer funding overcome its scale disadvantage. The definition is not conventional ROIC, however. Adding back depreciation without charging replacement capital makes rapidly obsolescing hardware look more attractive; “steady state” excludes ramp losses; and project-level gross investment may not fully allocate shared power, networking, land, leases and overhead. The claim should be treated as a falsifiable hypothesis, not dismissed or capitalized as fact.

4.3 Competitive comparisons

AWS has the largest installed infrastructure base, broadest service catalog, mature custom silicon and a Q2 segment margin near 39%. Its scale and customer captivity are stronger. Oracle competes through database adjacency, contracting flexibility, cluster density and sovereign deployments.

Microsoft Azure combines enterprise identity, productivity, security, developer and application distribution with a balance sheet Oracle cannot match. Oracle’s response is to place its database inside Azure and monetize the customer rather than insist on full-cloud displacement.

Google Cloud combines TPUs, frontier models, data/analytics and global infrastructure. It remains smaller than AWS/Azure but larger than Oracle. Oracle’s database installed base and regulated deployment options differentiate it; Google’s AI stack and developer position are stronger.

SAP, Workday and Salesforce contest applications. Oracle offers a broad integrated suite and infrastructure; SAP retains deep ERP captivity, Workday is strong in HCM and finance, and Salesforce dominates customer workflows. Oracle’s Q4 application growth does not show a full-stack premium.

Snowflake, Databricks and PostgreSQL ecosystems challenge the database at the workload margin. They need not displace existing Oracle systems to matter; winning new analytical, AI and cloud-native applications can gradually narrow Oracle’s share of the expanding data layer.

4.4 Competitive-position verdict

Oracle has a durable but segmented advantage. Database captivity is wide and multicloud strengthens its defense. Applications add a narrower shared moat. General-purpose OCI and AI-cluster rental have not demonstrated a Greenwald scale moat; present success could reflect scarcity and execution. The corporate moat is durable in the installed core and unproven in the capital-consuming growth engine.

5. Growth History and Forward Opportunities

5.1 Multi-year growth spine

Oracle’s recent history has three distinct phases: mature software growth, the Cerner acquisition step-up, and the AI-infrastructure acceleration.

FY2023’s revenue step-up was heavily acquisition-driven because Cerner closed in June 2022. FY2024–FY2025 returned to mid-to-high-single-digit growth while Cerner integration and OCI buildout continued. FY2026 is the first full year where infrastructure changed the consolidated growth rate: IaaS rose 77% and represented 27% of group revenue. The critical distinction is that revenue growth accelerated while cash economics collapsed.

5.2 RPO conversion

RPO rose from $130.2B at February 2025 to $552.6B at February 2026 and $638B at May 2026. Q4 added approximately $85B sequentially. The balance is almost 9.5x FY2026 revenue, an extraordinary contracted-demand signal. Its timing is equally extraordinary:

The bull case sees a capacity-constrained company converting contracted revenue for years, with prepayment or customer-supplied GPUs reducing financing and obsolescence risk. The bear case sees a long-duration obligation whose value depends on build schedules, power, customer solvency, renewal economics and low-cost capital. Both are partly right. The current 12% share did not improve from June, so the larger balance does not shorten the duration.

5.3 Opportunity set

OCI and AI infrastructure are the largest dollar opportunity. Oracle guides FY2027 revenue to $90B, which would imply approximately 34% growth, and Q1 total revenue growth to 27%–29% in dollars. IaaS grew 93% in Q4 and current utilization is high. The constraint is delivery rather than contracted demand. The economic test is whether segment margin bottoms, depreciation is matched by durable cash return and financing intensity falls.

Multicloud database is the highest-quality growth opportunity. Q4 database cloud revenue grew 29%, Multicloud AI Database revenue grew 404%, and bookings grew 325%, all from undisclosed bases. This strategy monetizes existing captivity without requiring Oracle to win every general-purpose cloud workload. The metric needed is absolute database-cloud revenue and net retention, not another percentage on a small base.

Applications and AI agents can deepen retention and expand consumption. SaaS grew 10% in Q4 and 11% for the year; deferred SaaS revenue grew 16%. Oracle is embedding agents and experimenting with paid token bundles and outcome pricing, while many current features remain included. Near-term AI may defend the suite before it lifts ARPU. Sustained mid-teens application growth or visible consumption revenue would establish a halo; current growth trails that threshold.

Oracle Health offers cloud migration, clinical-data and workflow automation optionality. The installed base is sticky and regulation raises entry barriers. Without segment revenue, margin, retention and migration disclosure, however, it remains a monitoring item rather than a modeled growth engine.

Sovereign and regulated cloud differentiates Oracle from smaller providers and can win government, healthcare and financial-services workloads. Dedicated Region and Cloud@Customer allow data residency and control that public-cloud-only architectures may not. This is a credible niche with strong barriers, but public disclosures do not isolate its economics.

5.4 Growth-quality verdict

Growth is accelerating and more contractually visible than at any point in Oracle’s recent history. Its quality is mixed. Multicloud database and applications extend customer captivity; large AI infrastructure relies on capital, scarce inputs and concentrated buyers. The June 9 bull test is tracking because OCI conversion accelerated. The test is not passed until conversion persists for multiple quarters while segment margin, free cash flow and credit stop deteriorating.

6. Financial Quality

6.1 Five-year statements: growth bought with capital

Oracle’s five-year financial series separates the franchise from the current investment cycle. Derived gross profit is revenue less the direct cloud/software, hardware and services costs because Oracle does not label a consolidated gross-profit subtotal.

Cerner explains much of the FY2022–FY2023 margin reset. FY2026 is different: infrastructure mix drove derived gross margin down 470 basis points in one year and 1,330 basis points from FY2022. Revenue grew 17%, but cloud/software direct expense grew 52% while cloud revenue grew 39%. Cloud/software segment revenue rose 19% to $58.5B, yet segment margin dollars rose only 11% and the margin rate fell from 63% to 59%.

Oracle preserved consolidated operating margin through expenses below gross profit. Sales and marketing fell from $8.65B to $8.33B and from 15.1% to 12.4% of revenue. R&D rose modestly to $10.27B but fell from 17.2% to 15.3% of revenue; G&A was nearly flat. Those are real efficiencies. Restructuring and other charges simultaneously rose to $1.84B. The economic interpretation is mixed: Oracle can harvest its installed sales/support base to fund new capacity, but finite overhead leverage is masking direct-cost deterioration. Product economics cannot be inferred from stable consolidated EBIT margin.

6.2 Quality of earnings

FY2026 net income grew 37%, twice operating-income growth. The gap was not recurring operations. Non-operating income rose to $3.55B from $60M, including $2.81B of investment gains—principally a $2.7B realized gain when SoftBank bought Ampere. The reported effective tax rate was 12.6%, versus Oracle’s adjusted 19.9% rate. Management reported $7.63 of FY2026 non-GAAP EPS, but only $6.83 excluding Ampere and Bloom one-time investment gains. Operating income and gain-adjusted earnings are cleaner anchors than reported net income.

Operating cash flow also needs reconciliation. OCF rose to $31.98B, exceeding net income by $14.89B. Depreciation/amortization of $9.29B and SBC of $4.81B are noncash add-backs, while $4.59B came from customer prepayments with a significant financing component. Those prepayments improve liquidity and validate commitment; they are also financing-like timing support. Excluding just that disclosed inflow, OCF would be $27.39B and conventional FCF approximately negative $28.28B.

Quarterly cash flow shows why one good quarter is not an inflection:

Q4’s negative $1.87B was much better than Q3, while IaaS grew 93%. Seasonality, prepayments and construction timing matter, and Oracle says capex should keep rising. The quarter validates conversion; it does not establish self-funding.

SBC remained a real economic cost. FY2026 expense was $4.81B, 7.1% of revenue, up from $2.61B in FY2022. FCF less SBC was approximately negative $28.5B. Subtracting SBC is not GAAP and can double-count compensation if valuation already uses diluted shares; it is still a useful owner-earnings warning because repurchases no longer offset awards.

6.3 Balance sheet and commitments

At May 31 Oracle had $31.89B of cash and marketable securities and $129.54B of notes and other borrowings, producing $97.65B of standard net debt. Debt rose from $92.57B one year earlier. Recognized operating-lease liabilities were another $30.19B; finance-lease liabilities $7.70B; mandatory-convertible preferred stock $4.95B. These should be shown separately rather than hidden inside one ambiguous “debt” number.

Physical capital expanded faster. Net PP&E rose 130% to $99.96B, including $39.97B of construction in progress. Operating-lease right-of-use assets rose 126% to $29.69B. Cash capex divided by depreciation climbed from 2.2x in FY2024 to 5.5x in FY2025 and 7.3x in FY2026. Oracle also acquired $18.25B of operating-lease ROU assets and $4.95B of finance-lease ROU assets in FY2026; cash capex alone understates total additions.

The largest capital item has not yet reached the balance sheet. Oracle disclosed approximately $260B of additional lease commitments, substantially datacenter arrangements expected to commence in FY2027–FY2029 for terms generally lasting 15–19 years. A lessor-borrowing guarantee can reach $3.3B. Year-end unconditional purchase and other obligations were $13.31B, mainly power; after year-end Oracle entered another $19B of five-year cloud-infrastructure purchase commitments. These are not all present debt and should not be added mechanically to enterprise value. They are legally committed future capacity whose value depends on long-duration customer performance.

Oracle’s filing explicitly warns that customer-contract and lease terms or prices may not match. If customers fail to perform or renew, Oracle can remain obligated to sites, power and financing and may be unable to repurpose capacity on equivalent terms. That duration mismatch is more decision-useful than the headline net-debt ratio.

6.4 Returns on capital

Reported ROE is analytically weak because historical buybacks and acquisitions left a thin equity denominator. A rough consolidated ROIC provides a better directional test. Applying Oracle’s 19.9% adjusted tax rate to FY2026 EBIT gives approximately $16.5B of NOPAT. Dividing by year-end borrowings plus Oracle equity less cash/securities yields roughly 11.8%; adding recognized operating-lease liabilities to the denominator lowers it to roughly 9.7%. Comparable FY2025 estimates were approximately 13.9% and 12.3%.

This is an intentionally simple point-in-time calculation, not an audited KPI. It uses year-end rather than average capital, does not allocate capital by segment, and excludes $260B of future leases until commencement. It nevertheless captures direction: after-tax operating profit grew, but funded capital grew faster. Legacy software can still earn exceptional returns while consolidated marginal ROIC declines.

Management’s high-20s project claim can coexist with a 10% group return during ramp if new projects have large construction balances and delayed revenue. It can also overstate true economics if depreciation is added back without replacement capital and shared assets. The test is a multi-period reconciliation: gross project investment, after-tax operating cash before and after depreciation, replacement capex, ramp losses, leases and terminal/renewal economics.

6.5 Financial-quality verdict

The legacy annuity remains excellent and demand is strong, but consolidated quality deteriorated. Revenue +17% and OCI +77% came with gross margin down 470 basis points, cloud/software margin down four points, negative $23.7B FCF, capex/depreciation at 7.3x and rough lease-adjusted ROIC near 10%. Reported net income and OCF flatter recurring economics through investment gains, a low tax rate and financing-component prepayments. Scale is improving revenue; it has not yet improved owner returns.

7. Capital Allocation

7.1 Sources and uses changed completely

Oracle once used excess software cash to shrink the share count. It now uses external capital and customer funding to build infrastructure while preserving the dividend.

Stopping buybacks was rational as valuation rose and funding needs expanded. The issue is the replacement use. Oracle issued $43B par of senior notes in FY2026, received approximately $5B from 6.50% mandatory-convertible preferred stock, and generated $3.35B of net short-term financing tied to capex. Some tranches extend to 2056 and 2066 at coupons of 6.70% and 6.85%, while the relevant hardware turns in years. Interest expense rose 29% to $4.60B.

Management expects approximately $40B more FY2027 debt/equity financing, including the up-to-$20B ATM. No ATM shares had been sold at May 31 and no subsequent filing through August 9 quantified sales, so realized dilution should not be fabricated. Prospective dilution is part of the stated plan. Common shares outstanding already rose 2.6% to 2.880B, and diluted weighted-average shares rose 1.7% to 2.914B.

Oracle maintained the $0.50 quarterly common dividend after year-end. The preferred adds $325M of annual dividend at 6.5%. Paying dividends while FCF is negative is not automatically irrational for a company with a durable annuity and temporary build; it does make the dividend externally financed at the consolidated level. Management has chosen payout continuity and capacity growth over balance-sheet repair.

7.2 What customer funding changes

The $75B of prepaid or customer-supplied hardware is the strongest defense of current allocation. Prepayments improve project cash-on-cash return and contract commitment. Customer-supplied GPUs transfer purchase and residual-value risk. Contractual cost pass-through can protect margin when components change. Oracle may therefore need less balance-sheet capital per revenue dollar than gross datacenter construction suggests.

The filing prevents an “asset-light” conclusion. Standard debt, recognized leases, future leases, power commitments, purchase obligations, construction in progress and FY2027 financing all increased. The $4.59B filed prepayment category is much smaller than $75B because the latter combines different arrangements. Economic risk was transferred and restructured—not eliminated. Investors need the split and matching of each customer contract to Oracle’s fixed commitments.

7.3 Acquisitions and investments

Cerner remains the five-year capital-allocation scorecard. Oracle paid $28.3B cash in June 2022. Four years later it still does not disclose Oracle Health revenue, operating profit, cash flow, invested capital or ROIC. Management expects double-digit FY2027 growth for a new AI-enabled clinical system; no filed segment result validates it. The correct conclusion is not that Cerner failed, but that management has not supplied evidence sufficient to claim value creation.

Oracle’s Ampere exit was successful financially. It received $4.3B for equity, debt and call-option interests and recognized a $2.7B gain when SoftBank acquired the company. That is a positive allocation outcome and a source of cash. It is nonrecurring and should not support a normalized earnings multiple. Oracle also holds 15% of TikTok USDS Joint Venture LLC, but the filing does not disclose invested cost or return expectations.

7.4 Incentives and governance

The latest proxy, filed September 2025, shows Larry Ellison owning 40.6% and pledging 346M shares. Founder ownership creates powerful alignment and decisive strategic control. The pledge and policy exception create a tail governance risk because lenders can influence outcomes outside ordinary corporate processes.

Executive performance design emphasizes share price, at least 30% operating margin for three of eight years, $20B of non-GAAP cloud revenue and $10B of non-GAAP PaaS/IaaS revenue. It contains no explicit FCF, ROIC or leverage gate. That design rewards scale and reported margin without directly charging for the capital used to create them—the exact mismatch the current build exposes.

New CFO Hilary Maxson received a $26M equity grant, 80% time-based and 20% performance-based; the performance portion uses revenue metrics consistent with senior leadership through 2028. No disclosed FCF/ROIC/leverage condition applies. Her call commentary focuses on investment grade and project returns, but incentives remain more growth-oriented than cash-return-oriented.

7.5 Five-year SEC and insider sweep

The trailing five-year corpus contains five 10-Ks, fifteen 10-Qs and 236 Form 4s; all 236 Form 4 index entries matched local accession files. Across the window, code-S rows represented 39.85M shares and $5.53B of sales; 27.70M shares and $4.01B were identified as 10b5-1 planned. Code-P purchases totaled only 24,543 shares and $2.65M.

In the last 24 months, 13.68M shares were sold for $2.77B; 13.42M shares and $2.70B were identified as planned. Safra Catz accounted for $2.53B, all planned. Two open-market director purchases correct the June report’s “zero purchases” error: Charles Moorman bought approximately $948,000 in February 2025 and Rona Fairhead approximately $112,000 in July 2025. Those are positive signals but tiny beside sales.

On June 24, 2026, Jeffrey Henley exercised 400,000 options at $40.93 and sold the shares for approximately $63.7M under a pre-existing 10b5-1 plan. The planned nature reduces incremental information. The absence of meaningful current-executive buying after the July collapse provides no positive confirmation. The aggregate signal is neutral-to-negative, not because planned selling proves concern, but because small director purchases do not offset it.

The SEC sweep also confirms the current financing facts: the $20B ATM is live, preferred conversion can dilute common shareholders, additional commitments sit outside recognized debt, and no newer proxy has refreshed the incentive design. No material unreported acquisition changes the current thesis.

7.6 Capital-allocation verdict

Current allocation is high-risk and not yet proven to create value despite three genuine offsets: contracted RPO, customer-funded/supplied hardware and the profitable Ampere exit. Oracle issued $43B of notes plus $5B mandatory preferred, plans another $40B of funding, maintained dividends through deeply negative FCF and committed to enormous leases/purchases. S&P’s BBB- downgrade externalizes the balance-sheet cost. Buybacks were correctly stopped; the replacement use—late-cycle AI infrastructure with falling segment margins and undisclosed after-replacement ROIC—has not yet cleared its cost of capital. Cerner opacity and growth-weighted incentives argue against granting an evidentiary benefit of the doubt.

8. Changes and Headwinds — Last Two Years

8.1 The company changed faster than the income statement

Oracle’s last two years contain a strategic discontinuity. FY2024 was still primarily a database, support and applications company: revenue was $53.0B, cloud infrastructure $6.9B, conventional FCF $11.8B and net PP&E roughly $21B. By FY2026, revenue was $67.4B, infrastructure $18.1B, conventional FCF negative $23.7B and net PP&E $100.0B. Operating-lease right-of-use assets reached $29.7B, and unrecognized data-center lease commitments reached approximately $260B. The revenue statement evolved; the capital structure was remade.

The build produced real commercial results. OCI growth accelerated, Oracle delivered more than 1.2 gigawatts in FY2026, RPO rose to $638B, and management said utilization reached 97.5%. Oracle also gained an estimated point of cloud-infrastructure share in Q1. The June 9 analysis correctly identified that demand was not the near-term bottleneck. It underestimated how quickly customer funding structures and new contracts would expand.

The funding model changed after June. Oracle disclosed that most Q3/Q4 large AI contracts involved customer prepayments or customer-supplied GPUs, totaling $75B across the two categories. Bring-your-own-hardware transfers the largest fast-obsolescing component to the customer and can improve Oracle’s project cash return. Prepayment brings forward liquidity and strengthens commitment. These arrangements partially “uncouple” Oracle’s cash outlay from headline capacity. They do not remove Oracle’s $70B FY2027 net cash-capex plan, leases, power and site obligations, network investment or operating risk.

8.2 Margin and cash flow became the headwind

Cloud and software revenue rose 19% in FY2026, while its direct expense rose 56% to $16.9B. Segment margin fell four points to 59%. Management said full-year gross margin fell about five points and expects further FY2027 pressure as datacenters ramp. Oracle held consolidated GAAP operating margin near 31% through operating leverage and sales-and-marketing efficiency, but cost control outside cost of revenue cannot indefinitely offset a lower-margin mix.

Cash flow deteriorated more sharply. OCF rose 54%, assisted by $4.6B of customer prepayments with significant financing components, yet capex rose to $55.7B and conventional FCF fell to negative $23.7B. FY2027 net cash capex is guided near $70B, and another $40B of debt/equity financing is planned. The company is effectively borrowing and issuing equity while preserving a $2.00 annualized common dividend. That can be rational if the new asset cohorts earn attractive returns; the filed group economics do not yet demonstrate it.

8.3 Credit crossed the June line

S&P downgraded Oracle on July 9 from BBB/A-2 to BBB-/A-3 with a stable outlook. The stable outlook reduces immediate downgrade pressure, but BBB- is the last investment-grade notch. This directly triggered the June report’s stated bear condition of a ratings downgrade. S&P’s prior analysis expected negative free operating cash flow of roughly $20B–$25B annually over the next three years and adjusted leverage in the high-4x area. Oracle’s customer funding and $20B ATM improve flexibility; they do so partly by transferring risk to common shareholders through dilution.

Credit-default-swap spreads reached record levels in July according to market reporting, showing that credit investors—not merely equity momentum—are repricing the build. That signal should not be confused with imminent insolvency: Oracle has a large recurring software annuity, $31.9B of cash and short-term investments, customer financing, and access to debt/equity markets. It does show that the marginal cost and availability of capital have become thesis variables.

8.4 Leadership, labor and disclosure

Oracle is executing under a co-CEO structure. Clayton Magouyrk represents OCI and infrastructure; Mike Sicilia represents applications and industries; Larry Ellison remains chairman, CTO and the dominant strategic voice. Hilary Maxson, the new CFO, used the Q4 call to foreground project returns, customer funding and commitment to investment-grade credit. This division can put accountable operators over both sides of the barbell. It can also create coordination and succession risk during the largest capital program in company history.

FY2026 restructuring charges rose to $1.84B from $374M, including $823M in Q4. Cost reductions support near-term operating margin while infrastructure gross costs ramp. They are not evidence that the new revenue mix has attractive unit economics. The report avoids unverified headcount estimates; the filing’s restructuring charge is the sourced fact.

Disclosure improved on project economics and funding structure but remains insufficient for full underwriting. Oracle still does not publish OCI contribution profit, database-cloud dollars, customer concentration, the cash-versus-hardware split within $75B, conventional project ROIC, Oracle Health economics, or replacement-capex assumptions. Management’s high-20s project-return estimate is a useful starting point precisely because investors can now specify what reconciliation is missing.

8.5 Prior-thesis scorecard

Changes verdict: commercial proof and funding quality improved, but margin, free cash flow and credit deteriorated. The June thesis therefore strengthened on demand and database distribution, worsened on shareholder economics, and remains unresolved on normalized project returns.

9. Risk Analysis

The risks interact. A customer delay is manageable if Oracle can redeploy capacity and refinance cheaply. It becomes severe if industry supply is loosening, hardware is aging, leases are fixed and the credit rating falls. The same positive feedback works in reverse: prepaid or customer-supplied equipment, high utilization and long contracts can make a levered build profitable if capacity arrives on time and renewal economics hold.

The most important downside is not a collapse in AI demand. It is demand that grows but slower than capacity, producing revenue without adequate return. In that state RPO can convert, headline growth can remain strong, and shareholder value can still be impaired through depreciation, interest, dilution and stranded fixed commitments.

Catastrophic loss remains remote because the database/support annuity, applications, customer base and operating cash flow provide substantial enterprise value. A large equity impairment is plausible through a less dramatic path: margin settles below software-era expectations, free cash flow remains negative, a further downgrade raises funding cost, and the ATM/share conversion spreads the eventual recovery over more shares.

10. Valuation Discussion — Embedded Expectations

10.1 Capitalization and EV bridge

At $147.02 and 2.880B common shares, equity value is $423.42B. Oracle’s unusual mix of funded debt, preferred stock, recognized leases and future commitments requires more than one EV convention:

The $260B of future leases for facilities not yet commenced is not current EV; neither are $13.31B of year-end purchase obligations or $19B of subsequent commitments. Adding every future nominal payment to current debt would be economically wrong. Omitting them from analysis would be equally wrong. Conventional EV is the comparable headline measure, lease-adjusted EV captures recognized claims, and future commitments belong in scenarios and discount rates.

Lease-adjusted EV/EBITDA also has a matching problem: operating rent remains in operating expense while liabilities are added to EV, and finance-lease economics split between depreciation and interest. It is a sensitivity, not a purer truth.

10.2 Current multiples

The apparent value depends on the denominator. An 18.3x guided non-GAAP P/E looks moderate for 34% guided revenue growth. That EPS excludes SBC, acquired-intangible amortization and restructuring and does not charge capex. EBITDA adds back depreciation precisely when capex is 7.3x depreciation. The owner-cash lens shows no current yield. None is inherently “correct”; each answers a different question, and the capital-cycle question demands EBIT, FCF and invested capital more than EBITDA.

Own-history context confirms a large but incomplete reset. AZI places P/E at 25.09x and the 48.0th percentile of Oracle’s own history, P/B at the 42.9th, P/S at the 75.9th and the three-factor composite at the 55.6th. P/B is weakened by negative tangible book/thin equity. P/S matters because it remains upper quartile just as the revenue mix becomes lower-margin and more capital-intensive. These are Oracle-versus-itself percentiles, never cross-sectional ranks.

10.3 Peer cross-read

Oracle is cheaper than Microsoft and Alphabet on revenue and equity earnings, but its EBITDA multiple is close to Microsoft/Amazon despite much weaker balance-sheet and FCF protection. SAP is the cleanest applications-core contrast: similar guided earnings valuation in the June snapshot, positive FCF and much lower capital intensity. Oracle’s growth is faster. Its discount is not automatically a mispricing; leverage, concentration, dilution and negative cash conversion explain much of it. Peer definitions differ, so the table is directional rather than a mechanical target. (Microsoft FY2026; Amazon Q2 2026; Alphabet Q2 2026; SAP Q1 2026)

10.4 Reverse DCF

The reverse DCF asks what current EV requires, not what the stock “should” be worth. The sensitivity fixes FY2027 revenue at management’s $90B guide and an analyst FY2027 FCF margin of negative 35%, consistent with approximately $70B of net cash capex and no evidence that OCF will fund it. Revenue then grows at a constant rate through FY2036 while FCF margin recovers linearly. WACC is 9% and terminal growth 3%.

At a 20% normalized FCF margin, conventional EV requires approximately 14%–16% annual revenue growth for the nine years after FY2027, depending on cash-margin recovery speed. If normalized margin is only 15%, the hurdle approaches 19%–20%. A 25%–30% margin lowers the growth requirement substantially; that is the unproven input while gross margin falls and future leases await commencement. A one-point WACC change or a different FY2027 cash trough moves the answer materially.

An earnings-power cross-check points the same way. Capitalizing approximate FY2026 NOPAT of $16.5B at 8%–9% produces unlevered earnings power of roughly $183B–$206B before growth value, less than 40% of lease-adjusted EV. The method undervalues a company in a purposeful build and ignores franchise reinvestment; its purpose is to show that most current enterprise value depends on growth and recovered returns, not present cash earnings.

10.5 Scenario framework

Each scenario holds FY2027 revenue at $90B and FCF margin at negative 35%. DCF support is shown only as a percentage of current conventional EV, not as equity value or a share-price target.

The dispersion is the conclusion. Current EV is supported near the base only by $90B FY2027 revenue, nine additional years of 15% growth and recovery to a 20% cash margin. The bear still assumes strong growth; it fails through slower cash recovery rather than demand collapse. The bull needs exceptional duration and cash economics above today’s infrastructure-heavy mix. Dilution does not alter FCFF enterprise value, but it changes each common share’s claim.

10.6 Valuation verdict

The June richness thesis weakened materially; the cash-economics thesis did not. Current multiples no longer assume perfection on guided earnings, yet the reverse DCF still assumes long-lived growth and a substantial FCF recovery. Valuation is best described as conditionally reasonable on management’s revenue guide, demanding on normalized owner cash, and unusually sensitive to lease, dilution and margin assumptions. The market has stopped pricing Oracle like a proven fourth hyperscaler. It has not priced the infrastructure build as a failed utility.

11. Variant Perception

11.1 What consensus appears to believe

The observed price path suggests consensus has moved through three regimes. At the September 2025 peak, the market treated RPO as near-certain high-value revenue and capitalized Oracle as an emerging hyperscaler. By July 2026, the market emphasized leverage, dilution, capex and customer concentration. The current rebound indicates a third view: the collapse may have over-discounted financing risk given customer-funded hardware, rapid OCI conversion and the durability of the software core.

The middle consensus now appears to be:

1. FY2027 revenue near Oracle’s $90B guide is plausible because contracted demand and capacity delivery are strong.
2. Infrastructure will dilute gross margin and keep FCF negative, but funding remains available.
3. Customer prepayments and supplied GPUs make the build less dangerous than headline capex suggests.
4. Database and applications provide a valuable floor.
5. Normalized OCI returns will eventually be positive, although the timing and magnitude are unclear.

This is not a straw man. Each point has evidence. The variant question is whether points three and five are strong enough to offset the fixed-capital and credit structure.

11.2 Strongest bull case

The strongest bull argument starts with contract quality, not a generic AI TAM. RPO is $638B and still grew $85B sequentially. IaaS growth reached 93% in Q4. Customers are not merely signing paper: Oracle says they prepaid for or supplied $75B of hardware portions, GPU utilization is 97.5%, and expiring capacity is rapidly resold. Four customers signed contracts above $8B in Q4, improving diversification at the margin. Bring-your-own-hardware transfers the fastest-obsolescing asset to sophisticated customers while Oracle earns on datacenter, networking and operations.

The bull then separates Oracle from a commodity lessor. Oracle operates very large clusters, secured scarce power, can pass through uncertain component costs, and uses its database franchise to attach higher-margin services. Multicloud grew 404%, letting Oracle monetize installed data wherever customers choose to compute. Management’s high-20s steady-state project return, if reconciled to true after-replacement economics, would exceed Oracle’s cost of capital by a wide margin.

Finally, the equity has reset. The own-history composite multiple fell near the middle of its range, guided non-GAAP P/E is approximately 18x, and the price is less than half its peak. If FY2027 revenue reaches $90B, infrastructure margins improve after ramp, financing peaks and FCF inflects later in the plan, today’s trailing multiples overstate normalized valuation. The database/support annuity absorbs the interim risk.

11.3 Strongest bear case

The strongest bear does not dispute revenue. It argues that the market still confuses contracted growth with value creation. Only 12% of RPO converts within a year, while more than half lies beyond three years. Oracle commits to sites and power for 15–19 years, debt and preferred capital remain outstanding, and accelerator economics can change within a few product generations. Customer-supplied GPUs reduce purchase cost but may also reduce Oracle’s differentiation to datacenter operation, where sophisticated buyers can negotiate renewals.

The filed economics are already adverse. Direct cloud/software expense grew far faster than revenue; segment margin fell four points; FCF was negative $23.7B; net PP&E more than doubled; restructuring supported consolidated margin; and S&P cut the rating to BBB-. Oracle will issue additional debt and equity before normalized returns are observable. Adding depreciation back to project returns while omitting replacement capex is exactly how a capital-heavy business can appear profitable in the ramp while destroying value over the asset cycle.

The bear also questions the “floor.” Database support is durable, but net-new workloads increasingly use open-source and cloud-native data platforms. Applications growth trails several peers. The core can remain healthy while becoming a smaller share of capital and enterprise value. If infrastructure deserves an asset-heavy multiple and the core is burdened by consolidated debt, a mid-history headline P/E is not necessarily cheap.

11.4 Variant conclusion

The differentiated conclusion sits between “RPO is fake” and “RPO is value.” The contracts are meaningful, conversion is visible, and customer funding materially improves the outcome distribution. The remaining variant is duration-adjusted and capital-adjusted backlog value. Oracle can beat revenue forecasts and still miss the return embedded in enterprise value. Conversely, even mediocre gross margins can produce acceptable equity returns if customers fund hardware, utilization holds, pricing passes through costs and financing intensity falls. The next two years should be judged on margin, replacement capital and credit—not another headline contract.

12. Fact vs. Interpretation

13. Open Questions

1. Reconcile the high-20s project return to conventional after-tax ROIC after depreciation, replacement capex, ramp losses, shared network/datacenter assets, leases, working capital and financing.
2. Split the $75B between cash received, enforceable future prepayments and customer-supplied equipment. Who owns the hardware, bears residual risk and controls redeployment?
3. What percentage of RPO belongs to the largest customer and top five customers? What are the termination, delay and minimum-payment protections?
4. Why did the current-RPO share remain at 12% despite $85B of sequential additions? What is the dollar conversion cadence by fiscal year?
5. What are OCI gross profit, contribution profit, depreciation and allocated invested capital by cohort?
6. What does renewal pricing look like for large GPU clusters delivered in 2024–2026, and how portable are workloads at contract expiry?
7. How much of the $70B FY2027 net cash-capex guide depends on timing rather than durable customer funding? What is gross capex before offsets?
8. How much of $260B of unrecognized lease commitments matches customer contracts in duration, price, credit quality and commencement?
9. What conditions would preserve investment grade if FCF remains negative through FY2028? How much ATM equity is assumed at what average price?
10. What are absolute database-cloud revenue, multicloud revenue, net retention and net-new workload share?
11. Why is applications growth not accelerating with the claimed OCI/database halo? What paid-agent or token consumption is visible?
12. What are Oracle Health revenue, margin, cloud-migration rate and acquisition-level return since Cerner?
13. How much of FY2026 operating-margin resilience came from restructuring and sales/marketing reductions versus sustainable gross economics?
14. What useful lives are assigned to accelerators and related equipment, and what historical impairment/replacement experience supports them?
15. How will the next proxy align co-CEO/CFO compensation with FCF, leverage and conventional ROIC rather than revenue and share price?

14. What Must Be True

Bull case

The bull requires more than revenue growth:

• RPO converts without material cancellation; current-RPO dollars rise even if the percentage remains low.
• FY2027 revenue reaches roughly the $90B guide and infrastructure growth remains broad enough that no single customer controls the trajectory.
• Customer-prepaid and supplied hardware meaningfully reduce Oracle-funded accelerator capital without surrendering margin.
• Cloud/software segment margin bottoms as datacenter ramps mature; depreciation and power do not overwhelm gross profit.
• Conventional project ROIC after replacement capital exceeds the cost of capital, not merely management’s depreciation-added metric.
• Net cash capex and external financing peak, credit remains investment grade, and FCF begins a credible recovery.
• Database support remains stable, multicloud database grows in absolute dollars, and applications regain at least mid-teens momentum.
• Dilution from SBC, mandatory preferred and the ATM does not absorb the incremental equity value.

Bull falsification test: the case fails if current-RPO conversion slips materially, cloud/software margin continues falling through FY2028, or Oracle cannot show a path to positive FCF without recurrent equity issuance. A further agency downgrade while capex remains elevated is a direct failure.

Bear case

The bear requires the infrastructure economics to remain structurally inferior:

• Industry capacity catches demand, weakening utilization, renewal pricing or cost pass-through.
• Customer funding proves less cash-like or less durable than implied, leaving Oracle with fixed leases and purchase obligations.
• Depreciation, replacement capex and financing costs keep consolidated returns near or below the cost of capital.
• Database and application growth fail to offset infrastructure’s lower margin and capital burden.
• Credit deterioration raises financing cost and forces more dilution or a reduction in strategic flexibility.
• The market ultimately values the core as mature software and the growth engine as an asset-heavy utility rather than granting one blended software multiple.

Bear falsification test: the case fails if Oracle produces at least two to three quarters of OCI/RPO conversion with rising cloud/software margin, independently reconcilable after-replacement project returns, declining net-capex intensity and a credible positive-FCF trajectory while the rating stabilizes. Strong revenue alone does not falsify it.

15. Source Appendix

Financial statement facts are anchored to Oracle’s FY2026 Form 10-K and Q4/FY2026 release. Credit status is anchored to S&P’s July 2026 rating action. Price history comes from AZI; factor diagnostics come from FactorsToday. Industry, competitor, regulatory, governance and filing-corpus sources—with dates, limitations and reconciliation notes—are listed in Appendix B — Source Appendix.

APPENDIX A — Standard Diligence Questionnaire

General

What thoughtful questions have other investors asked about this company?

The highest-value questions are not whether Oracle can grow OCI revenue—the FY2026 result already showed that it can—but whether the growth creates value after the full capital bill:

1. What is the conventional after-tax return on each large AI-infrastructure project? Management estimates a “high-20s” steady-state project return using after-tax operating margin plus depreciation divided by gross investment. Interpretation: that is closer to a pre-replacement cash return than conventional ROIC. Investors should ask for a reconciliation including ramp losses, leases, shared networking and data-center capital, working capital, depreciation, replacement capex, and renewal assumptions.
2. What is inside the $75 billion of prepaid/customer-supplied hardware? Fact: the figure combines cash prepayments and customer-supplied GPUs; the audited Form 10-K separately identifies only $4.6 billion of FY2026 customer prepayments with a significant financing component. Investors should ask for the cash-versus-equipment split, ownership, cancellation rights, residual-value risk, and remaining Oracle-funded amount.
3. How quickly will $638 billion of RPO become revenue and cash? Fact: only 12% is expected within twelve months; 34% in months 13–36, 34% in months 37–60, and 20% thereafter. The current percentage did not improve from Q3 even as headline RPO rose by $85 billion.
4. How concentrated is backlog by customer? Oracle says four customers signed contracts exceeding $8 billion in Q4, but it does not disclose the largest counterparty, cancellation protections, or revenue concentration. Large-customer diversification improved at the margin but remains unquantified.
5. Can Oracle disclose database-cloud dollars and retention? Multicloud AI Database revenue grew 404% and database cloud 29% in Q4, but both bases are undisclosed. The key question is whether Oracle wins net-new workloads or chiefly preserves captive installed-base workloads.
6. Why should OCI’s current scarcity economics persist? Management reported 97.5% GPU utilization and rapid resale of expiring capacity. Investors should test renewal pricing and utilization after AWS, Azure, Google, Oracle, and neocloud capacity arrives.
7. Is Oracle Health creating value after the $28.3 billion Cerner acquisition? Oracle forecasts double-digit FY2027 Health growth but does not report Health revenue, margin, or return on acquisition capital separately.
8. Can the database/applications “halo” be measured? Oracle argues that AI infrastructure pulls through database and applications. Applications grew only 10–11%, trailing several current peers, so attach rates and incremental ARPU are needed.
9. How much financial flexibility remains after the downgrade? S&P lowered Oracle to BBB-/A-3 in July 2026, one notch above speculative grade, as the company plans another approximately $40 billion of FY2027 debt/equity funding.

Sources: Oracle FY2026 Form 10-K, Oracle FY2026 results, and Q4 FY2026 call transcript, accessed 2026-08-09.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low?

Interpretation: consolidated earnings are not at a conventional demand-cycle peak, because contracted OCI capacity is still ramping and Oracle guides FY2027 revenue to $90 billion. They may, however, be near a scarcity-economics high for new AI-infrastructure contracts. Demand exceeds available supply, GPU utilization is 97.5% by management’s measure, and Oracle can pass through uncertain component costs. Simultaneous hyperscaler and neocloud capacity additions make those conditions unlikely to remain permanently favorable.

The legacy database-support and applications earnings are much less cyclical. Mission-critical renewals and subscription contracts stabilize revenue. The faster-growing infrastructure mix introduces data-center delivery, component-price, power, customer-credit, and capital-cycle sensitivity.

Are results driven by the external environment or internal actions?

Both. Fact: external AI-compute demand drove FY2026 IaaS revenue up 77% and RPO up 363%. Scarce GPUs, power, and data-center capacity support utilization and customer prepayments. Fact: Oracle’s internal actions—securing capacity, constructing more than 1.2 gigawatts in FY2026, offering BYOH/prepayment structures, extending Oracle Database into rival clouds, and cutting sales-and-marketing costs—determined how much demand converted into revenue and operating income.

Interpretation: demand is externally supplied; Oracle’s execution captures it. The durability of returns depends on internal contracting and operating discipline after external scarcity fades.

How stable are revenues?

Fact: cloud and software produced $58.5 billion, or 87% of FY2026 revenue. Software support alone was $19.8 billion and grew 1%; cloud applications were $15.9 billion and grew 11%. These streams are recurring or ratably recognized and relatively stable. Software licenses, hardware, and services are more transactional.

RPO provides visibility but must be duration-adjusted. Of $638 billion at May 31, only approximately $76.6 billion is expected within twelve months. Long-dated commitments lower near-term demand uncertainty but increase delivery, customer-credit, financing, and re-contracting risk.

What is the outlook for products and services?

• Cloud infrastructure: Fact: Q4 IaaS grew 93%; Oracle guides FY2027 total revenue growth of 34% in constant currency and FY2027 net cash capex around $70 billion. Capacity delivery should keep recognized growth high. Interpretation: margins and owner cash flow are the limiting factors, not near-term demand.
• Database: Fact: database cloud grew 29% and Multicloud AI Database 404% in Q4, while software support grew 1% for FY2026. Interpretation: multicloud is a credible defensive growth engine because it lets customers retain Oracle Database inside AWS, Azure, or Google Cloud.
• Applications: Fact: cloud applications grew 10% in Q4 and 11% for the year; SaaS deferred revenue grew 16% in Q4. Interpretation: growth is healthy but trails SAP Cloud ERP and Workday subscription growth; embedded AI has not yet produced a visible growth premium.
• Oracle Health: management expects double-digit FY2027 growth from a new AI-enabled clinical system. Interpretation: this remains an unverified forecast because segment revenue and margins are not separately disclosed.
• Legacy license/hardware/services: licenses fell 9% in FY2026 as workloads moved to cloud; hardware grew 5% and services 10%. These are supporting, not thesis-driving, lines.

How big will the markets be—growing or shrinking, domestic or international?

Fact: Synergy Research estimated global cloud-infrastructure spending at $143 billion in Q2 2026, approximately $572 billion annualized, after 35% Q1 growth. Gartner forecasts the DBMS market at approximately $161 billion in 2026 and $255 billion by 2029. Gartner forecasts worldwide software spending of $1.44 trillion in 2026, up 15.1%. Oracle operates globally; cloud, database, and applications demand is international.

Interpretation: the addressable pools are large and growing, but TAM growth does not guarantee returns. Oracle’s realistic database serviceable market is the enterprise operational/relational and hybrid pool where its installed base and Exadata/multicloud deployment matter. In cloud, Oracle’s approximately 4% Q1 share means its near-term revenue runway can be substantial even without approaching the top three, while its scale disadvantage remains material.

Sources: Gartner IT forecast, Gartner DBMS forecast, and Synergy data reported by CRN and ITPro, accessed 2026-08-09.

Business Quality & Competitive Moat

Is the industry getting more or less competitive?

Interpretation: database and ERP competition is stable at the installed-base core but intensifying for new workloads and adjacent applications. PostgreSQL, cloud-native databases, Snowflake, Databricks, AWS, Microsoft, and Google compete for new data workloads. SAP remains the closest large-enterprise ERP rival; Workday, Salesforce, ServiceNow, and Microsoft compete by application category.

Cloud infrastructure is becoming more competitive even while demand exceeds supply. AWS, Azure, Google, Oracle, neoclouds, and frontier laboratories are expanding simultaneously. Current scarcity masks the pricing regime that will emerge when capacity catches up.

How profitable is the business—ROIC and ROE?

Fact: FY2026 GAAP operating income was $20.6 billion and net income attributable to common shareholders approximately $17.0 billion. Cloud-and-software segment margin fell from 63% to 59%. Oracle’s reported ROE is elevated by its historically thin equity base and is not a clean economic-return measure.

Interpretation: the legacy software franchise earns very high returns on modest tangible capital, but consolidated ROIC is being diluted by explosive data-center asset growth. Management’s “high-20s” steady-state project return is not conventional ROIC because it adds back depreciation and does not disclose all allocated capital. Until Oracle reports a reproducible after-tax return including depreciation/replacement economics, infrastructure ROIC remains unknown.

How profitable is the industry; how many competitors and what barriers exist?

Enterprise database and ERP are highly profitable, concentrated incumbent markets. Barriers are migration risk, accumulated schemas and integrations, trained personnel, regulatory certifications, partner ecosystems, and R&D scale. The high-end ERP core is led by SAP and Oracle, with specialists competing around it.

Cloud infrastructure has a small number of scaled providers. Q1 2026 shares were AWS 28%, Azure 21%, Google 14%, and Oracle 4%. Barriers include power, sites, global networks, custom silicon, scarce accelerators, security, compliance, enterprise sales, operational experience, and tens of billions of dollars of annual capital. These barriers protect the oligopoly but do not ensure that every participant earns attractive incremental returns.

Can the business be easily understood?

Operationally, yes: Oracle sells cloud infrastructure, database/software support, enterprise applications, licenses, hardware, and related services. Economically, it is harder because a high-margin recurring software franchise and a capital-intensive infrastructure build are combined in one Cloud and License segment. Oracle does not separately disclose database-cloud dollars, OCI contribution profit, AI-contract concentration, Oracle Health economics, or conventional project ROIC. The absence of those splits is the main analytical obstacle.

Can foreign low-cost labor undermine the business?

Not directly. Software development and support can be globally sourced, but low-cost labor cannot cheaply replicate an installed database estate, enterprise integrations, hyperscale infrastructure, compliance certifications, or the capital required for AI clusters. Global engineering talent can pressure labor costs and improve competitors’ execution; it is not the principal moat threat. Open-source software and cloud-native architectures are more important threats than wage arbitrage.

Do brands matter?

Yes, as enterprise trust rather than consumer preference. Oracle’s name reduces perceived operational and career risk for mission-critical databases, ERP, healthcare, and government systems. Interpretation: brand reinforces customer captivity but is not an independent moat; switching cost, installed data/processes, and service reliability do the economic work.

What is the nature of competition?

• Database: compatibility, reliability, performance, security, deployment flexibility, AI/data features, and migration cost.
• Applications: suite integration, functional depth, implementation ecosystem, AI features, pricing, and total migration risk.
• Infrastructure: price/performance, accelerator availability, network design, geographic/power capacity, platform breadth, reliability, sovereign deployment, and commercial financing.

Cloud competition currently resembles entry/preemption: incumbents build capacity to defend future share while Oracle enters through AI clusters, database adjacency, and sovereign/private deployment. If scarcity ends, the game can shift toward price/quality competition.

What are customers’ switching costs?

Database and ERP switching costs are high and real. Customers face application rewrites, data migration and validation, integration replacement, retraining, downtime risk, regulatory testing, and potential failure of financial/operational workflows. The financial evidence is a $19.8 billion support stream that grew despite declining new licenses and $5.4 billion of annualized cloud revenue created over three years from migrating support customers.

Switching costs are lower for portable net-new cloud workloads and sophisticated frontier-model laboratories that can multi-source compute. OCI contracting and cluster operations can create relationship friction, but renewal behavior after normalized supply is not yet observable.

Moat verdict: Oracle has a durable demand-side captivity moat in installed database workloads and a narrower shared switching-cost moat in applications. Multicloud extends the database moat defensively. OCI does not yet have a proven Greenwald scale moat: 4% share is improving but subscale, and normalized renewal economics remain untested.

Financial Condition & Balance Sheet

Are there assets not fully recognized on the balance sheet?

Yes. The database installed base, accumulated customer schemas/integrations, Java ecosystem, application process knowledge, partner relationships, engineering know-how, and enterprise trust are internally developed and not recorded at reproduction value. Power/site options and signed customer relationships may also have strategic value not captured by book value.

Conversely, $62.3 billion of recorded goodwill largely reflects acquisition premiums, including Cerner, rather than a liquid or independently reproducible asset. Book value therefore both omits valuable internally developed franchise assets and includes acquisition accounting that may not earn its carrying value.

Are there off-balance-sheet liabilities?

Materially so. Fact: as of May 31, 2026 Oracle had $260 billion of additional lease commitments, substantially all data-center arrangements, expected to commence from FY2027 to FY2029 for terms generally lasting 15–19 years; these were not reflected in the balance sheet or lease-maturity table. Recognized operating- and finance-lease payments totaled $41.9 billion and $11.5 billion before imputed interest. After year-end Oracle entered another $19 billion of unconditional five-year cloud-infrastructure purchase commitments beginning in FY2027.

Interpretation: these commitments are central, not footnotes. They increase fixed operating leverage, create asset/customer duration mismatch, and mean conventional net debt understates committed capital. Customer prepayments and BYOH arrangements offset some funding need but do not extinguish lease and infrastructure obligations.

How conservative is the accounting?

Revenue accounting is conventional for cloud/support and license arrangements, but the unusually long RPO duration makes backlog presentation easy to over-interpret. The $75 billion customer-funded/supplied-hardware figure is not an audited cash balance and combines different economics.

Oracle capitalizes infrastructure and depreciates it over estimated useful lives. Risk: actual accelerator and component lives can be shorter than accounting lives; Oracle itself warns that changing strategy, technology, performance, or failure rates may require shorter lives, retrofits, replacement, or impairment. Segment margins exclude R&D, G&A, SBC, amortization, restructuring, interest, and other allocations, so the 59% cloud-and-software margin is not operating margin or ROIC. Interpretation: GAAP statements are usable, but management’s segment and project-return measures require conservative adjustment.

How CapEx-hungry is the business?

Extremely. Fact: FY2026 operating cash flow was $32.0 billion and free cash flow negative $23.7 billion, implying approximately $55.7 billion of cash capex. Net PP&E rose from $43.5 billion to $100.0 billion; operating-lease ROU assets rose from $13.1 billion to $29.7 billion. Management guides FY2027 net cash capex outlay to approximately $70 billion, excluding $20–25 billion of customer-prepayment/timing effects included in reported capex.

This is the defining change in Oracle’s business quality. A formerly asset-light software company is becoming one of the world’s most capital-intensive infrastructure builders.

Capital Allocation & Management

How much free cash flow does the business generate; how is it used; what is the philosophy?

Fact: FY2026 operating cash flow was $31.977 billion, but conventional FCF was negative $23.7 billion after infrastructure capex. Oracle paid $5.787 billion of dividends, repurchased only $93 million of common stock, raised approximately $43 billion of debt and $5 billion of preferred equity, and ended with $31.3 billion of cash.

Interpretation: the current philosophy is to prioritize OCI capacity, preserve the dividend, minimize discretionary buybacks, and fund the gap with debt, mandatory-convertible preferred equity, customer prepayments, BYOH, and an ATM program. This is aggressive growth capital allocation, not balanced capital return. The decisive test is whether infrastructure returns exceed the fully loaded cost of capital after replacement investment.

Have there been significant acquisitions recently?

The most significant recent strategic acquisition remains Cerner, acquired in June 2022 for approximately $28.3 billion. No similarly sized FY2026 acquisition defines the current period. Oracle sold its Ampere interest when SoftBank acquired Ampere in November 2025 and recorded investment gains; those gains should not be treated as recurring operating earnings.

Interpretation: Cerner’s strategic adjacency—healthcare workflows, data, and OCI—makes sense, but value creation remains unproven because Oracle Health revenue, margin, and return on capital are not separately reported.

Is the company buying back shares?

Only minimally. Fact: repurchases fell from $1.2 billion in FY2024 to $600 million in FY2025 and $93 million in FY2026. The authorization has no expiration and no required amount. Interpretation: the historical buyback is effectively suspended as infrastructure consumes capital; investors should not underwrite meaningful share-count support.

Is Oracle issuing large amounts of new shares to insiders?

Fact: FY2026 stock-based compensation was $4.811 billion, and Oracle issued shares under compensation and employee purchase plans. It also issued $4.954 billion of 6.50% mandatory-convertible preferred stock and authorized an ATM common-equity program of up to $20 billion. Those financing issuances are not insider compensation, but they create potential common dilution.

The filed-count-controlled sweep matched all 236 Form 4s since 2021-08-09. Over the last 24 months, insiders reported 13.678 million shares of sales for approximately $2.765 billion; 13.419 million shares and $2.701 billion were identified as 10b5-1 planned. Two directors made open-market purchases totaling only approximately $1.06 million: Charles Moorman in February 2025 and Rona Fairhead in July 2025. Interpretation: the signal is neutral-to-negative, not “zero purchases.” Small director buys are outweighed by planned liquidity and the lack of meaningful current-executive buying after the 2026 collapse; planned sales are less informative than discretionary sales.

What is the compensation policy for directors and management?

Oracle uses salary, annual incentives, and large equity awards; FY2026 SBC was $4.8 billion. The 2025 proxy describes performance and time-based equity and longstanding founder control. Interpretation: equity aligns executives with share value, but public disclosure does not provide an equally prominent, independently verifiable infrastructure-ROIC or FCF hurdle commensurate with the current capital program. The next proxy should be checked for changes under the new co-CEO/CFO structure.

What motivates management?

Larry Ellison remains chairman, chief technology officer, founder, and an approximately 40% owner, making long-term equity value and strategic legacy powerful motivations. Co-CEOs Clayton Magouyrk and Mike Sicilia are identified with OCI and applications/industries execution, respectively; new CFO Hilary Maxson has emphasized investment-grade credit and project returns.

Interpretation: insider ownership creates genuine alignment but also governance concentration. Management is motivated to establish Oracle as a fourth hyperscaler and AI platform; that ambition may encourage capital deployment beyond what a purely FCF-oriented owner would choose.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer?

No. Fact: Oracle Corporation is a Delaware corporation listed directly on the New York Stock Exchange under ORCL. It is not an ADR, MLP, or K-1 issuer.

What is the dividend policy?

Fact: Oracle declared a $0.50 quarterly common dividend after Q4, or $2.00 annualized if maintained. FY2026 common dividends declared were $2.00 per share and cash dividends paid totaled $5.787 billion including preferred dividends. Dividend changes remain at board discretion.

How profitable is the business?

Fact: FY2026 revenue was $67.357 billion; GAAP operating income $20.606 billion; common net income approximately $17.0 billion; cloud-and-software segment margin 59%. GAAP operating margin was approximately 30.6%. The core software franchise is more profitable than the consolidated average, while fast-growing infrastructure dilutes gross and segment margins.

Is net income diverging from cash from operations?

Fact: FY2026 net income was $17.087 billion and operating cash flow $31.977 billion. The positive gap reflects $9.294 billion of depreciation/amortization, $4.811 billion of SBC, $4.592 billion of deferred revenue from customer prepayments with significant financing components, and other working-capital/noncash adjustments, partly offset by investment gains and receivable growth.

Interpretation: cash conversion before capex is strong, but the owner-economics conclusion reverses after capex: FCF was negative $23.7 billion. CFO minus net income is not evidence of superior earnings quality when depreciation and customer funding accompany an even larger infrastructure investment.

Current market context

Fact: Oracle closed at $147.02 on August 7, 2026. Rebuilt from approximately 2.88 billion shares, equity value was approximately $423 billion. Conventional enterprise value was approximately $526.6 billion; adding $37.9 billion of recognized operating- and finance-lease liabilities produced lease-adjusted EV of approximately $564.5 billion. On the lease-adjusted measure, that equaled roughly 8.4 times FY2026 revenue, 27.4 times GAAP operating income, and 18.9 times operating income plus D&A; conventional FCF was negative. The $260 billion of uncommenced future leases is disclosed separately and is not booked into EV. These are descriptive market facts, not a recommendation or target.

Risks & Downside

What factors would cause the stock to decline?

• RPO conversion delays, cancellations, customer nonperformance, or disclosed concentration.
• OCI margins remaining low or falling as infrastructure becomes a larger revenue share.
• FY2027 capex, leases, or funding exceeding plan; additional credit downgrades or a higher cost of capital.
• Customer prepayments/BYOH proving smaller, less cash-like, or less protective than assumed.
• AI infrastructure supply overtaking demand, reducing utilization and renewal pricing.
• Accelerators becoming obsolete faster than depreciation schedules.
• Database-support erosion or failure of multicloud to win beyond the installed base.
• Applications growth continuing to trail SAP, Workday, Salesforce, or Microsoft.
• Cerner/Oracle Health failing to return to growth or requiring impairment.
• Cybersecurity, privacy, AI, healthcare, export-control, antitrust, or data-sovereignty actions.
• Dilution from preferred conversion, SBC, and the ATM program.

Is there a risk of catastrophic loss?

Yes, but low in probability and potentially high in impact. A major cloud or healthcare-data breach, prolonged outage, loss of critical AI customer payments after Oracle committed to long-duration capacity, power/supply-chain disruption, or regulatory exclusion from important markets could produce large liabilities and impairments. The mismatch between customer-contract terms and 15–19-year data-center leases is especially important: Oracle warns that it may be unable to re-lease or repurpose capacity if customers do not renew.

What is the chance of a total loss?

Interpretation: remote. Oracle owns a durable database-support annuity, sticky applications, a global customer base, valuable intellectual property, and substantial operating cash flow. Total loss would require simultaneous destruction of the core franchise, failure of the infrastructure build, and inability to refinance or restructure commitments. The more realistic severe downside is not zero but a large equity impairment from lower margins, stranded capacity, dilution, and leverage.

Recent News & Events

Has the business environment changed recently?

Yes. Fact: Q4/FY2026 confirmed OCI acceleration, $638 billion of RPO, customer-funded hardware structures, negative $23.7 billion FCF, and another large FY2027 funding program. Fact: S&P downgraded Oracle to BBB-/A-3 on July 9, 2026. Fact: Q2 cloud spending reached approximately $143 billion while AWS, Azure, Google, Oracle, and neoclouds all expanded. Interpretation: revenue visibility improved, but so did financing and capital-cycle risk.

Regulation also moved. The European Commission preliminarily found AWS and Azure should be DMA gatekeepers, which can help OCI as a challenger. EU AI Act transparency duties became applicable on August 2, 2026, while the EU Data Act is lowering cloud switching barriers and eliminating switching/egress charges from January 2027. These measures create both compliance cost and multicloud opportunity.

Have there been significant acquisitions?

No new acquisition comparable to Cerner was identified since the June 2026 report. The capital program, not M&A, is the transformative transaction. Oracle sold its Ampere investment through SoftBank’s acquisition and recognized nonrecurring investment gains.

Have accounting policies changed?

No thesis-changing accounting-policy change was identified in the FY2026 Form 10-K. The important presentation changes are economic and disclosure-related: Oracle emphasizes net cash capex after customer prepayments/timing effects, while GAAP capex and FCF remain much larger; and the $75 billion customer-funded/supplied-hardware disclosure combines distinct arrangements. Investors should keep GAAP cash flow, gross commitments, and funding disclosures separate.

What recent changes occurred in markets, facilities, or management?

• Facilities/capacity: Oracle delivered more than 1.2 gigawatts in FY2026 and expected close to another gigawatt in Q1 FY2027; net PP&E more than doubled. It holds $260 billion of additional unrecognized lease commitments and added $19 billion of post-year-end infrastructure purchase commitments.
• Markets/products: Multicloud AI Database grew 404% in Q4; Oracle continues to deploy database services inside AWS, Azure, and Google. It is extending embedded agents, token bundles, and outcome-based pricing across Fusion and industry applications.
• Funding: FY2026 included $43 billion of debt and $5 billion of preferred equity; FY2027 funding is expected to be approximately $40 billion including a $20 billion ATM.
• Management: Oracle operates with co-CEOs Clayton Magouyrk and Mike Sicilia, with Hilary Maxson newly serving as CFO for the Q4 call. Founder Larry Ellison remains chairman and CTO.
• Credit: S&P’s BBB- downgrade is the clearest external change in the financial-risk assessment.

Overall diligence conclusion: Oracle’s database and applications franchises remain durable, understandable, and financially validated by customer captivity. The new Oracle is nevertheless dominated at the margin by an infrastructure program whose revenue is contracted but whose normalized return is not independently demonstrated. The central diligence discipline is to separate RPO from current revenue, customer-supplied equipment from cash, management’s pre-replacement project-return metric from conventional ROIC, and operating cash flow from free cash flow.

APPENDIX B — Source Appendix

Company: Oracle Corporation (NYSE: ORCL)

Report date: 2026-08-09

Access date for all sources below: 2026-08-09 Scope: financial statements, governance, price/momentum/factor positioning, the five-year event map, post-2026-06-09 news, industry and peer evidence, and falsification-test status. Primary sources precede aggregators. Management statements are hypotheses unless supported by filings or realized financial results.

Primary company and regulatory sources

Primary competitor, industry and regulatory sources

Market-size and share research

Credit, market-event, and corroborating sources

Quantitative third-party sources