RPO Vaults to $523B and OCI Reaccelerates to +68%, but a Capex Shock to ~$50B and −$10B Free Cash Flow Sink the Stock 11%: Maintaining Outperform

Results vs. Consensus

Q2 FY2026 Scorecard

Year-Over-Year Comparison

Sequential Comparison (vs. Q1 FY2026)

Quality of the Print

Revenue: The $16.06B result grew 14% in USD, a third consecutive quarter of double-digit total growth and an acceleration off the +9% Q2 last year, but it landed roughly 1% below the ~$16.2B consensus and at the low end of the company's own +12–14% constant-currency guide. The composition is a barbell: cloud infrastructure reaccelerated to +68% and cloud overall grew 34%, while software fell 5% constant-currency as the license line dropped 23%. The software erosion is not new and not alarming on its own (it is the legacy base being cannibalized by cloud), but it was heavy enough this quarter to pull the total below the Street. This is the first Oracle revenue miss of the current arc, and after a year in which every print was framed against the RPO story, the market's willingness to look past a soft top line has clearly narrowed.

Margins & EPS: The non-GAAP operating margin slipped to 42% from 43% a year ago as data-center ramp costs load ahead of the associated revenue, and non-GAAP operating income grew 10% against 14% revenue growth, and the depreciation-and-ramp drag we have flagged is now visible at the operating line. The reported non-GAAP EPS of $2.26 looks spectacular, but Oracle recognized a ~$2.7B pretax gain from selling its Ampere stake and, unusually, did not exclude it from the non-GAAP figure. Backing the gain out at the 20.8% non-GAAP tax rate leaves underlying operating EPS near $1.53, which is below both the ~$1.64 consensus and the $1.61–$1.65 USD guide the company set three months ago. The honest read is that operating EPS modestly missed; the gain did the rest. A GAAP tax rate of 3.3% (versus the 20.8% non-GAAP rate), driven by deferred-tax effects from last year's legal-entity realignment, further flattered the GAAP line to $2.10.

Cash flow: This is the number that moved the stock. In-quarter operating cash flow was only $2.1B while capex was $12B, producing −$10B of free cash flow against a StreetAccount consensus near −$5.2B, a miss of roughly 2x. On a trailing-twelve-month basis, capex has reached $35.5B and free cash flow is −$13.2B, down from −$5.9B just one quarter ago. The balance sheet is absorbing it through debt: cash rose to $19.2B only because Oracle issued $17.9B of senior notes in the first half, and total borrowings now sit near $108B against $30.5B of equity. The build is real, the demand behind it is contracted, and management insists the funding is manageable, but the cash-flow trough deepened faster this quarter than any prior guide implied.

Segment Performance

Revenue by Line — Q2 FY2026

Cloud Infrastructure (OCI) — +68%, Reaccelerating

OCI grew 68% in USD (66% constant currency) to $4.08B, an acceleration from +55% in Q1, and rose 22% sequentially, the fastest sequential step of the arc. GPU-related revenue grew 177%, and management characterized demand as continuing to outstrip supply across both training and inferencing. The capacity numbers underneath were the most concrete Oracle has disclosed: 147 live customer-facing regions with 64 more planned, close to 400 megawatts of data-center capacity handed to customers in the quarter, 50% more GPU capacity delivered than in Q1, more than 96,000 NVIDIA GB200 accelerators delivered at the Abilene, Texas supercluster, and the first deliveries of AMD MI355 capacity to customers.

“The four segments of our business are growing at incredible rates. This will contribute to a continued acceleration in our infrastructure revenue in the coming quarters.” — Clay Magouyrk, co-CEO

Assessment: OCI is doing exactly what the bull case requires, and the reacceleration to +68% is direct evidence that the backlog is beginning to convert as capacity comes online rather than sitting idle in RPO. The AMD MI355 deliveries and the framing that OCI growth is broad-based (cloud natives, dedicated regions, and multicloud, not just AI training) both matter for the concentration debate. The unresolved question is the same as last quarter: at what through-cycle gross margin does this revenue land once the depreciation on $50B of annual capex fully loads.

Cloud Database & Multicloud — +817% Consumption, the Capital-Light Engine

Cloud database services revenue grew 30%, with Autonomous Database up 43%, but the standout was multicloud database consumption up 817% year-over-year. Oracle launched 11 new multicloud regions in the quarter to reach 45 live across AWS, Azure, and Google Cloud, with 27 more planned imminently and 72 multicloud datacenters now more than halfway built. Two new commercial programs launched: multicloud universal credits (commit once, consume in any cloud) and a multicloud channel-reseller program. Management flagged “billions in identified pipeline.”

Assessment: Multicloud is the most underappreciated line in the model because it monetizes Oracle's database moat inside competitors' data centers without Oracle having to own the building or the power, a fundamentally better capital profile than OCI. An +817% consumption figure is off a small base, but the region build-out and the universal-credits program are exactly the go-to-market removals of friction that turn a demo into a run-rate. This is the cleanest high-margin growth in the P&L and the best structural offset to the FCF-trough bear case.

Cloud Applications (SaaS) — +11%, an Acceleration Thesis Not Yet in the Numbers

Cloud applications revenue grew 11% to $3.9B, with Fusion ERP +17%, Fusion SCM +18%, Fusion HCM +14%, NetSuite +13%, and the industry-cloud verticals (hospitality, construction, retail, banking, restaurants, government, communications) up 21% combined. Strategic back-office SaaS grew 16%. Management completed a major sales reorganization combining the industry-application and Fusion sales teams into a single “One Oracle” selling organization, took 330 customers live in the quarter, and reported 274 healthcare customers live on its clinical AI agent. Crucially, cloud-applications deferred revenue grew 14%, faster than the 11% recognized-revenue growth.

“The cloud application deferred revenue is up 14%. That is higher than the cloud apps revenue growth of 11% ... we expect continued apps growth acceleration.” — Mike Sicilia, co-CEO

Assessment: The 11% headline looks pedestrian against OCI, but the deferred-revenue acceleration is the leading indicator that usually precedes a reacceleration, and the same pattern preceded last year's Q4 apps step-up. The differentiated argument, that Oracle sells complete suites with built-in AI while peers sell best-of-breed point products, is credible and is showing up in the industry-vertical growth rate of 21%. We do not yet model the acceleration; it remains a “show me,” but the setup is real and it is the quiet counter to the sector-wide “SaaS is dead” narrative.

Key Operating KPIs

Key Topics & Management Commentary

Overall Management Tone: Confident on demand, deliberately reassuring on funding, and noticeably lighter on the multi-year framework than the euphoric Q1 call. With the architect of that framework off the call, management leaned on operational proof points (megawatts delivered, regions live, hours to redeploy a data center) rather than on grand five-year revenue arcs, and spent most of its energy pre-empting the funding and cash-flow anxieties that had already knocked the stock down 23% in November. The posture was a company that believes the demand is settled and is now managing the market's discomfort with how it gets paid for.

1. The Backlog Keeps Inflecting: RPO $523B

RPO reached $523.3B, up 438% in USD (433% constant currency) and up $68B sequentially, a 15% quarter-on-quarter step on top of last quarter's historic $317B jump. New commitments from Meta and NVIDIA drove the additions, and management emphasized diversification of the backlog beyond any single counterparty. The composition improved on the axis that matters most for near-term modeling: RPO expected to convert within twelve months grew 40% year-over-year, up from 25% last quarter and 21% a year ago.

“RPO increased by $68 billion in Q2—up 15% sequentially to $523 billion—highlighted by new commitments from Meta, NVIDIA, and others.” — Doug Kehring, Principal Financial Officer

Assessment: The RPO beat (versus the ~$502B Street figure) is the reason the thesis is intact despite the ugly cash statement. More important than the headline is the acceleration in near-term RPO: a 40% year-over-year jump in the twelve-month bucket is the first hard evidence that the mega-backlog is beginning to translate into recognizable revenue on a visible schedule, not just sitting in the long-dated tail. That is precisely the conversion the bull case needs.

2. The Capex Shock: FY26 to ~$50B and the Funding Question

The single most damaging disclosure was the FY26 capex revision. After guiding ~$35B at Q1 (itself up from $25B at Q4 FY25), management now expects FY26 capex about $15B higher, roughly $50B, because the newly added RPO can be monetized quickly and warrants pulling capacity forward. For scale, all of FY25 capex was $21.2B. Kehring framed the equipment as purchased “very late in the data center production cycle,” allowing quick conversion to revenue, and stressed a “foundational” commitment to the investment-grade rating.

“We've been reading a lot of analyst reports ... that show an expectation of upwards of $100 billion for Oracle to go out and complete these build-outs. And based on what we see right now, we expect we will need less, if not substantially less, money raised than that amount to fund this build-out.” — Clay Magouyrk, co-CEO

Assessment: A $15B upward capex revision in a single quarter is the kind of number that resets a stock, and it did. The mitigant is that management tied the increase explicitly to contracted, near-term-convertible RPO rather than speculative capacity, and pushed back directly on the ~$100B funding fears circulating on the Street. But “less, if not substantially less than $100B” is a rebuttal, not a plan, and the refusal to size the raise is the disclosure gap that leaves the bear free to assume the worst.

3. Free Cash Flow: −$10B and a Deepening Trough

Operating cash flow was only $2.1B in the quarter against $12B of capex, producing −$10B of free cash flow; on a trailing-twelve-month basis, FCF is now −$13.2B against $35.5B of capex. This is a regime change for a company that generated double-digit-billions of positive FCF as recently as last year. Management's framing is that the cash mechanics are more favorable than the gross numbers imply, because Oracle incurs no data-center expense until a facility is delivered, and because alternative structures (customers bringing their own chips, vendors renting rather than selling capacity) let it “synchronize payments with receipts.”

“There are other financing options through customers that may bring their own chips ... and suppliers who may lease their chips rather than sell them. Both of these options enable Oracle to synchronize our payments with our receipts and borrow substantially less than most people are modeling.” — Doug Kehring, Principal Financial Officer

Assessment: The bring-your-own-hardware and vendor-rent structures are genuinely important (they are the mechanism that could keep the funding survivable), but they were described this quarter as options, not as a quantified share of the pipeline. Until Oracle puts numbers on how much of the build is customer-funded or vendor-financed, the −$13.2B trailing FCF is what the market will underwrite against, and it is deepening faster than any prior guide implied. This is the binding constraint on the stock.

4. The Ampere Sale and “Chip Neutrality”

Oracle sold its stake in Ampere, the Arm-based CPU designer, recognizing a ~$2.7B pretax gain that flowed through non-operating income and inflated both GAAP and non-GAAP EPS. Ellison framed the divestiture as a strategic pivot to “chip neutrality.”

“Oracle sold Ampere because we no longer think it is strategic for us to continue designing, manufacturing and using our own chips in our cloud datacenters. We are now committed to a policy of chip neutrality ... we will continue to buy the latest GPUs from NVIDIA, but we need to be prepared and able to deploy whatever chips our customers want to buy.” — Larry Ellison, Chairman & CTO

Assessment: The strategic logic is sound: in a market where customers increasingly specify NVIDIA, AMD, or their own silicon, owning a CPU designer is a distraction and a conflict, and chip neutrality aligns Oracle with the customer's chip choice. The problem is presentational: recognizing a $2.7B gain and leaving it in the non-GAAP EPS number, in the same quarter operating EPS missed, invited exactly the “low-quality beat” read the stock delivered. The gain is real and non-recurring; investors should model off the ~$1.53 ex-gain figure.

5. The New Leadership Structure: First Call Under the Co-CEOs

This was the first earnings call since Oracle installed co-CEOs Clay Magouyrk (infrastructure/OCI) and Mike Sicilia (applications/industries), with Larry Ellison remaining Chairman & CTO and Doug Kehring serving as Principal Financial Officer. Safra Catz, who had presented every prior quarter's multi-year framework and had personally underwritten the FY26–FY30 arc, was not on the call, having moved to Executive Vice Chair. The prepared remarks split cleanly along the new lines: Magouyrk on infrastructure and capacity, Sicilia on applications, Ellison on the AI data-platform strategy.

Assessment: The operational logic of the split is defensible: infrastructure and applications are now genuinely different businesses with different capital profiles, and a dedicated operator on each is sensible. But the transition lands at the most financially scrutinized moment in Oracle's history, and it removed from the call the executive most associated with the forward framework precisely when the market wanted reassurance on funding. Ellison remains the strategic anchor, which limits the risk, but the co-CEO structure and the absence of a permanent, named CFO are new variables on the watch list that did not exist a year ago.

6. Fungibility and the Counterparty Question

The most pointed exchange of the call asked directly what Oracle would have to do to repurpose a data center if a large customer “was unable to pay,” a thinly veiled probe of the OpenAI-concentration risk that has shadowed the RPO story since Q1. Magouyrk's answer was that Oracle's AI infrastructure is the same cloud as its general-purpose cloud, that unused capacity can be reallocated to another customer “in less than an hour,” and that with 700-plus AI customers already on OCI, redeployment happens “all day every day.”

“When you ask the question of how long does it take to transfer capacity from one customer to another? It's on the order of hours ... whenever we find ourselves with capacity that's not being used, it very quickly gets allocated and provisioned.” — Clay Magouyrk, co-CEO

Assessment: This is the most reassuring answer management has given on concentration risk, and it is credible: because OCI's AI capacity is architecturally the same as its commercial cloud, a defaulted contract is a reallocation problem, not a stranded-asset problem. It does not eliminate the risk (a large customer failing to pay would still leave a revenue hole and a financing gap), but it reframes the tail from “billions of stranded GPUs” to “capacity re-sold into a demand-constrained market.” That is a materially better place to be.

7. The AI Data Platform: Reasoning on Private Data

Ellison devoted his prepared remarks to the AI data platform, Oracle's pitch that the next and larger wave of AI value is models reasoning over private enterprise data, not training on public data. The platform vectorizes data across Oracle databases, Oracle applications, and non-Oracle sources (object stores, other clouds, competitor databases like MongoDB or Snowflake, bespoke apps), and lets any of the major models (ChatGPT, Gemini, Grok, Llama) reason across all of it from a single query while keeping the data private.

“Training AI models on public data is the largest, fastest-growing business in history. AI models reasoning on private data will be an even larger and more valuable business. Oracle databases contain most of the world's high-value private data.” — Larry Ellison, Chairman & CTO

Assessment: This is the same inferencing-on-installed-base thesis Ellison has advanced for several quarters, now packaged as a named product. It is the strategic option that, if it lands, differentiates Oracle from both the neoclouds (which have no database) and the applications vendors (which have no infrastructure). It contributes little to the FY26 numbers and we treat it as a free option, but it is the reason the applications and database businesses sit at the center of the long-term bull case rather than as afterthoughts to OCI.

8. Guidance: A Confident Q3, a Held FY26, and a $4B FY27 Uplift

Management guided Q3 FY26 to total cloud revenue growth of 37–41% constant currency (40–44% USD), total revenue growth of 16–18% constant currency (19–21% USD), and non-GAAP EPS of $1.64–$1.68 constant currency ($1.70–$1.74 USD, +16–18%). Full-year FY26 revenue expectation was held unchanged at $67B, and the newly added RPO was said to add $4B of incremental FY27 revenue given its near-term convertibility.

Assessment: The Q3 guide is an acceleration (total revenue stepping to +19–21% USD from +14% this quarter, with cloud approaching +44%), and it is the clearest statement that conversion is beginning. The held FY26 revenue against a $15B higher capex bill is the tension in miniature: Oracle is spending far more to deliver the same near-year revenue, with the payoff pushed into FY27 and beyond. The $4B FY27 uplift is a down-payment on that payoff.

Guidance & Outlook

Read as a whole, the guide is coherent and mildly bullish on the top line: revenue growth steps up to +19–21% USD in Q3 as contracted RPO begins to convert, and the near-term RPO bucket accelerating to +40% supports the trajectory. The cost side is what unsettles: the FY26 revenue line is unchanged while capex jumps $15B, so the entire incremental spend is in service of FY27 and beyond, which lengthens the interval investors must fund before the cash turns.

Implied ramp: A held $67B FY26 revenue against a Q3 guide of +19–21% implies the second half runs hotter than the first, consistent with capacity coming online. The revenue is now gated by capacity delivery, not demand.

Street at: Consensus FY26 revenue was already near the $67B guide, so the reaffirmation was neutral on the top line; the shock was entirely on capex and free cash flow, where the Street had modeled a materially shallower trough.

Guidance style: Oracle has a multi-quarter pattern of guiding cloud conservatively and clearing it. The +40–44% USD cloud guide continues that cadence; the credibility question this quarter is not the revenue guide but the unquantified funding plan behind the capex.

Analyst Q&A Highlights

How much capital does the build actually require?

The dominant opening line of questioning went straight at the funding fear that had driven the pre-print sell-off: how much money Oracle needs to raise to fund its AI growth. Management declined to give a single number but explained why the answer is a range (the mix of self-funded hardware, customer-brought hardware, and vendor-rented capacity determines the cash need), and pushed back hard on the ~$100B figure circulating in analyst notes.

Q: “Oracle is clearly the destination of choice for the most sophisticated AI customers, but this is a far more capital-intensive proposition unlike any business Oracle's ever been in before. Very specifically, how much money does Oracle need to raise to fund its AI growth plans ahead?”
— Brad Zelnick, Deutsche Bank

A: “We have a lot of different options for how we go about delivering this capacity ... customers can actually bring their own chips ... some vendors are actually very interested in a model where they rent their capacity rather than selling ... We've read quite a few [reports] that show an expectation of upwards of $100 billion ... we expect we will need less, if not substantially less, than that amount.”
— Clay Magouyrk, co-CEO

Assessment: The “less, if not substantially less than $100B” line was the most-quoted moment of the call and the clearest attempt to cap the funding narrative. It is directionally reassuring, but a refusal to size the raise leaves the single most important capital-structure variable unquantified — the bear's best remaining hook.

How long until OCI reaches the 30–40% margin target?

A recurring concern was the path of OCI margins, given how central they are to eventual cash generation. Management reframed the question away from a per-contract ramp toward a fleet-mix dynamic: the money-losing window on any single data center is only a couple of months, so blended margin is a function of how much of the fleet is online versus still building.

Q: “At the analyst meeting, you said margins for AI workloads for OCI would be in the 30 to 40% range over the life of a customer contract. How long will it take your AI margins across all your OCI data centers to ramp to that level? And what needs to happen to get there?”
— Ben Reitzis, Melius Research

A: “The period of time where we're incurring expenses without that revenue and the gross margin profile ... is really on the order of a couple of months ... What actually matters much more is the overall mix of the data centers that we have online ... the best way to improve margins quickly is to actually go out and deliver capacity faster.”
— Clay Magouyrk, co-CEO

Assessment: The “deliver faster to improve margins” framing is counterintuitive but internally consistent: if the drag per site is short, the blended margin is dominated by the online-versus-building mix, and accelerating delivery pulls the mix toward maturity. It is also unfalsifiable until Oracle discloses a segment-level OCI margin, which it still does not.

What happens to the data center if a large customer cannot pay?

The most pointed question of the call probed the counterparty-concentration risk directly, asking what Oracle would have to do to convert a facility from one customer to another. Management's answer leaned on the architectural fungibility of OCI capacity and the depth of its AI-customer demand.

Q: “Question for Larry and Clay on the fungibility of your infrastructure. What would you have to do to convert a data center from one customer to another, such as [if] one of the larger customers was unable to pay?”
— Brent Thill, Jefferies

A: “How long does it take to transfer capacity from one customer to another? It's on the order of hours ... we have more than 700 AI customers on our platform ... when we give them capacity, they typically spin that capacity up in the order of two to three days ... whenever we find ourselves with capacity that's not being used, it very quickly gets allocated and provisioned.”
— Clay Magouyrk, co-CEO

Assessment: This is the exchange that most matters for the thesis. By establishing that AI capacity is the same fungible cloud as commercial OCI and can be redeployed in hours into a demand-constrained market, management reframed the concentration tail from a stranded-asset catastrophe to a manageable reallocation. It does not erase the revenue-and-financing hole a default would create, but it meaningfully de-risks the worst case.

Selling more up the stack to AI customers

One line of questioning explored whether Oracle can layer database, middleware, and other services onto its AI-infrastructure customers, mirroring how the traditional hyperscalers grew platform revenue on top of raw compute. Ellison used the opening to walk through the three-step database strategy: multicloud availability, vectorization, and the AI data platform.

Q: “How are you thinking about the opportunity to sell additional services such as database, middleware, other pieces of the portfolio ... and what might be some of the similarities or differences with the emerging AI platform-as-a-service market versus the traditional cloud platform-as-a-service market?”
— Tyler Radke, Citi

A: “First multi-cloud. Second, vectorize all the data and make it accessible by all of the popular AI models. Third ... we built an AI lakehouse, the AI data platform, that points to and vectorizes all of your data ... so you can ask a single question and the AI models can find the answer regardless of what data store it's in.”
— Larry Ellison, Chairman & CTO

Assessment: The answer confirms the strategy is to attach high-margin database and data-platform revenue to the low-margin infrastructure land-grab, the same motion that turned raw compute into a profitable platform business for the incumbents. It is the right playbook; the evidence it is working is the +817% multicloud consumption, not yet a broad platform-revenue disclosure.

The per-data-center cash-flow curve

Building on the margin discussion, a question asked management to walk through the cash-flow profile of a single data center (from commitment through hardware purchase to cash-flow-positive) and how those curves aggregate. Management gave an unusually concrete answer that doubled as a defense of the funding model.

Q: “Clay presented a slide at the financial analyst meeting showing the revenue and expenses for a single data center. Can you talk about the cash flow for that same data center, starting with the commitment ... and then how that rolls up across multiple data centers?”
— Mark Moerdler, Bernstein

A: “We incur no cash expenses until [the data center is] fully delivered and provisioned ... In some cases customers bring their own hardware, in which case we don't have any capital expense ... We don't need calculus for this one. Basic arithmetic is enough, because [the cash flows] just layer on top of each other.”
— Clay Magouyrk, co-CEO

Assessment: The “no cash expense until delivered” point is the crux of Oracle's defense that the reported capex overstates the true near-term funding need. It is analytically sound for the lease-covered land/buildings/power, but the revenue-generating equipment (the bulk of the $12B in-quarter capex) still hits cash when purchased, which is why FCF is −$10B. The arithmetic layers both ways.

Why will applications accelerate when SaaS peers are decelerating?

The lone applications question challenged the acceleration thesis head-on: why should Oracle's apps business speed up when every large SaaS peer is slowing? Sicilia's answer combined the One Oracle go-to-market reorganization with a product argument about complete suites and built-in AI.

Q: “Mike, you said applications are gonna accelerate this year. Why the confidence when all your large SaaS peers are seeing just the opposite ... Is it mostly go-to-market? Is there something more about the products?”
— John DiFucci, Guggenheim Securities

A: “We are the only applications company in the world that's selling complete application suites ... over 400 AI features live in Fusion already ... the deferred revenue for apps [is] growing at 14% now, faster than the in-quarter revenue growth of 11%. For all those reasons, I'm optimistic on our apps business going forward.”
— Mike Sicilia, co-CEO

Assessment: The deferred-revenue-above-recognized-revenue signal is the strongest objective evidence for the acceleration claim, and the complete-suite-plus-built-in-AI argument is genuinely differentiated against best-of-breed peers. It remains a forward promise rather than a delivered result, but it is the credible counter to the sector's SaaS-deceleration narrative and worth tracking into Q3.

What They're NOT Saying

1. The size of the funding raise. Management rebutted the ~$100B figure but would not put a number on how much debt and equity Oracle will actually issue to fund a ~$50B capex year against −$13.2B trailing FCF. “Less, if not substantially less” is a direction, not a plan, and the financing mechanics (tenor, mix of debt vs. equity, timing) remain undisclosed.
2. The BYOH / vendor-rent mix. Customer-brought-hardware and vendor-rented-capacity structures were described as the reason the cash need is lower than modeled, but Oracle gave no figure for what share of the pipeline runs on those structures. That single percentage is the swing variable in the entire funding debate.
3. Customer concentration in the $523B RPO. The fungibility answer addressed what happens if a large customer defaults, but management still did not quantify how concentrated the backlog is. With OpenAI plausibly a large fraction of the RPO added over the past two quarters, the concentration of the backlog remains the most important undisclosed number.
4. A segment-level OCI margin. Six-plus calls in, Oracle still asserts a 30–40% OCI margin target without disclosing the current figure. With depreciation on $50B of capex about to load, the standalone economics of the infrastructure business remain a black box.
5. The Ampere gain's effect on the “beat.” Management reported non-GAAP EPS of $2.26 including the ~$2.7B gain and did not lead with the ex-gain operating figure. The omission let the headline read as a large beat when the underlying operating result modestly missed.
6. A permanent CFO. Doug Kehring presented as Principal Financial Officer, but the call gave no timeline for a permanent CFO appointment, a notable gap in the most capital-intensive, financing-dependent stretch in the company's history.

Market Reaction

• Pre-print setup: ORCL closed the regular December 10 session at $223.01, entering the print already bruised, down about 23% in November (its worst month since 2001) and roughly 32% below the September all-time closing high of $328.33. The stock was still +33.8% year-to-date and +25.5% over the trailing twelve months, but down 7.4% over the trailing 30 days. The 52-week closing range of $122.82–$328.33 captures the round trip from the Q1 RPO euphoria into a steady autumn de-rating.
• After-hours / reaction move: Shares fell roughly 11% immediately after the release and closed the December 11 session at $198.85, down 10.8% ($24.16) on volume of 100.6M shares versus a 22.1M 30-day average (4.5x). The AI-infrastructure complex moved in sympathy, with NVIDIA and AMD each off about 1%.

The decline tracked the cash statement, not the backlog. Every metric the bull case is built on (RPO $523B, OCI +68%, near-term RPO +40%, multicloud +817%) moved the right way, and the stock fell anyway. The market is not disputing Oracle's demand; it is repricing the near-term cost of servicing it. A capex revision to ~$50B, −$10B of free cash flow against a −$5.2B expectation, and a revenue miss whose headline EPS was rescued by a one-time gain were more than enough to override a backlog beat. Against a stock that had already given back nearly a third of its value from the September peak, the reaction was the same funding anxiety re-triggered, not a new fear.

Street Perspective

Debate: Is the capex acceleration a demand signal or a red flag?

Bull view: Oracle raised capex $15B because contracted, near-term-convertible RPO justified pulling capacity forward: capacity follows signed demand, and the +40% near-term RPO and +68% OCI prove the conversion is real.

Bear view: A $15B single-quarter revision on top of a doubling last quarter is a company chasing demand it cannot fund from cash flow, with −$13.2B trailing FCF and an unquantified raise ahead, the definition of a capital-intensity trap.

Our take: The bull has the better of it on the demand question (the RPO beat and the near-term-bucket acceleration are hard evidence), but the bear is right that the funding plan is the missing piece. We side with the bull on demand while flagging the unsized raise as the legitimate reason conviction is lower than at Q1.

Debate: Was this a beat or a miss?

Bull view: RPO smashed consensus, OCI reaccelerated, cloud beat, and near-term RPO accelerated, and the metrics that determine the multi-year revenue arc all improved, and one soft revenue line is noise against a $523B backlog.

Bear view: Revenue missed, software fell 5%, operating EPS missed ex the Ampere gain, and free cash flow was 2x worse than expected: strip the accounting and this was a miss dressed up by a one-time gain.

Our take: Both are describing the same quarter accurately, which is the point. This was a good backlog quarter and a poor income-and-cash quarter. For a stock priced on the multi-year cascade, the backlog matters more, but only if the cash trough is fundable, which is why the market weighted the bad half this time.

Debate: Does the leadership transition matter?

Bull view: The co-CEO split puts dedicated operators on two now-distinct businesses, Ellison remains the strategic anchor, and the strategy and framework are unchanged — continuity with sharper focus.

Bear view: Losing the executive who owned the multi-year framework, with no permanent CFO named, at the most financing-dependent moment in the company's history, concentrates execution risk exactly when the market wants steady hands.

Our take: We lean bull but acknowledge the timing is awkward. Ellison's continued presence caps the risk, and the operational proof points were delivered credibly; the watch-item is whether a permanent, credible CFO is named to own the funding narrative.

Model Update Needed

Valuation impact: We trim our fair-value range to $285–340 from $310–360, reflecting a higher funding-and-dilution risk premium on an unchanged backlog thesis (roughly 34–40x an FY27 non-GAAP EPS framework of ~$8.00–$8.50, before the FY27 uplift). At $198.85 post-print, ORCL trades around 24–25x that FY27 framework, a meaningful discount to fair value that exists precisely because the market is capitalizing the cost of the build while the $523B backlog it serves is recognized over years. The risk to fair value is asymmetric to two variables: the size and mix of the funding raise, and the still-undisclosed customer concentration.

Thesis Scorecard Post-Earnings

Overall: Thesis intact on visibility, weaker on quality and cash. The three infrastructure/backlog bull pillars all tripped bullish again; the bear case shifted from “is the capex justified?” toward the harder, more concrete question of “how does Oracle fund a −$13B free-cash-flow run-rate without diluting or over-levering?” The demand is not in doubt; the financing is the binding constraint, and the print's reliance on a one-time gain to clear EPS did the thesis no favors this quarter.

Action: Maintain Outperform; fair value $285–340. A stock down ~39% from its September peak, trading at ~24x an FY27 framework while its contracted backlog grows to $523B, is the setup the rating framework rewards, provided the funding is manageable, which management asserts and we provisionally accept. Own it with lower conviction than at the Q1 euphoria. We would upgrade conviction on a sized, credibly-financed funding plan and a customer-concentration disclosure; we would move to Hold if the FY27 capex framework balloons past ~$100B without a matching funding plan, if equity issuance becomes the primary funding lever, or if OCI margins fail to inflect as the FY26 data-center cohort seasons.