A Clean 20%+ Beat Reverses the Q2 Cash-Flow Panic: OCI Reaccelerates to +84% and the Build Goes Customer-Funded — Upgrading to Outperform

Results vs. Consensus

Oracle guided to a solid quarter at the December call and delivered above it on every line that matters, with the crucial difference from Q2 being that the earnings beat is real operating performance rather than a portfolio gain. This is the "cleaner data" the Hold rating said it was waiting for.

Q3 FY2026 Scorecard

Year-Over-Year Comparison

Sequential Comparison (the acceleration tell)

Quality of the Beat

Revenue: The +22% USD print (+18% cc) beat the ~$16.9B Street bar by roughly $270M and, more importantly, accelerated from the +13% cc pace of Q2. The mix is doing exactly what the thesis requires: cloud (+44% USD) is carrying the company while software erodes on plan (+3% USD, −1% cc) as customers migrate. There were no one-time revenue items; the acceleration is organic, and management underscored that by explicitly calling the quarter "organic" in the release. Currency was a modest tailwind (USD growth ran four points ahead of constant currency), so the cleaner underlying figure is the +18% cc, which is still an acceleration.

Margins: Non-GAAP operating income grew 19% USD to $7.4B, and the non-GAAP operating margin held at 43%, down about a point year over year. That one-point slip is the AI-infrastructure mix at work: cloud and software cost of revenue grew 66% USD against 22% total revenue growth, the same COGS pressure visible in Q2, but this quarter Oracle grew operating income faster than that pressure rather than being swamped by it. The most important margin disclosure was volunteered, not extracted: management put Q3 AI-capacity gross margin at 32%, above the 30% floor of the 30-40% guided band. After five calls with no OCI margin figure, that single data point removes the cleanest bear hook on the model.

EPS: Non-GAAP EPS of $1.79 grew 21% USD and 16% cc, and there is no portfolio gain in it, which is the entire point relative to Q2. GAAP EPS of $1.27 grew even faster (+24%), helped by declining intangible amortization flowing through the GAAP line. A lower non-GAAP tax rate (17.5% vs. 19.9% a year ago) was a tailwind, and rising interest expense (+32% YoY to roughly $1.18B) was the offsetting drag as the debt funding the build compounds. Net, this is the first quarter in the fiscal year where the EPS number can be taken at face value.

Segment Performance

Cloud Infrastructure (OCI) — +84%, the Reacceleration

OCI grew 84% USD (81% cc) to $4.9B, an acceleration from +68% in Q2 and the fifth straight quarter of a rising growth rate (the sequence now reads 55%, 68%, 84% across the last three quarters). Within OCI, the AI-infrastructure component grew 243% year over year. The operational proof points were concrete: more than 400 megawatts handed to customers in the quarter, with 90% of committed capacity delivered on or ahead of schedule, and manufacturing capacity that Oracle says it has tripled while quadrupling rack output over the last year. Crucially, management disclosed the segment margin for the first time in a long while.

"Looking at the AI capacity we delivered in Q3, our gross margin for that remained above our 30% guidance at 32%. Now combine that with our other segments of OCI, which have much higher margins, like our database services, and you can see why Oracle Corporation is growing so quickly and profitably." — Clay Magouyrk, co-CEO

Assessment: OCI is doing everything the bull case requires and then some: it reaccelerated at scale, it delivered capacity on schedule, and it finally carries a disclosed margin above the guided floor. The reacceleration to +84% is the single cleanest confirmation that the $553B backlog is converting to recognized revenue rather than sitting as a slide. The remaining question is not whether OCI grows, but what the blended, fully-loaded GAAP margin looks like once depreciation on the CapEx wave seasons; the 32% AI-capacity figure is an encouraging first read but not yet a full-fleet GAAP number.

Multicloud Database — +531%, the Second Engine

Multicloud database revenue grew 531% year over year, and Oracle hit a milestone it framed as strategically important: global region coverage across all three partner clouds. It now runs 33 regions live with Microsoft Azure and 14 with Google Cloud, and it scaled AWS from two live regions at the start of Q3 to eight at the end, with a plan to exit Q4 at 22. Management pegged the margin on this line at 60-80%, far above AI-infrastructure margins, and tied its growth directly to AI adoption: customers are moving their most valuable private data into cloud database services so it can be colocated with the AI agents that reason over it.

"Those partnerships unlock an enormous backlog of demand: our database customers who want to use our database in other clouds. This quarter, we achieved an important milestone: we have global region coverage in all of our partner clouds." — Clay Magouyrk, co-CEO

Assessment: This remains the most under-appreciated line in the model, and it is the highest-quality growth Oracle has. It monetizes the database moat without Oracle owning the data center, so it carries a fundamentally better capital profile than OCI and a much richer margin. A +531% rate off a growing base, plus the "all three clouds now global" milestone, makes multicloud a credible standalone driver within a few quarters and a structural counterweight to the margin dilution of AI infrastructure.

Cloud Applications (SaaS) — +13%, the Acceleration Thesis Firms

SaaS revenue grew 13% USD (11% cc) to $4.0B, a $16.1B annualized run-rate, with the leading indicator once again running ahead of the print: cloud applications deferred revenue grew 14% cc versus 11% in-quarter revenue growth. Underneath the headline, the modules that matter are inflecting: Fusion ERP +14% cc, Fusion SCM +15%, Fusion HCM +15%, NetSuite +11% cc, and industry SaaS +19% combined. Oracle took more than 2,000 customers live in the quarter and reported that median time-to-go-live continues to fall, with a long list of competitive wins over Workday and SAP.

"In constant currency, cloud applications deferred revenue was up 14% versus in-quarter cloud applications revenue growth of 11%, which further supports our acceleration thesis." — Mike Sicilia, co-CEO

Assessment: The apps acceleration we flagged as "neutral, need two quarters to confirm" at Q2 is now firming into a modest, credible reacceleration. Deferred revenue running three points ahead of recognized revenue for a second straight quarter, plus industry SaaS at +19% and the multi-pillar competitive wins, argues 2026's low-teens apps growth is a floor. It will not bend the consolidated growth rate on its own, but it moves apps from a slight drag to a slight contributor and, importantly, is the margin ballast that offsets AI-infrastructure dilution.

Software (Legacy License + Support) — the Expected Bleed

Software revenue was $6.1B, up 3% USD but down 1% cc, as on-premise license and support continue to migrate into the cloud lines. This is the right kind of decline: every dollar that leaves the legacy book tends to reappear at a multiple in cloud revenue. It is also the reason a strong cloud quarter still shows a total-revenue number gated by the drag of the shrinking legacy base.

Assessment: Immaterial to the thesis except as a reminder that cloud has to keep overachieving to hold consolidated growth in the low-20s. The cc decline is consistent with prior quarters and requires no change to the model.

Key Topics & Management Commentary

Overall Management Tone: Confident and, notably, reorganized around disclosure. Management moved the supplemental metrics into the written release ahead of the call and kept prepared remarks tight, then spent its airtime on the two questions the Street cared about: how the backlog becomes profitable recurring revenue, and how the build gets funded without more Oracle cash. The posture was a clear step-change from the December call, where the loudest feature was the silence on cash flow; this time management leaned into the funding mechanics rather than deflecting, and volunteered the 32% AI-capacity margin unprompted. Where it stayed guarded was on customer concentration and on any explicit free-cash-flow crossover date.

1. The Quality-of-Beat Reversal: 20%+ Organic Growth on Both Lines

The framing management chose was deliberate. After a December quarter where the earnings beat was a portfolio gain and the stock fell 12%, Oracle led with the cleanest possible quality signal: organic total revenue and organic non-GAAP EPS both above 20% in USD, the first time in more than 15 years. Coming off a quarter the market punished for low earnings quality, this was the message the print most needed to send.

"Our momentum continues to accelerate with Q3 being the first quarter in over 15 years where both organic total revenue and organic non-GAAP EPS grew at 20% or better in USD." — Doug Kehring, PFO

Assessment: The word "organic" is carrying the weight here, and it is warranted. This is the disclosure that most directly answers the Q2 critique. It re-anchors the earnings-quality debate on operations, and it is the foundation of the rating upgrade.

2. The Customer-Funded Pivot: $29B Signed With No Oracle Cash

The most consequential new disclosure was the operationalization of the capital-light model management sketched at the December call. Oracle has now signed more than $29B of new AI-infrastructure contracts since then using a combination of bring-your-own-hardware and upfront customer prepayments, structures that let it expand capacity without incurring negative cash flow. That reframes the entire funding debate: the incremental backlog no longer requires incremental Oracle capital.

"We have signed more than $29 billion of contracts since then, across multiple customers using that new model. A combination of bring-your-own-hardware and upfront customer payments enables us to continue expanding without any negative cash flow from Oracle Corporation." — Clay Magouyrk, co-CEO

Assessment: This is the single most important development of the quarter for the investment case. The Q2 bear thesis was that a $50B CapEx run-rate against deeply negative FCF made the backlog a balance-sheet endurance test. If new bookings are increasingly customer-funded, the endurance test loses its teeth on the margin, and Oracle keeps the operating economics while shedding the cash burden. The caveat is mix: management would not say what share of the total backlog is customer-funded, so this de-risks the future more than the present.

3. The February Raise: $30B Oversubscribed, a $50B Cap on CY2026

Between quarters, Oracle announced its intent to raise up to $50B in debt and equity in calendar 2026 and, within days, executed $30B of it through investment-grade bonds and mandatory convertible preferred stock. Management stressed two things the credit-anxious Street wanted: the order book was a record and substantially oversubscribed, and the company will not issue additional bonds in calendar 2026 beyond the envelope.

"Within days of the announcement, we raised $30 billion through a combination of investment-grade bonds and mandatory convertible preferred stock, with a record order book that was substantially oversubscribed. As noted in our release, we have not yet initiated the at-the-market equity portion of the financing program." — Doug Kehring, PFO

Assessment: An oversubscribed $30B raise is the market's answer to the "can they even fund this?" question that the November-December credit-spread widening embodied. It functions as the credit-acceptance signal we set as an upgrade trigger at Q2. The two open items are the mandatory convertible preferred, which is future dilution dressed as financing, and the not-yet-initiated at-the-market equity sleeve, which is an explicit dilution overhang investors should carry in the model.

4. RPO at $553B: Deceleration in the Add, Not in the Trajectory

RPO ended Q3 at $553B, up 325% year over year but up only $29B sequentially, a sharp step-down from the +$68B of Q2 and the +$317B of Q1. It also came in marginally below the roughly $556B the Street modeled. Management framed the composition as the reassuring part: most of the increase came from large-scale AI contracts that require no incremental Oracle funding.

"Demand for AI infrastructure, both GPU and CPU, continues to exceed supply. This is directly visible in our $553 billion RPO. I want to share a model for how that RPO turns into profitable recurring revenue." — Clay Magouyrk, co-CEO

Assessment: This is the quarter's one genuine yellow flag, and it deserves to be named rather than buried. A +$29B add after two quarters of step-function bookings signals the backlog is normalizing off an extraordinary base, and a slight miss to the Street bar is the kind of thing that matters more the longer it persists. The bull rebuttal is that RPO of this size makes the forecasting problem one of delivery, not demand, and that conversion (OCI +84%) is now the metric that counts. We agree, with a note to watch the bookings line for two consecutive quarters of soft adds.

5. The RPO-to-Revenue Cascade and the Power Pipeline

Management's centerpiece was an operational walk of how backlog becomes cash-generative revenue: secure power and data-center capacity, construct, install compute, fund the capacity, deliver, recognize. The disclosure with the most forward signal was on power: Oracle has secured more than 10 gigawatts of capacity coming online over the next three years through partners, and more than 90% of that capacity is fully funded through those partners.

"Through our partners, we have secured more than 10 gigawatts of power and data capacity coming online over the next three years. Those infrastructure investments also need funding, and greater than 90% of that capacity is fully funded through our partners, with the remainder planned to finish this month." — Clay Magouyrk, co-CEO

Assessment: Power availability, not demand, is the binding constraint on this business, so a 10-gigawatt three-year pipeline that is more than 90% partner-funded is a substantive de-risking of both the delivery schedule and the capital plan. It is also the operational bridge between the $553B backlog and the +34% FY27 revenue guide. We treat it as the most credible single support for the forward revenue framework.

6. AI-Data-Center Profitability and the Cost of Capital

Pressed on whether the value created by AI data centers justifies the cost of building and funding them, management separated the accelerator economics (the 30-40% gross-margin band, holding) from the adjacent services (general compute, storage, load balancing, security) that ride along at 10-20% of spend and higher margins, and from the multicloud database layer at 60-80%. The framework: blended OCI margin strengthens as the fleet scales, and the only current drag is the volume of capacity simultaneously under construction.

"The reason we are not even more profitable right now, despite the fact that we are continuing to grow EPS, is because we have so much under construction at one time... That capacity, when we deliver it, is all already contracted for at a very profitable rate." — Clay Magouyrk, co-CEO

Assessment: The argument is internally coherent and, for once, partly quantified (the 32% AI-capacity figure). The "drag is construction, not the delivered fleet" framing is the right lens for a hypergrowth build, and if it holds, negative interim FCF is an investment rather than a value leak. The number still to see is a full-fleet, fully-depreciated GAAP OCI margin; management is disclosing the numerator willingly and the denominator reluctantly.

7. Multicloud Database and the "Move Private Data to Cloud" Flywheel

Management tied the +531% multicloud line to an AI-adoption flywheel: to use the best models on their most valuable data, customers first have to move that data into cloud database services, where it can sit next to the agents and support vector embeddings, model-context connectivity, and advanced security. That colocation is the reason the partner-cloud footprint matters and the reason the database business is inflecting alongside AI infrastructure.

"AI is also accelerating the adoption of our database cloud services. The rapid improvement in model coding skills and agentic abilities pushes customers to move their most valuable data into our cloud services." — Clay Magouyrk, co-CEO

Assessment: This is the cleanest articulation yet of why the database moat compounds in the AI era rather than eroding. If AI adoption structurally pulls high-value data into Oracle's cloud database services, the highest-margin line in the portfolio grows as a function of the same AI wave driving the low-margin infrastructure line. That is the mix shift the margin bulls need.

8. The SaaS Apocalypse Rebuttal: Suites, Agents, and the "Disruptor" Framing

A recurring theme was the "AI kills SaaS" narrative weighing on the entire application-software complex. Management rejected it forcefully, arguing that Oracle is the disruptor rather than the disrupted: it is using AI coding tools to build new applications with small teams, it has embedded more than 1,000 AI agents into Fusion and industry suites at no additional charge, and it sells complete, regulated, mission-critical suites that a bolt-on collection of AI features cannot replace. Larry Ellison extended the point into ecosystem automation, using healthcare as the example.

"That is why we think we are a disruptor. That is why we think the SaaS apocalypse applies to others, but not to us." — Larry Ellison, Chairman & CTO

Assessment: The argument is credible for Oracle specifically because of data gravity and suite completeness, and the deferred-revenue acceleration is early evidence the AI features are helping rather than cannibalizing. It is also a tacit acknowledgment that the narrative is real for weaker, single-product SaaS vendors, which is a reason the whole group de-rated and thus part of why Oracle got cheap. The competitive-win list against Workday and SAP is the tangible proof the suite is still taking share.

9. Inferencing and Data-Center Location Strategy

Asked whether the shift toward inferencing forces data centers closer to population centers, management argued the opposite: latency is proportional to the workload, and for the multi-second reasoning tasks that dominate enterprise inferencing, an extra 40 milliseconds of network latency is immaterial. The real latency lever is the accelerator architecture, not the site, which frees Oracle to put capacity where power and land are abundant.

"When you actually talk to customers about use cases where they need lower latency, the latency problem right now is not actually the location of the hardware, it is the type of hardware that is being deployed... it makes it much more flexible for us to go out and put data centers where power is abundant, land is plentiful." — Clay Magouyrk, co-CEO

Assessment: A sensible answer that reinforces the power-first siting strategy behind the 10-gigawatt pipeline. It also signals Oracle intends to stay chip-neutral and flexible, referencing the range of accelerators (NVIDIA, AMD, and emerging designs) it can deploy. This lowers the risk that inferencing demand strands the large centralized campuses now under construction.

10. Sovereign Cloud and Alloy: Full-Stack, Not an Edge Zone

Management positioned sovereignty as a widening pipeline and a genuine differentiator, arguing that its Alloy model delivers sovereign data, sovereign operations, and sovereign contracting with the full OCI stack (including applications and the AI Data Platform) rather than a thin edge zone. The flexibility to draw the sovereignty boundary at a country, a region, or a single enterprise's footprint was cited as a unique selling point.

"We are not simply putting an edge sovereign zone in. We are putting full-stack OCI, which has all of our OCI services... in that sovereign zone as well." — Mike Sicilia, co-CEO

Assessment: Sovereign and dedicated-region demand is a real, higher-margin adjacency that the neocloud competitors cannot easily serve, and it dovetails with the halo effect management described (AI infrastructure pulling through database, apps, and sovereign deals). It is not yet a broken-out number, so we treat it as optionality on the OCI and apps lines rather than a standalone driver.

11. TikTok Stake and the Counterparty-Strengthening Narrative

Two items reframed the counterparty question that has shadowed the OpenAI concentration debate. First, Oracle disclosed a 15% equity stake and a board seat in the newly independent TikTok US entity, accounted for under the equity method with earnings flowing to nonoperating income beginning next quarter, and with no change to the technology-vendor revenue it already earns. Second, management noted in the release that some of its largest AI-cloud customers have recently strengthened their financial positions substantially, a clear reference to the large funding rounds raised across the AI complex.

"Some of the largest consumers of AI Cloud capacity have recently strengthened their financial positions quite substantially. These market dynamics enable Oracle to comfortably meet and likely exceed our revenue growth rate forecast for FY27 and beyond." — Oracle Q3 FY26 press release

Assessment: The counterparty-credit tail risk that has been the loudest bear point is being addressed indirectly rather than head-on: Oracle is not quantifying concentration, but it is pointing at the improving balance sheets of its largest customers. Combined with the fungibility argument from the December call and the diversification of the customer roster, it softens the concentration risk without eliminating it. The TikTok stake is a small, additive item, not a needle-mover.

Guidance & Outlook

The forward story reads as a coherent acceleration. Q4 total-cloud growth of 46-50% USD implies no deceleration in the engine, and the FY27 revenue guide was lifted to $90B, a +34% step-up from FY26's ~17% that is the clearest statement that RPO conversion is beginning in earnest. The revenue and EPS growth rates diverging in the Q4 guide (revenue +20% at midpoint, EPS +16%) is the whole Oracle debate in two numbers: the top line is accelerating as the backlog converts, while margin mix, depreciation, and interest expense temper the bottom line. Management framed FY27 as comfortably meetable, citing the strengthened balance sheets of its largest customers.

Implied Q4 ramp: With nine-month revenue at roughly $48.2B against the reaffirmed $67B FY26 figure, Q4 needs about $18.8B, consistent with the +19-21% USD guide and offering little cushion for a slip. On CapEx, the more interesting math is the step-down: nine-month CapEx of $39.2B against the held $50B FY26 guide implies a Q4 of roughly $10.8B, a sharp sequential drop from Q3's $18.6B and the first tangible sign the customer-funded model is bending Oracle's own capital curve.

Street at: Consensus entering the print sat near $16.9B revenue and $1.70 EPS; the beat plus the FY27 raise to $90B (versus roughly $86.6B on the Street) pushed estimates and targets higher across the board.

Guidance style: Oracle has a multi-quarter pattern of guiding cloud conservatively and clearing it. The notable change this quarter is on CapEx: after two consecutive raises (to $35B, then $50B), management held the FY26 figure, which if delivered would validate the capital-light pivot rather than extend the funding anxiety.

Analyst Q&A Highlights

The halo effect turning into pipeline, and FY27 CapEx visibility

The opening exchange probed whether the AI-infrastructure business is dragging the rest of the portfolio along with it, and pressed, in a second question, for any read on next-year capital spending. Management confirmed a broad halo (apps benefiting from model proximity, OCI acting as a "budget creator" by running workloads cheaper, and a rising sovereign and Alloy pipeline) but deferred on FY27 CapEx, pointing only to the decoupling of CapEx from Oracle's own cash requirements.

Q: "We are hearing from the field now that that halo effect is actually turning into business... can you talk about what seems to be an underlying momentum building in these businesses?... And if I could, on a sort of related topic, can you give us any visibility into CapEx for fiscal 2027?"
— John DiFucci, Guggenheim Securities

A: "The most interesting thing that you should start thinking about is the uncoupling of CapEx with capital requirements from Oracle Corporation... when we have these additional funding mechanisms, there may be additional CapEx, but it does not require out-of-pocket cash from Oracle Corporation... So more to come, John, on the CapEx after next quarter."
— Doug Kehring, PFO

Assessment: The halo answer supports the reacceleration story, and the CapEx deferral is less concerning than it would have been a quarter ago because of the decoupling framing. But investors wanted a FY27 CapEx anchor and did not get one; the "reported CapEx may rise even as Oracle's cash outlay does not" distinction is the key thing to model, and it is real but unquantified.

Data-center siting as inferencing scales

The question tested whether a pivot toward inferencing forces Oracle to move capacity closer to users and fiber, away from the power-rich but remote Texas and Wyoming campuses. Management argued the latency concern is overstated for enterprise reasoning workloads and that the real latency lever is accelerator architecture, preserving the flexibility to site data centers where power and land are abundant.

Q: "As you make a move more into inferencing, are you seeing any reason to try to pivot those locations a little closer to where the users and the traffic are?"
— Mark Murphy, JPMorgan

A: "If what you are doing is you are asking a question of your business that is going to take an AI model several seconds to think about, an extra 40 milliseconds of latency from New York to Wyoming is not going to hurt you... the data center location is actually a very tiny part of that."
— Clay Magouyrk, co-CEO

Assessment: A confident, technically grounded answer that de-risks the concern that the centralized campuses under construction could be stranded by an inferencing shift. It reinforces the power-first siting strategy that underpins the 10-gigawatt pipeline. The tell is the chip-neutral posture: Oracle is optimizing for whatever accelerator delivers the best cost and latency, not betting the siting plan on a single architecture.

AI Database and AI Data Platform inflection

The question asked whether the AI Database inflection management teased at the October Analyst Day is materializing, and what customers are actually doing with private data. Management said the early assumption that enterprises would train bespoke private LLMs has largely not held; instead, customers want to combine the best off-the-shelf models with their private data, which is driving the migration of high-value data into cloud database services.

Q: "What are you hearing from customers about training their private data and building their private LLMs? And how confident are you in seeing the inflection in your AI Database growth that you talked about at the Analyst Day in October?"
— Siti Panigrahi, Mizuho

A: "I think in the early days, a lot of people thought that most customers would be doing very specific training of their own large language model. I think that has largely proven to not be the case. Instead, what I think is incredibly popular... is people taking the best models and wanting to combine that in a private way with their private data."
— Clay Magouyrk, co-CEO

Assessment: The candor about the private-LLM thesis not panning out is a credibility marker, and the pivot (models plus private data via the AI Data Platform and multicloud database) is exactly the flywheel that explains the +531% multicloud line. It reframes the AI Database opportunity from a training story to a data-gravity story, which plays to Oracle's actual moat.

AI-data-center value creation versus the cost of capital, and sovereign cloud

With the major debt raise complete, the question asked how comfortable management is with the value created by the AI-data-center business net of the cost of building and funding it. Management walked the layered economics (30-40% on accelerators, higher-margin adjacent services, 60-80% multicloud database) and argued the only current profitability drag is the volume of capacity simultaneously under construction, all of which is pre-contracted at profitable rates.

Q: "Given the blend of the cost of building out the AI data center and the cost of raising capital to fund the AI data center, how comfortable are you with the values you are creating from the AI data center business itself?"
— Mark Moerdler, Sanford Bernstein

A: "When you combine all of these pieces together, the overall margin profile of OCI continues to strengthen and grows rapidly... The reason we are not even more profitable right now... is because we have so much under construction at one time... That capacity, when we deliver it, is all already contracted for at a very profitable rate."
— Clay Magouyrk, co-CEO

Assessment: This is the most substantive margin answer Oracle has given, and the 32% delivered-capacity figure gives it a factual anchor the December version lacked. The layered-margin logic (accelerators plus adjacencies plus database) is how blended OCI margins expand as the fleet matures. It does not settle the fully-loaded GAAP question, but it moves the burden of proof toward the bulls.

Is AI going to kill application software?

A recurring investor worry (that AI will hollow out packaged SaaS) drew a direct rebuttal. Management said no customer it talks to intends to replace core banking, retail merchandising, or EHR systems with a cobbled-together set of AI features, and that customers instead want to consume as much out-of-the-box AI as Oracle can embed into those suites.

Q: "The theme of SaaS software, application software: is AI going to kill it? I just wanted to hear what you guys are hearing when you talk with customers."
— Raimo Lenschow, Barclays

A: "I have not yet met a customer who tells me they are ready to give away their retail merchandising system, their core banking system... and some cobbling together of niche AI features are going to replace all of that overnight. In fact, you hear quite the opposite... how can we consume as much AI out of the box as you are putting into your applications?"
— Mike Sicilia, co-CEO

Assessment: The answer is grounded in customer conversations rather than assertion, and the deferred-revenue acceleration backs it up. It also implicitly explains part of the setup: the SaaS-apocalypse fear helped de-rate the whole group, and Oracle's suite-plus-agents model is the reason it can argue it is on the right side of the disruption.

Oracle's role as the AI interaction layer

The closing exchange asked how Oracle's role evolves as many vendors vie to be the AI interaction layer across enterprise systems. Management leaned on data gravity, arguing that agents are best built inside the system of record where the mission-critical data lives, and that the AI Agent Studio in Fusion (plus the open AI Data Platform development environment) lets customers and partners build their own agents against that data using any model in the Oracle Cloud.

Q: "How do you see Oracle Corporation's role evolving in a world where many other players are vying to be the AI interaction layer across multiple different enterprise systems and workflows?"
— Brad Zelnick, Deutsche Bank

A: "Data gravity matters here, and I think mission-critical data gravity matters even more... If you are going to build a bunch of AI agents... where would you start? You would start inside the system of record... because that is the data... that is going to be highly relevant and highly specific and add a bunch of context to AI."
— Mike Sicilia, co-CEO

Assessment: The data-gravity argument is Oracle's strongest defensive moat and its best offensive wedge into the agent layer, and Ellison's healthcare ecosystem-automation vision extends it into a TAM-expansion story. The open, model-agnostic Agent Studio is a smart answer to the "interaction layer" competition: rather than fight to own the interface, Oracle sells the data and the environment the interfaces have to run against.

What They're NOT Saying

1. No explicit free-cash-flow crossover framework. Nine-month FCF is running around −$22B and trailing-twelve-month FCF is roughly −$25B, yet management still gave no quarter, year, or cumulative-spend milestone for when consolidated FCF turns positive. The customer-funded pivot makes the future look better, but the crossover date remains undefined, which is the single biggest modeling gap.
2. No quantified customer concentration. The +$29B RPO add was again attributed to "large scale AI contracts" and "multiple customers" without naming them or disclosing what share of the $553B backlog sits with the largest counterparty. The concentration question that has dominated the bear case since Q1 is being softened by reference to customers' strengthened balance sheets, not answered with a number.
3. No FY27 CapEx figure. Management explicitly deferred FY27 capital spending to the year-end call, offering only the decoupling framing. Given two consecutive CapEx raises earlier in the year, investors are right to want the number before treating the $50B FY26 hold as a durable ceiling.
4. The customer-funded mix is undisclosed. The $29B of BYOH/prepaid signings is a powerful de-risking, but management would not say what fraction of the total backlog or of go-forward bookings uses the capital-light model. That mix determines how quickly the FCF trough narrows, and it was left qualitative.
5. The at-the-market equity sleeve and the mandatory convertible. The ATM equity portion of the $50B program has not been initiated, and the $30B already raised includes mandatory convertible preferred; both are dilution that the per-share model must carry, and neither was sized on the call.
6. Governance and legal housekeeping went unmentioned. A securities-fraud class action filed in February and continued insider selling by a co-CEO earlier in the quarter received no airtime. Neither is thesis-breaking, but in a quarter about rebuilding credibility, a clearer posture would have helped.

Market Reaction

• Pre-print setup: ORCL closed at $149.40 on March 10 entering the print, down 23.3% year to date (versus the S&P down 0.9% over the same span) and roughly 57% below the September 2025 all-time high near $328. The 52-week closing range was $122.82 to $328.33, capturing the full round trip from the Q1 FY26 euphoria to a deeply de-rated, oversold base. Trailing-twelve-month performance was essentially flat (+0.4%). This was a stock that had priced a great deal of pessimism.
• Reaction session (March 11): Shares gapped up 11.4% at the open, traded as high as $171.76, and closed at $163.12, up 9.2% (+$13.72) on volume of 83.2M shares versus a 30-day average of 29.4M (2.8x). The move was a high-conviction relief rally, not a low-volume drift.

Why it worked. This is the mirror image of the December reaction. In Q2, an optical EPS beat over a revenue miss and a cash-flow blowout produced a 12% decline as algorithms stripped the Ampere gain. In Q3, a clean operating beat, an OCI reacceleration to +84%, and, above all, the reframing of the funding question (an oversubscribed $30B raise, a hard cap on 2026 bond issuance, and $29B of customer-funded signings) gave the market permission to buy the de-rated stock. The single most important sentence for the tape was management's assurance that it would not issue additional bonds in calendar 2026 beyond the announced envelope.

The de-rate did the heavy lifting. A 9% pop looks large in isolation, but it recovered only a fraction of a stock that had fallen 57% from its high. Entering at roughly 20x a ~$7.4 FY26 EPS and under 20x the newly raised $90B FY27 revenue trajectory, ORCL carried none of the AI-narrative premium it held in September. The reaction reflects a market that had over-discounted the funding risk and, confronted with structural evidence that the risk is smaller than feared, re-rated the stock only partway back toward fair value.

Street Perspective

Debate: Is the funding problem actually solved, or just deferred?

Bull view: The bull case being made on the Street is that the customer-funded model changes everything: $29B of new signings require no Oracle cash, more than 90% of the 10-gigawatt power pipeline is partner-funded, and the oversubscribed $30B raise plus a firm no-more-bonds-in-2026 stance proves credit markets will fund whatever gap remains at investment grade.

Bear view: The bear camp contends nothing structural has changed on cash: Q3 CapEx was the highest yet at $18.6B, trailing FCF is roughly −$25B, the mandatory convertible and the untapped ATM equity are dilution, and management still refuses to name an FCF crossover date. Deferring the FY27 CapEx number is a tell.

Our take: The bull has the better of it on trajectory, the bear on the present. The customer-funded pivot is real and is the most important development of the quarter, but it de-risks the future more than the current cash statement. The right read is that the funding tail risk that drove the November-December sell-off has narrowed materially, which is enough to change the rating even though the FCF trough is not yet behind the company.

Debate: Does the RPO deceleration matter?

Bull view: A growing consensus view holds that at $553B (roughly eight times trailing revenue), RPO has already de-risked years of growth, and the metric that now matters is conversion, where OCI at +84% and total cloud at +44% show the backlog turning into revenue exactly on schedule. A +$29B add is still an add.

Bear view: Skeptics argue the sequential add collapsing from +$317B to +$68B to +$29B, plus a slight miss to the Street bar, is the first crack in the demand narrative that justified the entire re-rating, and that a backlog that stops growing is a backlog whose present value the market will discount.

Our take: We lean bull but flag this as the real watch item. One decelerating add off an extraordinary base is normalization, not a crack; two in a row would be a signal. With conversion accelerating and power capacity secured, the delivery story is intact, but the bookings line has earned closer scrutiny than it got when it was compounding by hundreds of billions.

Debate: What is the right multiple after a 57% drawdown?

Bull view: The bull view is that at roughly 20x FY26 EPS and under 20x a FY27 top line growing 34%, with a disclosed 32% AI-capacity margin and a customer-funded model, ORCL is priced for a funding accident that just became far less likely; a re-rate toward the mid-to-high 20s on FY27 implies substantial upside from the mid-$160s.

Bear view: The bear view is that a company with roughly −$25B trailing FCF, over $130B of debt, and equity dilution ahead should trade at a discount to positive-FCF hyperscaler peers, not at parity, and that the right multiple stays compressed until cash generation is visible.

Our take: We are closer to the bull on the multi-year setup while respecting the bear's cash-flow discipline. The appropriate framing is a high-visibility, long-duration backlog paired with a near-term cash trough that is now partially self-funding. That combination, at a valuation that discounts a receding tail risk, is the setup the rating framework rewards, which is why we upgrade rather than wait for the FCF inflection.

Model Update Needed

Valuation impact: We raise our fair-value range to $215–250 from the $208 midpoint carried at Q2. The change reflects three things this print delivered: a clean operating beat that restores earnings-quality credibility, a customer-funded model that shrinks the funding tail, and a disclosed 32% AI-capacity margin that removes the cleanest bear hook. At the March 11 close of $163.12, the midpoint implies roughly +43% upside, and even the low end implies a low-30s percentage gain. The range sits below the Street's average target because we still discount for the undefined FCF crossover, the concentration that remains unquantified, and the dilution embedded in the financing plan. The framework anchors to roughly 25-29x a FY27 EPS base building toward the high-$8s as the $90B revenue guide converts.

Thesis Scorecard Post-Earnings

Overall: Thesis strengthened. The demand-and-delivery side of the bull case advanced on every front (OCI reacceleration, multicloud inflection, apps firming, and a first hard margin data point), while the capital-side bear case moved from Confirmed against us to Contained: the oversubscribed raise, the held CapEx guide, and the customer-funded signings collectively defuse the funding endurance test that drove the December de-rate. Two risks are genuinely open: the still-unquantified customer concentration, and a newly-emerging watch item in the decelerating bookings add. Neither breaks the thesis; both cap conviction below the Q1 FY26 euphoria.

Action: Upgrading to Outperform from Hold; fair value $215–250. At Q2 we set three upgrade triggers: a clean operating beat with apps acceleration building, a credit-market signal that the funding plan is accepted, and an FCF-inflection framework. The first two cleared decisively (a 20%+ organic beat on both lines; a record oversubscribed raise); the third did not, but the customer-funded pivot substitutes structural de-risking for the explicit FCF framework we asked for. Combined with a stock that fell 57% from its high and now discounts a receding tail risk, the risk/reward has flipped to favorable. Our December note flagged that we would "lean to add toward $175 if reached without a thesis-breaking disclosure"; the stock printed well below that on a thesis-strengthening quarter. Add on any pullback toward the low-$150s; the binding watch-items are the FY27 CapEx figure at year-end, two consecutive quarters of soft RPO adds, and any disclosure on customer concentration.