ORCL — Oracle Corporation

Oracle sits on top of one of the most durable software annuities in enterprise technology — a $44 billion recurring support business anchored by database and ERP systems that Fortune 500 companies cannot remove without years of risk and cost — while simultaneously executing the most capital-intensive infrastructure bet in its history, taking on $162 billion in debt to build the data center capacity that AI hyperscalers are contractually committed to fill. The RPO backlog has exploded from $138 billion to $553 billion in less than a year, driven by AI infrastructure contracts with the largest names in the industry, and OCI revenue is growing at 84% year over year. The business is interesting but not yet actionable: the annuity is exceptional and the cloud trajectory is legitimately compelling, but the stock requires OCI revenue to ramp materially through FY2027 and contracted obligations to convert to cash before the math clears a reasonable entry threshold.

---

Enterprise technology spending has entered one of its most dislocated periods in twenty years. The AI infrastructure buildout has created a capital expenditure arms race among cloud providers, with Microsoft, Amazon, Google, and Oracle collectively committing hundreds of billions of dollars to data center expansion in a compressed timeframe. The conventional narrative holds that this benefits the hyperscalers and disadvantages the legacy vendors — that Oracle is a company of the prior era, fighting a rearguard action as workloads migrate to AWS and Azure. That narrative was plausible in 2020. By early 2026, the evidence has substantially inverted it. Oracle's remaining performance obligations — the contracted future revenue that has not yet been recognized — stood at $553 billion as of February 2026, up from $98 billion two years earlier. That figure is not a projection or a sales pipeline estimate. It is signed contract value. Something has changed at Oracle that the conventional narrative did not anticipate.

The change is structural. AI model training and inference at scale requires not general-purpose cloud compute but specifically configured infrastructure: dense GPU clusters, high-bandwidth networking, low-latency storage, and the operational expertise to run them at near-perfect uptime. The three dominant hyperscalers — AWS, Azure, and Google Cloud — are capacity-constrained. They are simultaneously building for their own AI workloads and serving external customers, creating a genuine shortage of available GPU capacity for frontier AI labs. Oracle Cloud Infrastructure entered this environment with a specific technical advantage: its network fabric, built from the ground up with RDMA over converged Ethernet, delivers lower latency between GPU nodes than the incumbents' general-purpose networks. For distributed AI training jobs where inter-GPU communication is the bottleneck, this matters enough to drive contract decisions at companies whose infrastructure costs run into the billions of dollars annually.

The enterprise software industry that Oracle built its original business on operates by different rules than consumer technology. Enterprise software is purchased to run critical processes — financial systems, supply chains, human resources, manufacturing operations — and its replacement requires not just technical migration but organizational change management, re-training, and the assumption of meaningful operational risk during transition. Oracle's database has been the foundation of enterprise transaction processing since the 1980s. Its Fusion ERP applications, the on-premise E-Business Suite, and the PeopleSoft and JD Edwards systems it acquired in the early 2000s collectively process the financial records of the majority of the world's largest corporations. These systems do not get replaced because a competitor builds a slightly better product. They get replaced, if they get replaced at all, as part of multi-year transformation programs that typically cost $50 million to $150 million for a Fortune 500 company and carry substantial failure risk. Oracle's customer retention is not a measure of customer satisfaction — it is a measure of the cost of defection. The two things are often confused, but only one of them is an economic moat.

Oracle generates revenue from three primary sources: cloud services and license support (the annuity), cloud license and on-premise licenses (new license sales, declining), and hardware (legacy, immaterial). The support revenue — maintenance contracts on existing software installations that renew at approximately 22% of original license cost per year — is the economic core of the business. In fiscal year 2025, cloud and license support generated $44.0 billion in revenue, growing 12% year over year. This segment runs at operating margins that are rarely disclosed separately but that analysts and the company's own disclosures imply are north of 80% — the customer has already paid for the software, the incremental cost of supporting it is predominantly labor, and Oracle's pricing power on support renewals has historically been approximately 8% annual increases accepted without meaningful churn. Oracle has raised support prices every year for fifteen years, and the renewal rate on its installed base is above 90%. This is not a business that competes on price. It is a business that sets its price and collects.

The moat in the database segment is specific and demonstrable. Oracle Database runs inside the transaction processing systems of roughly 95% of the Fortune 500. The specific mechanism of lock-in is not brand loyalty but technical dependency: Oracle SQL extensions, Oracle-specific stored procedures, Oracle-specific performance tuning, and Oracle-specific backup and recovery procedures accumulate over decades of system development into a technical profile that cannot be replicated by open-source alternatives without a clean-sheet rewrite. Migrations from Oracle to PostgreSQL or MySQL — which are technically possible and have been completed at smaller companies — at enterprise scale require re-certification of every application that touches the database, re-testing of every business process, and acceptance of a multi-year window during which both systems must run in parallel. The cloud providers' managed database services (Amazon RDS, Azure SQL) are direct competitors for new workloads, but they have not meaningfully displaced Oracle in the mission-critical core where the lock-in is deepest. Oracle's autonomous database, which automates patching, tuning, and backup without human intervention, extends the moat by reducing the administrative overhead that made on-premise Oracle expensive — while keeping revenue in Oracle's ecosystem rather than migrating it to AWS.

The financial profile of Oracle looks very different depending on where you draw the lens. The underlying annuity business — support revenue plus the cloud applications business that is cannibalizing legacy on-premise installations at higher margins — is exceptional. Total revenue for fiscal year 2025 was $57.4 billion, up 8%, with cloud services growing 24%. GAAP operating income was $17.7 billion, representing a 30.8% operating margin. Non-GAAP operating income, which adds back $4.7 billion in stock-based compensation and acquisition-related amortization, was $25.0 billion, or 43.6% of revenue. GAAP net income was $12.4 billion, or $4.34 per share. These are the numbers that describe the business as it has operated historically.

The numbers that describe the business as it is operating now tell a different story. Operating cash flow for fiscal year 2025 was $20.8 billion — strong. Capital expenditures were $21.2 billion — up from $6.9 billion in fiscal year 2024 — producing free cash flow of negative $394 million for the full year. In the trailing twelve months through February 2026, capital expenditures have accelerated further as Oracle builds data center capacity against its contracted RPO commitments. Capex guidance for fiscal year 2026 is approximately $50 billion. Against operating cash flow that has run at $18 to $21 billion annually, this implies free cash flow of negative $25 to $30 billion for fiscal year 2026 before any improvement from the higher-revenue contracts coming online. Oracle has funded this capital deficit with debt: long-term debt stood at approximately $162 billion as of February 2026, up from $88 billion at the end of fiscal year 2024. After a capital raise in early 2026, cash on the balance sheet stood at approximately $39 billion, producing net debt of approximately $123 billion. A company that generated $11.8 billion in free cash flow in fiscal year 2024 now carries net debt equal to roughly ten times that figure.

The debt load is the central risk in the Oracle thesis, and it deserves examination without euphemism. Oracle's debt structure has been used aggressively for decades — the company funded the PeopleSoft acquisition, the Cerner acquisition, and successive share buyback programs with borrowed capital, and it has retired that debt reliably as cash generation allowed. The current leverage is different in scale. At $123 billion in net debt against a business generating $20 billion in operating cash flow before the capex surge, Oracle's leverage ratio is roughly six times operating cash flow — a level that would typically indicate distress in an industrial company. Oracle is not an industrial company. Its annuity revenue is contractual and does not decline in recessions. The interest coverage on the existing business, exclusive of the OCI buildout, is adequate. But the bet is explicit: Oracle must generate sufficient revenue from the $553 billion in contracted obligations to service this debt at reasonable cost. If OCI revenue ramps as the RPO backlog implies, the debt is manageable and the equity value is substantial. If the AI infrastructure spending cycle stalls, contracts get renegotiated, or execution disappoints, Oracle's balance sheet will require addressing at a time when the annuity may be funding both debt service and continued capex obligations.

Larry Ellison founded Oracle in 1977 and owns approximately 41% of the outstanding shares, a position worth north of $175 billion at current prices. This is the dominant fact in the management alignment analysis. Ellison's incentives are aligned with Oracle shareholders in the most direct way possible: he is the largest single shareholder by an enormous margin. In July 2025, he sold approximately 6.5 million shares at prices near the all-time high of $345 — a transaction that represented less than 0.6% of his total position. The sale is worth noting but not alarming: a founder liquidating a fraction of a percent of an enormous position near all-time highs is a normal liquidity event, not a signal of reduced conviction. Ellison, now in his eighties, has driven the OCI buildout personally — he relocated to Hawaii partly to oversee the construction of a major data center facility there, and his public statements on AI infrastructure have been specific and consistent with the capital allocation decisions the company is making.

CEO Safra Catz has run Oracle's financial operations since 2004 and has been co-CEO since 2014. Her track record on capital allocation is mixed. The share buyback program has been large — Oracle has reduced its share count from approximately 5.6 billion shares in 2010 to approximately 2.8 billion shares today — but much of this buyback occurred at prices that in retrospect were not cheap relative to intrinsic value. The Cerner acquisition in 2022 for $28.3 billion remains unproven: Cerner is the largest electronic health records system in the United States, and Oracle has invested substantially in integrating it with cloud infrastructure, but the financial contribution to operating margins has not yet been demonstrable. The OCI buildout itself is the largest capital allocation bet in Oracle's history, and its outcome will define Catz's legacy as a capital allocator. The company has consistently met or beaten its own guidance — non-GAAP EPS growth has been within or above guidance range in each of the last ten quarters — which speaks to execution discipline even as the strategic bets are large.

The growth runway for Oracle has two distinct components that need to be understood separately. The annuity business — database support, ERP support, Fusion Applications — grows at mid-single to low-double digits annually as Oracle increases prices on existing customers and converts on-premise installations to cloud subscriptions at higher annual contract values. This growth is predictable and not dependent on new customer acquisition: Oracle has essentially 100% penetration of the Fortune 500 in its core database category. The incremental revenue comes from price increases, from selling additional cloud modules within the installed base, and from the Autonomous Database and Oracle Cloud Applications that upsell existing customers into higher-value, higher-margin subscription arrangements.

The OCI growth trajectory is the thesis-altering variable. It is quantified in the table below:

Quarter	OCI Revenue	Total Cloud Rev	RPO	Capex (Ann.)	Net Debt
Q4 FY2024	$2.0B	$5.6B	$98B	$6.9B	~$58B
Q4 FY2025	$3.0B	$7.7B	$138B	$21.2B	~$74B
Q1 FY2026	$3.3B	$7.2B	$455B	~$50B	~$108B
Q2 FY2026	$4.1B	$8.0B	$523B	~$50B	~$123B
Q3 FY2026	$4.9B	$8.9B	$553B	~$50B	~$123B

OCI revenue grew from $2.0 billion quarterly to $4.9 billion quarterly in five quarters — a 145% increase in annualized run rate, from approximately $8 billion to approximately $20 billion. The RPO jump from $138 billion to $455 billion between Q4 FY2025 and Q1 FY2026 reflects the signing of the Stargate partnership with OpenAI and related AI infrastructure contracts with Meta and others. These are not speculative numbers. RPO is audited contracted revenue. The question is not whether Oracle has secured large AI infrastructure contracts — the $553 billion backlog makes that undeniable. The question is the pace at which that backlog converts to recognized revenue and then to cash, and whether the debt accumulated to build the capacity can be serviced in the interim.

At $4.9 billion quarterly, OCI revenue annualizes to approximately $20 billion. If the growth trajectory sustains at even half its current pace — say 40% year over year from here — OCI would reach approximately $28 billion in annual revenue by fiscal year 2027. Adding this to the $44 billion annuity base and the $9 billion in cloud applications would put total revenue approaching $85 billion. At the current operating leverage of the annuity business, a revenue base of that size would generate operating cash flow in the range of $30 to $35 billion. Against a capex cycle that management has indicated will peak around $50 billion in FY2026 before stepping down as capacity comes online, the free cash flow trajectory would turn sharply positive in FY2027 and FY2028 — if OCI executes. The penetration math supports the runway: Oracle has captured a fraction of the enterprise cloud infrastructure market, which is currently dominated by AWS (~32% share), Azure (~23%), and Google (~12%), with OCI at approximately 2 to 3% of global cloud revenue today. The $553 billion in contracted obligations is roughly six times Oracle's current annual cloud revenue, implying that the company has sold capacity it has not yet built against demand it has not yet recognized.

The management narrative and the numbers are broadly consistent at this moment in Oracle's history, which is unusual for a company undertaking a transformation at this scale. Ellison has described the AI infrastructure opportunity in public forums as the largest capital spending cycle in the history of the technology industry, and Oracle's financial commitments are consistent with that characterization — not aspirational statements but actual capex. The divergence worth noting is on free cash flow timing: Oracle's guidance and investor communications emphasize the RPO backlog and the revenue trajectory without equivalent emphasis on the cash deficit being funded by debt during the buildout phase. The negative $24.7 billion in LTM free cash flow is a fact that appears in the financials but not in the investor narrative. For a company that spent the prior decade generating $10 to $12 billion in annual free cash flow and returning most of it to shareholders through buybacks, the current state of the balance sheet represents a genuine discontinuity — not necessarily a permanent impairment, but a material change that the investment thesis must account for.

The intelligent bear on Oracle makes the following argument: $553 billion in RPO represents contracted obligations from counterparties — primarily AI companies and hyperscalers — whose own capital spending plans are subject to revision as AI model training economics, regulatory constraints, and competitive dynamics evolve. OpenAI is not a Fortune 500 company with forty years of Oracle database dependency; it is a startup whose strategic priorities and financial capacity can change. If AI infrastructure spending decelerates — whether from model efficiency improvements that reduce training compute requirements, from regulatory intervention, or from investor pressure on AI companies to achieve profitability — Oracle's contracted backlog could be renegotiated or cancelled. At that point, Oracle would hold $162 billion in debt, data centers that are difficult to redeploy, and an annuity business that is healthy but insufficient to service the leverage it has assumed. The bear is not arguing that Oracle is a bad business. The bear is arguing that the AI infrastructure bet is a binary outcome at the current balance sheet configuration, and binary outcomes at $162 billion in net debt are dangerous.

The answer to the bear is specific and conditional rather than dismissive. The RPO backlog is not a single large bet but a portfolio of contracts across multiple AI companies, hyperscalers, and enterprise customers. The OpenAI Stargate commitment is the most prominent, but the $553 billion represents diversified demand from multiple sources with varying risk profiles. Oracle's technical differentiation in GPU cluster networking — the low-latency RDMA fabric that AI training workloads specifically value — is validated by the contracts themselves: hyperscalers do not sign multi-year commitments totaling hundreds of billions of dollars without evaluating the technical alternatives. And the annuity business, generating $20 billion in operating cash flow before capex, is not going to zero in a scenario where AI infrastructure spending slows. The bear's scenario requires simultaneous contract cancellations across Oracle's customer base, not a slowdown in AI spending generally. That scenario exists, but it is not the base case, and it does not describe what is currently observable in Oracle's order book. The weaker version of the bear — that growth slows, OCI ramps later than expected, and the stock performs poorly for a few years while debt gets serviced — is entirely plausible and is adequately captured in the watchlist designation.

At approximately $153.75 per share, Oracle trades at roughly 35 times trailing GAAP earnings per share of $4.34, and approximately 25 times trailing non-GAAP earnings of $6.03. The enterprise value is approximately $563 billion — market cap of $440 billion plus $162 billion in debt minus $39 billion in cash. Against trailing revenue of approximately $59 billion, the EV/Revenue multiple is approximately 9.5 times. Against Oracle's fiscal year 2026 revenue guidance of approximately $67 billion, the forward EV/Revenue is approximately 8.4 times. These multiples are not cheap by any traditional measure for a business generating negative free cash flow. They are substantially below the stock's peak valuation in September 2025, when the stock traded at $345 — a level at which the market was pricing in the full OCI scenario without adequate risk adjustment for the balance sheet commitments required to achieve it. The 55% decline from that peak has restored some rationality to the valuation without yet bringing it to the level where the annuity business alone justifies the price.

The normalized pre-tax earnings of the business — taking the annuity as the durable core and excluding the distortions of the OCI buildout — are approximately $14.9 billion on a GAAP basis, or $5.32 per share at 2.8 billion diluted shares outstanding. At 15 times that figure, the annuity business alone would be worth approximately $80 per share. The current price of $153.75 implies that the market is attributing roughly $74 per share — approximately $207 billion of enterprise value — to the OCI growth trajectory above and beyond the annuity. Whether that attribution is reasonable depends on whether OCI revenue ramps as the RPO backlog implies. If it does, $207 billion for a cloud infrastructure business growing at 80% year over year with $553 billion in contracted obligations is arguably conservative. If it does not — if the AI spending cycle stalls and contracts get renegotiated — $207 billion for a partially-built data center empire backed by $162 billion in debt is not conservative at all.

The stock is not compelling at $153 because it requires the OCI scenario to work. It is not avoidable at $153 because the underlying annuity is exceptional and the OCI evidence base is genuinely stronger than the market appears to credit, given the 55% decline from peak. The right characterization is conditional: if OCI revenue reaches $30 billion annually by FY2028, the current price will prove to have been an attractive entry point in retrospect — the contracted backlog makes a path to that number visible, not just plausible. If OCI execution disappoints materially or AI infrastructure spending stalls, the stock has meaningful downside from here as the market re-prices the annuity alone and discounts the leveraged balance sheet. The investor who buys Oracle today is making a specific bet: that the $553 billion in contracted obligations is real, that Oracle can build the capacity to satisfy them on schedule, and that the AI infrastructure demand driving those contracts does not abruptly reverse before Oracle's free cash flow turns positive. That bet may be correct. It is not a bet that should be made at any price without recognizing its conditional character.

The annuity is worth owning. The bet on top of it is interesting enough to watch. At or below $80 per share — where the annuity alone justifies the entry price and the OCI trajectory is genuinely free — Oracle becomes straightforwardly compelling regardless of how the AI infrastructure cycle resolves. At $153, the investor is paying for both, and that price requires the catalyst to arrive.