Oracle Corporation

RPO is the second battlefield. The headline number is enormous, but the quality of that number matters more than the size. Oracle’s Q4 FY26 release said 76.56 billion USD of the 638 billion USD RPO is expected to be recognized in the next year, which means only about 12% converts within 12 months. Oracle’s March 2026 10-Q said 12% would be recognized over the next twelve months, 31% over months 13–36, 35% over months 37–60, and the rest thereafter. None of that is bad by itself, but it does mean investors should stop treating RPO as near-cash. The backlog is long-dated and increasingly infrastructure-heavy, and bound up with counterparties whose credit quality varies widely. Public reporting and credit commentary now point to OpenAI as a particularly important exposure, but Oracle’s own filings do not fully disclose the customer-level concentration needed to prove the exact split. That opacity deserves a valuation discount.

Oracle today is best described as a company in transition whose financial quality has weakened while its growth profile has improved, rather than structural decline, a clean mispricing, or a classic value trap. The underlying software franchise remains healthy and cloud demand is plainly strong. That mix matters for capital-market treatment. At 119.90 USD, Oracle is no longer priced like the euphoric AI winner it was in September 2025 or even in late May 2026. But it is still not cheap enough to erase the financing, concentration, and execution risks now embedded in the model.

My independent conclusion differs in emphasis from the previous report. The old “Watch” call was right on price. Today the stock sits much closer to fair value than it was, but the business has also become riskier in ways the market is right to care about. I move the rating up one notch to Hold. The reason is that the valuation reset has already absorbed a large part of the new financing anxiety; Oracle itself has not become safer. The valuation bands still come down relative to the previous report, because the evidence since May 20 has shown that the OCI and AI push carries more balance-sheet strain, more long-duration contractual risk, and less near-term cash conversion than the old bands allowed for.

Company vertical history

Origins and listing path

Oracle was founded in 1977 by Lawrence J. Ellison, who remained CEO until September 2014 and is still executive chair and chief technology officer. The company existed to do one thing: commercialize the relational database model at a time when enterprise data management was becoming a strategic bottleneck. Oracle did not invent the relational model. Its early edge was productizing it and selling it aggressively as a general-purpose system for business computing. Oracle’s own investor FAQ says the IPO took place on March 12, 1986, with 2.1 million shares offered at 15 USD. In public markets the original story was straightforward: Oracle was a database company in a world where databases were becoming the central nervous system of corporate computing.

The modern Oracle story only makes sense if one remembers what kind of customer base that original database machine built. Oracle sells into mission-critical workloads, not consumer curiosity. Once the database became deeply embedded in ERP, transaction systems, and regulated workloads, switching costs turned operational rather than aesthetic, which is what let the company spend the next three decades layering applications, support, hardware, and eventually cloud around the original data estate. The installed base funded the future. It still does.

The stages that mattered

The first stage was database establishment and scale-up: Oracle’s rise from relational database specialist to one of the core vendors in enterprise IT. The durable impact of that phase is the installed base itself. Without it, the company would not have had the pricing power, renewal economics, or cross-sell opportunities that later funded applications and cloud. Even today, Oracle’s own business description still centers on enterprise applications and infrastructure serving enterprise IT needs across on-premise and cloud deployment models.

The second stage was consolidation through M&A. Oracle’s 2004 agreement to buy PeopleSoft for about 10.3 billion USD was a decisive turn that changed Oracle from a database-centric company into a broader enterprise-application consolidator. The deal was strategically important: it let Oracle close the gap with SAP in applications and established the company’s long-running playbook: use balance-sheet strength, installed-base credibility, and aggressive M&A to expand the product stack and hold customers inside a wider enterprise architecture. Later deals such as Sun Microsystems and, much later, Cerner extended that logic.

Sun made Oracle more vertically integrated. The company was no longer only software sitting atop someone else’s hardware, which mattered for Exadata, engineered systems, and Oracle’s broader argument that customers could buy performance, security, and reliability as an integrated stack. Cerner did something different. The 2022 acquisition pulled Oracle deeper into healthcare workflows and data, a large but operationally demanding market. It did not change the whole company overnight, but it reinforced Oracle’s shift from pure software vendor to enterprise platform owner in industries where data gravity, compliance, and long contracts matter.

The third stage began with the cloud transition and Larry Ellison’s 2014 handoff of the CEO role. Oracle did not win the first wave of public cloud mindshare. AWS and Microsoft reached scale faster and shaped developer ecosystems more effectively. Oracle’s answer was to focus on enterprise workloads, database performance, security isolation, and increasingly multicloud deployment rather than to out-AWS AWS in general-purpose cloud breadth. The 2014 leadership change and the 2025 leadership reshuffle should be read in that context: the company eventually pushed operators with deeper OCI and applications roots to the top as AI infrastructure moved from side business to capital-markets story. Clay Magouyrk and Mike Sicilia became co-CEOs in September 2025, with Safra Catz moving to executive vice chair; Hilary Maxson became CFO in April 2026. None of that was cosmetic: it signaled that OCI had become central to Oracle’s identity.

The fourth stage is the current one: Oracle as a software company trying to become a serious AI infrastructure landlord without losing the economics of software. It is the most ambitious phase in the company’s history since the database era. Oracle’s Q1 FY26 RPO was already 455 billion USD; Q2 reached 523 billion USD; Q3 reached 553 billion USD; and Q4 reached 638 billion USD. OCI growth accelerated from 68% in Q2 to 84% in Q3 to 93% in Q4. At the same time, Oracle said in March 2026 that many AI contracts were supported either by upfront customer prepayments or by customers purchasing and supplying the GPUs themselves, and in June 2026 the company guided to as much as 95 billion USD of fiscal 2027 capex with about 70 billion USD of net cash outlay by Oracle after customer reimbursements. This phase has already changed the company’s fate in one respect: Oracle is no longer being valued or punished as a sleepy legacy software vendor.

Price and valuation history

Oracle’s modern capital-market history has three distinct moods. For years it traded as a mature large-cap software name: solid margins, buybacks, dividends, limited excitement. Then the market began to reward the cloud transition more seriously as OCI growth improved and multicloud database offerings gained traction. The September 2025 euphoria after the OpenAI contract reporting and rapid RPO expansion pushed Oracle toward a narrative premium that looked more like AI infrastructure than conservative enterprise software. That was the period when the stock moved into valuation territory the previous report rightly flagged as stretched.

The latest phase began in June 2026, when the market stopped asking whether Oracle could win AI demand and started asking how it would pay for the required steel, power, leases, and chips. Reuters tied the post-Q4 selloff directly to debt-load and cash-burn concerns, and S&P’s July downgrade reinforced the message that this was not just equity volatility. The valuation center shifted because the business mix shifted. Oracle is now rewarded less for software-like resilience and more discounted for infrastructure-like capital appetite.

Business model and financial review

Revenue machine and where profits really come from

Oracle still makes most of its money from what can be called contractual enterprise software gravity. In fiscal 2026 the company reported total revenue of 67.4 billion USD, including 34.0 billion USD of cloud revenue, 24.5 billion USD of software revenue, 5.7 billion USD of services revenue, and 3.1 billion USD of hardware revenue. In fiscal 2025 total revenue was 57.4 billion USD and total cloud services and license support revenue was 44.0 billion USD. The old software-and-support layer remains the economic foundation underneath the new cloud expansion; Oracle has not abandoned legacy software.

The real profit source is still the cloud-and-license complex, especially database support, applications support, and higher-margin software layers sitting around the data estate. OCI is growing the fastest, but Oracle does not separately disclose OCI operating profit in a way that lets outside investors prove today’s unit economics. That gap matters: fiscal 2026 cloud-and-software operating expenses jumped to 17.6 billion USD from 11.6 billion USD in fiscal 2025, and Oracle’s Q4 call explicitly said cloud gross margins stepped down by five percentage points as the company scaled infrastructure. A business can survive that if volume outruns cost absorption. It cannot survive it forever if capital intensity stays enormous and pricing power turns out weaker than the market hoped.

Customer dependence is where the old Oracle story gets less comforting. The company disclosed on its Q4 FY26 call that it signed more than 8 billion USD of contracts with each of four customers during the quarter. Public reporting and credit commentary later pointed to OpenAI as a major concentration inside the AI backlog, though Oracle’s filings do not disclose enough customer-level detail to fully verify the exact proportion. So the honest posture is to treat RPO concentration as a real risk without pretending the company has fully disclosed the underlying customer map.

The conversion path from RPO to revenue to cash is much more nuanced than the headline number suggests. Oracle’s Q4 FY26 release said 76.56 billion USD of RPO is due within the next year, roughly 12% of the total. Oracle’s March 2026 10-Q said 12% was expected in the next twelve months and 31% over months 13–36. Most of the backlog is therefore long-dated and execution-heavy. Investors should also distinguish between a software backlog, where delivery tends to be lighter-capital and margin-richer, and an AI infrastructure backlog, where delivery demands power, land, servers, networking, and often long lease commitments before the revenue becomes cash.

Cost structure, moat, and governance

Oracle’s cost structure is now bifurcated. The old business has classic software economics: large fixed development and sales costs, then high incremental margin on renewals and support. The new AI infrastructure business has very different economics: lease and power costs, procurement obligations, rapid server depreciation, financing costs, and construction risk. Oracle’s own filings say its primary uses of operating cash include employee expenses, data-center leases and power, taxes, and interest. The moment OCI became the center of the growth story, Oracle’s cost base moved closer to a built-asset model.

The moat is still real, but it is not uniform across the whole company. Oracle’s strongest moats are customer stickiness, switching costs, and ecosystem depth around its database and mission-critical applications. These are not marketing moats. They have survived decades of competition. A second moat is architectural adjacency: Oracle can sell database, applications, support, and cloud around the same customer’s data estate. A third, newer moat is multicloud database placement. Oracle’s Q3 FY26 release said multicloud database revenue rose 531% year over year, which shows Oracle has found a way to profit from other clouds’ distribution rather than fight them on every front.

What Oracle does not yet have is an unambiguous moat in general-purpose infrastructure breadth. Azure, AWS, and Google Cloud still dominate developer mindshare, tooling ecosystems, and the broadest set of cloud-native workloads. Oracle wins where enterprise customers want Oracle database compatibility, lower-friction migration, strong isolation, and increasingly AI compute tied to a specific contract. That slice of the market is narrower but still meaningful, which is why Oracle can grow fast without needing to become the default cloud for everyone.

Governance cuts both ways here. The positive side is continuity. Larry Ellison still shapes product and strategy; Safra Catz’s era disciplined operations and capital markets; the 2025 co-CEO transition elevated the executives closest to OCI and industry applications. The negative side is concentration of influence. Oracle is still heavily shaped by insiders, and the current capital program requires investors to trust management’s ability to see several years ahead in power, customer demand, and financing markets. With a stable business that might be comforting. With 260 billion USD of off-balance-sheet lease commitments, it becomes a source of key-person and judgment risk.

Financial vertical review

The most useful way to see Oracle’s last three fiscal years is to look at what held firm and what broke shape.

Sources: Oracle FY2025 and FY2026 annual reports.

The business reason is straightforward. Revenue and earnings continued to improve because demand for OCI, cloud applications, and multicloud database services accelerated. But free cash flow stopped resembling software free cash flow because Oracle rebuilt itself into a capital spender. PP&E nearly doubled in fiscal 2025 and more than doubled again in fiscal 2026, while depreciation began to catch up. The market has accordingly stopped valuing Oracle on GAAP EPS alone; the balance-sheet expansion is too large to ignore.

The balance sheet is still liquid in the narrow sense, but it is no longer conservatively financed in the old Oracle sense. At May 31, 2026 Oracle had 31.3 billion USD of cash and equivalents, but also 130.1 billion USD of borrowings, 30.2 billion USD of operating lease liabilities, 7.7 billion USD of finance lease liabilities, 260 billion USD of additional lease commitments not yet on the balance sheet, and 13.3 billion USD of unconditional purchase and other obligations primarily tied to data-center power arrangements. Those non-balance-sheet commitments are not debt-equivalent on a one-for-one basis, but they do tell you that the company has locked itself into a much more fixed-cost operating model.

Interest burden has started to bite. Interest expense rose from 3.5 billion USD in fiscal 2024 to 3.6 billion USD in fiscal 2025 and 4.6 billion USD in fiscal 2026. Oracle remained compliant with debt covenants and still had no borrowings under its 10 billion USD revolving credit agreement at May 31, 2026, but the direction of travel is clear: more debt, more refinancing risk, and less room for disappointment if capex returns come in late.

A final point on earnings quality: Oracle’s operating-cash-flow-to-net-income ratio has been very high, at 178% in fiscal 2024, 167% in fiscal 2025, and 187% in fiscal 2026. That looks excellent until one opens the hood. In fiscal 2026 Oracle said the increase in operating cash flow included 4.6 billion USD of customer prepayments with a significant financing component. It also added back 4.8 billion USD of stock-based compensation. Cash conversion is real, but some of that “cash quality” is front-loaded by contract structure and financing, not purely by recurring software economics.

Industry and horizontal analysis

Industry structure and cycle

Oracle sits at the intersection of enterprise applications, databases, and cloud infrastructure. Those markets do not move in perfect sync. The applications side is relatively defensive and renewal-heavy, and the database side is sticky, deeply embedded in regulated and performance-sensitive workloads. Infrastructure is increasingly tied to a capex cycle driven by AI demand, power availability, financing conditions, and customer willingness to commit to long-duration compute contracts. The result is a hybrid cycle profile: part defensive software, part technology-iteration race, part capex cycle, and now part rate cycle because debt and leases matter much more than they used to.

That hybrid profile is why the market’s May–July 2026 selloff in AI infrastructure names hurt Oracle, but did not explain Oracle by itself. Reuters reported a broad AI-equity wobble in early June as growth stocks sold off on higher-rate fears and valuation fatigue. Hotter inflation concerns and rate anxiety matter because higher discount rates punish long-duration growth and because expensive infrastructure is harder to justify when funding costs rise. But Oracle’s move was sharper because it had its own financing story layered on top. Sector beta was the weather. Oracle’s capital structure was the lightning strike.

Policy and geopolitics matter mostly through power, data, and financing. Oracle’s AI build-out depends on data-center capacity, local power availability, and customer demand in an environment where governments increasingly care about national AI infrastructure and data sovereignty. The Stargate-related expansion with OpenAI and others fits that backdrop. This is a tailwind for demand, but it also pulls Oracle toward larger fixed commitments and more government-adjacent execution risk than the legacy software story ever carried.

Peer portrait

Oracle has enough direct and indirect peers that this should be treated as a scenario-C landscape: several meaningful rivals, but each competing on a slightly different axis. Microsoft and Amazon dominate broad public cloud and enterprise AI infrastructure. Alphabet is now the fastest-improving large cloud competitor in AI-heavy workloads. SAP is the clearest application-layer comparator in ERP and process software. Salesforce remains relevant as a cloud-software and AI-software benchmark, but it is less directly exposed to infrastructure capital intensity than Oracle is.

Sources: market caps and P/E from live market feeds; company growth and cloud metrics from latest official company disclosures.

The business reasons behind the differences are more important than the table. Microsoft became the default enterprise AI platform because it married Azure with productivity software, developer tools, and OpenAI distribution, while still funding capex from a vastly larger profit base. Amazon remains the broadest infrastructure utility, with AWS still generating segment-level profits that can be measured directly. Alphabet has become the most impressive late improver in AI cloud momentum, with far quicker cloud growth and materially higher disclosed cloud margins than Oracle. SAP became the cleanest process-software counterpoint: slower than the hyperscalers, but with a backlog and ERP story the market reads as more software-like and less balance-sheet-heavy.

Oracle’s niche is narrower but real. Customers choose Oracle when database compatibility, performance for Oracle-centric workloads, support of regulated or legacy enterprise environments, and increasingly multicloud database deployment matter more than having the broadest cloud catalog. Oracle’s Q3 multicloud database revenue growth of 531% shows that it has found a lucrative wedge: sell Oracle technology on everyone else’s real estate if customers prefer that architecture. That is clever positioning, and not the same thing as winning the whole cloud market.

The horizontal judgment comes down to this: Oracle’s competitive edge is strongest where data gravity and application switching costs matter, and weakest where general developer ecosystem breadth matters. Enough to sustain real growth, but not by itself enough to justify hyperscaler-like capital intensity unless management can prove the new assets earn durable returns. That proof is still incomplete.

Current fundamentals, valuation and tracking

Last four quarters and what the market is trading

Oracle’s last four reported quarters show a business whose demand line kept getting steeper even as its financial quality became more contested.

Sources: Oracle official quarterly releases and call summaries.

Read sequentially, the pattern is obvious: revenue, cloud, OCI, and RPO all accelerated. What did not accelerate in a comfortable way was free cash flow. Oracle’s own March release still tried to calm the market by saying many large AI contracts were being supported either by customer prepayments or customer-supplied GPUs, and that Oracle did not expect to need incremental funds for those specific deals. By June, the message had changed in tone: capex could reach 90–95 billion USD in fiscal 2027, around 70 billion USD would be Oracle’s own net cash outlay, and the company still planned major debt and equity financing. The stock reaction told you which message investors took more seriously.

What the market is trading right now is the gap between AI demand and AI funding, not simple “AI enthusiasm.” Oracle’s shares fell because investors suddenly realized Oracle was trying to finance infrastructure at a pace that moved it out of the comfort zone for a software issuer, not because backlog disappeared. Reuters reported that the June selloff was driven by higher-than-expected capital spending, a deepening free-cash-flow deficit, and fears about rising debt. S&P’s downgrade in July made that narrative official enough for credit investors to care.

Price-move attribution and RPO quality

The period since May 20 can be decomposed into four forces.

First, company-specific financing concern did most of the work. Oracle’s fiscal 2026 capex was 55.7 billion USD; free cash flow was negative 23.7 billion USD. In fiscal 2026 the company issued 43.0 billion USD of senior notes, 5.0 billion USD of mandatory convertible preferred stock, enlarged its revolving credit facility to 10.0 billion USD, and set up a 20 billion USD at-the-market equity program. That is the funding architecture of the new Oracle.

Second, capital intensity changed the equity story. Oracle’s own June commentary framed fiscal 2027 capex at up to 95 billion USD, but also said about 20–25 billion USD should be repaid by customers, leaving about 70 billion USD of Oracle net cash outlay. Even on that friendlier definition the gap against fiscal 2026 operating cash flow is enormous. Using reported fiscal 2026 capex, the cash-flow gap was 23.7 billion USD. Using Oracle’s net-cash framing for fiscal 2027, the gap still implies another year in which internal cash generation does not remotely cover the build. This does not prove value destruction, but it does prove that investors are now underwriting execution and financing, not just demand.

Third, RPO quality is good enough to matter, but not clean enough to deserve a software multiple without discount. The positive case is that Oracle is signing real contracts, receiving some customer prepayments, and in some cases having customers fund their own GPUs. The negative case is that only about 12% of RPO converts in the next year, a large portion sits beyond three years, and public reporting now points to meaningful concentration in a few AI customers. The company disclosed four customers signed for more than 8 billion USD each in Q4, but has not disclosed enough customer detail for investors to evaluate exact counterparty dispersion, termination rights, and credit quality. That gap alone lowers backlog quality relative to SAP-like subscription backlog or traditional software support renewals.

Fourth, macro and sector beta amplified but did not originate the move. Early June saw a wider AI-stock wobble as rate fears and sky-high expectations collided. The 10-year U.S. Treasury yield was around 4.64% on July 27, 2026, high enough to keep pressure on long-duration growth valuations. But Oracle’s underperformance versus many software peers reflects company-specific cash-flow and balance-sheet risk, not just higher discount rates.

The mandatory classification follows from that. Oracle today is closer to deteriorating financial fundamentals than to a pure market mispricing, but it is not a classic value trap in the sense of a collapsing franchise. Demand has improved, the installed-base moat holds, and OCI has genuine product-market fit in a specific slice of AI and database infrastructure. What has deteriorated is financial quality: capex intensity, fixed commitments, and concentration risk. So the right reading of the 34% share-price drop is “good business, riskier balance sheet, lower valuation bands,” rather than “buy the dip” or “business broken.”

Valuation analysis

Start by refusing to read headline earnings in isolation. Oracle’s reported P/E on the live market feed is about 21.5x, and that multiple flatters the equity if one ignores the capital the company now needs to sustain and grow OCI. A lower-bound maintenance-capex proxy is depreciation, which rose to 7.6 billion USD in fiscal 2026 from 3.9 billion USD in fiscal 2025. Because Oracle’s server estate mostly carries a six-year useful life and the company is also building long-lived data-center assets, a prudent maintenance-capex estimate is higher than depreciation alone. I use a normalized maintenance-capex range of 10–12 billion USD, which is still conservative versus the scale of the current asset base.

Under that approach, owner earnings are materially below the figure implied by operating cash flow. Fiscal 2026 operating cash flow of 32.0 billion USD included 4.6 billion USD of customer prepayments with a significant financing component and 4.8 billion USD of stock-based compensation addbacks. After adjusting for those items and normalized maintenance capex, Oracle’s stricter equity-owner earnings are much closer to 12–14 billion USD than to 17–18 billion USD of GAAP net income or 32 billion USD of reported operating cash flow. On that basis, the true owner-earnings multiple is meaningfully higher than the headline P/E, which is why I default to owner earnings rather than accounting earnings in the scenario work below.

Historically, Oracle’s current valuation is lower than the AI-euphoria phase that carried the stock near its prior highs, but it is not washed-out in the way distressed capital consumers become washed-out. The stock is roughly 65% below its 52-week high of 345.72 USD, according to market data around May 2026, yet still trades on more than 21x trailing earnings and has not reached a level where the financing and execution risks are obviously over-discounted. This is cheaper than it was. It is not comfortably cheap.

Absolute valuation is more useful than peer spread here because peers are contaminated in different ways. Microsoft and Alphabet have much larger cash engines; Amazon’s AWS economics are much better established; SAP’s backlog has cleaner software characteristics. Oracle should trade at a discount to cleaner software names until management proves that OCI’s incremental return on capital is durably adequate, earning neither the full software premium of SAP nor the infrastructure premium once granted to AI darlings.

This is scenario analysis within a research framework, not investment advice. The assumptions reflect my judgment built on Oracle’s disclosed capex, RPO timing, debt load, and peer context.

The expectation gap is now clear. The market no longer doubts that Oracle can sign big AI contracts. It doubts the quality and economics of fulfilling them. The next event does not need another huge RPO number to move the stock. What investors care about most is narrower: how much fiscal 2027 capex is truly net of customer reimbursement, whether lease commitments are still rising at the same pace, whether OCI margins stabilize, and whether any customer concentration becomes clearer. Those are the variables that can still change both bull and bear judgments.

Margin of safety needs unpacking. The current price is above my ideal-buy range and only modestly below the center of my base-case fair value. The most fragile base-case assumption is that the very large AI backlog converts with acceptable utilization and pricing. If I haircut that conversion assumption to 70% of the base case, my fair value drops from roughly 122 USD to roughly 108 USD. If Oracle’s earnings simply flatten for three years and the market continues to award a mid-teens multiple, the expected annualized return is little better than the dividend yield and sits below the 4.64% area of the 10-year Treasury. At this buy price, there is no meaningful margin of safety for new money.

Risk, catalysts, and tracking dashboard

The biggest permanent-loss risks are specific. Counterparty and utilization risk comes first: if a small number of AI customers drive a large share of long-duration obligations, Oracle could be left with underused capacity and fixed lease expenses. Financing risk follows, since further debt or equity issuance at weak prices would both raise the hurdle rate and dilute the equity case. Third is return-on-capital risk, because OCI can keep growing while still earning mediocre returns if compute pricing falls or utilization disappoints. Fourth is valuation risk from higher rates and sector-style rotation: Oracle’s new story is too long-duration and too capital-hungry to be rate-insensitive. Last comes governance and execution risk, with the company making unusually large judgment calls on power, site selection, customer quality, and contract structure under a leadership team that is new in its current form.

The positive catalysts are the mirror image. The best one would be hard evidence that fiscal 2027 net capex is running below the feared path because customer reimbursements and financing structures work better than the market expects. Clearer disclosure on backlog timing and customer diversity ranks next. Others would be OCI margin stabilization while growth stays above peer expectations, a pause in credit deterioration, and any sign that Oracle can taper lease growth without losing the growth narrative.

Sources for observable variables: Oracle filings, quarterly releases, live market data, S&P action, and Treasury data. Oracle’s FAQ shows its fiscal pattern and typical release cadence, but as of this report date I did not find a formal company posting for the next earnings date.

These indicators are worth watching because they map directly to the current disagreement. OCI growth tells you whether demand is real. The RPO timing mix separates improving backlog quality from backlog that is merely getting longer-dated. Net capex burden and lease commitments show whether Oracle is becoming a more self-funding business or a more leveraged one. Credit posture reveals whether bond investors are calming down. And the 10-year Treasury matters because a company trying to turn long-duration growth into present value cannot ignore the discount rate.

Cross-synthesis summary

Bull and bear reasons

The strongest bull reason is that Oracle’s installed base is still exceptionally valuable. The company is not trying to invent relevance from scratch; it is monetizing decades of database and applications entrenchment into cloud, multicloud, and AI infrastructure demand. Hence cloud revenue of 34.0 billion USD in fiscal 2026 and multicloud database revenue growth of 531% in Q3 FY26.

Second, demand is not theoretical. OCI growth accelerated from 68% in Q2 to 84% in Q3 to 93% in Q4, while RPO rose from 455 billion USD in Q1 to 638 billion USD in Q4. The debate is not whether Oracle found AI demand. It did.

Third, Oracle may be exploiting a differentiated position rather than trying to become a me-too hyperscaler. Multicloud database offerings let it monetize AWS, Azure, and Google Cloud footprints rather than fight for every workload on raw ecosystem breadth. The market gives that strategy less credit than it deserves.

Fourth, valuation is far less demanding than it was in late May. The stock at 119.90 USD is roughly one-third below the prior report’s reference price and far below the 2025 AI-euphoria highs. Some of the prior excess has already burned off.

The core bear case begins with capital intensity. Oracle spent 55.7 billion USD on capex in fiscal 2026 after 21.2 billion USD in fiscal 2025, yet still guided to up to 95 billion USD in fiscal 2027. That is industrial-scale buildout, not software capex.

Second, fixed obligations are now enormous. Oracle had 130.1 billion USD of borrowings, 37.9 billion USD of on-balance-sheet lease liabilities, 260 billion USD of additional lease commitments off balance sheet, and 13.3 billion USD of unconditional purchase obligations at May 31, 2026. Those figures do not guarantee trouble, but they eliminate the old balance-sheet serenity that used to justify a higher-quality multiple.

Third is backlog quality. Only about 12% of RPO is due within the next year, much of it sits far beyond three years, and public reporting now points to meaningful concentration in a handful of AI customers, with OpenAI drawing particular scrutiny from credit investors. Oracle has not disclosed enough detail to let the market price that risk cleanly.

Fourth, returns on incremental capital remain unproven. Oracle’s blended operating-income improvement over the last two years is real, but set against the huge step-up in capex it still suggests only high-single-digit to low-double-digit current returns, and that likely overstates OCI-specific economics because the legacy software core still carries most of the profitability. The market is right to demand proof before rerating the shares.

Pre-mortem

Three years from now, a 50% drawdown from today would most likely come through one of two scripts.

The first script is the concentrated-contract failure. In 2027 or 2028, one or two major AI customers delay or resize commitments, Oracle still takes delivery of data-center capacity and power obligations, utilization falls far below expectations, OCI gross margins remain compressed, interest expense rises above 6 billion USD, and the market re-rates Oracle from a premium software multiple toward a low-teens earnings multiple. In that script the stock could move toward 60–70 USD even if the legacy database franchise remains intact. The causal chain would be fixed commitments hitting before the revenue does.

The second script is the slower, quieter version. Demand stays real, but returns stay mediocre. Oracle keeps posting strong OCI growth, yet capex never normalizes, free cash flow remains negative or barely positive, the market decides that Oracle is really an asset-heavy compute lessor wearing a software badge, and the valuation settles around 14x–15x stricter owner earnings rather than 20x. That outcome does not require a collapse. It only requires the market to stop believing that every dollar of AI backlog will eventually earn software-like returns.

Final research conclusion

Vertically, Oracle has proved one thing beyond argument: it knows how to turn deeply embedded enterprise technology into long-lived contractual economics. That held in databases, in maintenance and support, in ERP and adjacent applications, and it is at least directionally true in the first phase of OCI and multicloud database services. Oracle’s historic success came from embedding itself where data, compliance, performance, and operational risk made switching painful, not from fashion. That capability is still present.

Horizontally, AWS, Azure, and Google Cloud still own more breadth. Oracle’s edge is the ability to extract high-value workloads from an installed customer base that already trusts Oracle with mission-critical data and applications, then place those workloads into Oracle-controlled or Oracle-enabled cloud environments. The weakness is not temporary in the sense that breadth will suddenly equalize; Oracle will likely remain a more focused infrastructure player. What is temporary, if management is right, is the worst of the cash-flow pain. That still has to be proved.

The market’s main misjudgment in 2025 was to pre-spend future success. The market’s current risk is the opposite error: to infer from a necessary derating that the operating franchise is broken. I do not think it is broken. I do think financial quality has deteriorated enough that the stock deserves lower valuation bands than the previous report used. So the right call sits between enthusiasm and dismissal. Oracle at 119.90 USD is no longer an obvious valuation bubble. It is also not the kind of bargain that lets an investor ignore the possibility of further dilution, more leverage, or disappointing returns on a gigantic build program.

【Company-profile scores】

• Fundamental quality: high
• Growth: medium
• Moat: strong
• Financial soundness: medium
• Management credibility: medium
• Valuation attractiveness: medium
• Risk level: medium
• Suitable investor type: long-term growth

【Investment rating】

• Rating: Hold
• One-line thesis: Oracle’s software moat is intact, but OCI growth now comes with capital intensity and fixed commitments large enough to cap the rerating.
• Three price signals:
  • Ideal buy price: 【Ideal Buy Price】80–90 USD Basis: at least a 20% margin of safety below my conservative valuation range of roughly 101–112 USD per share.
  • Acceptable hold price: 104–140 USD
  • Clearly overvalued price: 155 USD and above
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. I would prefer a new-money entry below 90 USD, or above that only if Oracle shows two things at once: lease growth moderating and OCI margins stabilizing. The opportunity cost of waiting is missing a rerating if fiscal 2027 capex proves more customer-funded than feared.
• Target holding horizon: 3–5 years
• Expected annualized return: conservative about -3% to 0%; base about 2% to 4%; optimistic about 7% to 9%
• Max-loss risk: about 40%–50% in a stressed case, with the trigger being a large-customer slowdown or underutilized AI capacity arriving into a still-rising lease and interest bill
• Reassessment-trigger signals: if OCI revenue growth falls below 35% for two consecutive quarters; if additional lease commitments continue to rise materially above the already-disclosed 260 billion USD level; if annualized interest expense moves above 6 billion USD without similar OCF growth; if RPO due within twelve months falls further without better customer disclosure; if Oracle needs materially more equity financing than the market currently expects.

The rating moves up relative to the previous report, from Watch to Hold. The reason is valuation, not improved safety. The old report’s caution at around 181 USD was validated by events. Since then, the core quality of the business has been confirmed, but the financial quality of the AI build-out has deteriorated. I therefore disagree with any simplistic reading that the stock is merely “cheap versus the old bands.” It is cheaper, yes. It is also a different risk object than it was in May. My bands move down because the fixed-obligation burden and backlog-quality uncertainty are now more visible than they were 69 days ago.

【Valuation Range】

• current: 119.90 (close as of 2026-07-27)
• bear (conservative · ideal buy zone): [80, 90]
• base (fair · acceptable hold zone): [104, 140]
• bull (optimistic · above the clearly-overvalued line): [155, 175]

Key data tables

The direction of change matters more than the simple labels. The old report’s business-quality judgment was confirmed. The old bull-band ceiling was too high for today’s Oracle because the company now carries more funding and commitment risk than was visible in May. The rating rises even though the bands come down because price corrected faster than the full underlying franchise deteriorated.

Research uncertainties

The biggest blind spot is customer concentration. Oracle’s disclosures do not let outside investors fully map the RPO across named AI counterparties, termination rights, or effective credit support, so any judgment on backlog quality is necessarily probabilistic.

OCI profitability is the second. Oracle discloses enough to show cloud acceleration and margin pressure, but not enough to isolate OCI return on capital with precision, which forces investors to infer from group-level economics.

Third comes the exact timing of lease commencement and customer reimbursement inside the 260 billion USD additional lease commitments. The gross number is disclosed; the project-by-project timing and utilization assumptions are not.

Last is future financing mix. Oracle has disclosed frameworks for debt, preferred, and potential common-equity issuance, but the exact cadence and market terms remain unknown. That matters for per-share value.

Sources

This report is grounded primarily in Oracle’s FY2026 and FY2025 annual reports, Oracle’s FY2026 quarterly earnings releases and Q4 call transcript, Oracle investor-relations materials on financing, live market data for current prices and market caps, peer companies’ latest official earnings disclosures, S&P’s July 2026 downgrade headline, and U.S. Treasury market data for the 10-year yield. The most important non-company media sources used here are Reuters and the Financial Times, chiefly for market reaction, financing concerns, and sector context.

Other tickers mentioned

• US MSFT.US: principal cloud and enterprise-software benchmark, especially Azure and AI monetization
• US AMZN.US: AWS is the broadest public-cloud infrastructure comparator and the clearest infrastructure-margin benchmark
• US GOOGL.US: Google Cloud is the fastest-improving AI-cloud peer and a key valuation and execution comparison
• US SAP.US: closest large-scale ERP and enterprise-process software comparison for backlog quality and software economics
• US CRM.US: cloud-software and enterprise-AI applications peer with cleaner cash-flow characteristics than Oracle’s current infrastructure push

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.