A Clean, Gain-Free Beat Ends the Slide: OCI Accelerates to +84%, the $30B Funding Plan Lands, and ORCL Rallies 9%. Maintaining Outperform

Results vs. Consensus

Q3 FY2026 Scorecard

Year-Over-Year Comparison

Sequential Comparison (vs. Q2 FY2026)

Quality of the Print

Revenue: The $17.19B result grew 22% in USD and 18% in constant currency, beat the ~$16.91B consensus, and landed above the USD guide range. There was a genuine four-point FX tailwind this quarter, so the constant-currency 18% is the cleaner number, but even on that basis growth accelerated from the +13% cc of Q2. The composition finally broadened: OCI at +84% was the driver, but software returned to modest USD growth (+3%) as the license line stopped falling, services grew 12%, and cloud reached 52% of the total. This is the acceleration the RPO backlog promised, now showing up in recognized revenue, and it is why the guide-topping print mattered more than the modest percentage beat suggests.

Margins: Margins held up better than the depreciation load implied. Non-GAAP operating margin was 43% versus 44% a year ago, a single point of compression against a 22% revenue print, and GAAP operating margin expanded to 32% from 31%. The standout was the first hard OCI-economics disclosure: Magouyrk stated AI-capacity gross margin ran at 32% in the quarter, above the 30% floor, with adjacent services (general-purpose compute, storage, networking, security) adding a higher-margin 10–20% of the spend and the multicloud database layer carrying a 60–80% margin. After a year of asserting a 30–40% target without a number, management finally put one on the table, and it cleared the floor.

EPS & cash: Non-GAAP EPS of $1.79 was a clean, operational beat with only a trivial $132M of non-operating income, in sharp contrast to Q2's gain-flattered $2.26. The blemish remains the cash statement: trailing-twelve-month free cash flow deepened to −$24.7B on ~$18.6B of in-quarter capex, total borrowings reached ~$134.6B after the $30B raise, and interest expense rose 32% to $1.18B. The mandatory convertible preferred introduced a modest new drag ($22M of preferred dividends and eventual share dilution). The trough is real and still deepening; what changed is that the funding for it is now largely secured and capped.

Segment Performance

Revenue by Line — Q3 FY2026

Cloud Infrastructure (OCI) — +84%, Accelerating and Now Profitable-on-Delivery

OCI grew 84% in USD (81% constant currency) to $4.89B, up sequentially from +68%, the fifth straight quarter of acceleration (49% to 52% to 55% to 68% to 84%). AI-infrastructure revenue alone grew 243% year-over-year. The demand picture is unchanged (supply-constrained across GPU and CPU), but the disclosure was richer than usual: Oracle delivered more than 400 megawatts to customers in the quarter with 90% of committed capacity on or ahead of schedule, secured more than 10 gigawatts of power and data-center capacity coming online over three years, tripled its manufacturing sites and quadrupled rack output in the past year, and confirmed AI-capacity gross margin of 32%.

“It is unprecedented to scale a capital-intensive business so quickly while also increasing profitability. Looking at the AI capacity we delivered in Q3, our gross margin for that remained above our 30% guidance at 32%.” — Clay Magouyrk, co-CEO

Assessment: This is the quarter OCI stopped being a faith-based line item. The +84% acceleration proves capacity is converting rather than sitting in backlog, and the 32% gross-margin disclosure retires the single most durable bear hook (that Oracle would not, or could not, show the economics). Magouyrk's framing that the only drag on profitability is construction-in-progress, not delivered capacity, is the right way to read it: delivered halls are profitable, and the blended margin rises as the build-to-online mix matures.

Multicloud Database — +531%, the High-Margin Second Engine

Multicloud database revenue grew 531% year-over-year, and Oracle reached a milestone of global region coverage across all three hyperscaler partners: 33 regions live with Microsoft, 14 with Google, and AWS scaling from 2 regions at the start of the quarter to 8 at quarter-end, with a plan to exit Q4 at 22 AWS regions. Magouyrk put the gross margin on this business at 60–80%, far above the infrastructure line.

Assessment: Multicloud remains the most attractive line in the model because it monetizes Oracle's database moat inside competitors' data centers at software-like margins without Oracle owning the building. The AWS ramp (2 to 8 to a planned 22 regions) is the tell that the go-to-market friction is gone. At a 60–80% gross margin and triple-digit growth, this is the mix-shift that most directly offsets the lower-margin AI-infrastructure land grab, and it is why the blended OCI margin can rise even as infrastructure scales.

Cloud Applications (SaaS) — +13%, Acceleration Still on the Deferred Line

Cloud applications revenue grew 13% USD (11% cc) to $4.03B, an annualized run rate of $16.1B, with Fusion ERP +14%, Fusion SCM +15%, Fusion HCM +15%, NetSuite +11%, and the industry verticals up 19% combined. Deferred applications revenue again grew 14%, ahead of the 11% constant-currency recognized growth. Oracle took more than 2,000 customers live in the quarter and logged a string of suite 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 line is doing what it did last quarter: a steady low-teens recognized print with a faster-growing deferred balance that management keeps pointing to as the leading indicator of a coming acceleration. The competitive wins over Workday and SAP and the “SaaS apocalypse” rebuttal are credible, but the reacceleration is still on the deferred line rather than the recognized line, so we keep this pillar at “on track, show me.”

Key Operating KPIs

Key Topics & Management Commentary

Overall Management Tone: Confident and, for the first time in two quarters, on offense. Having spent the December call managing anxiety, management arrived with the funding plan already executed, the OCI margin finally quantified, and a rewritten press release that front-loaded the supplemental metrics so the call could focus on strategy. The posture was a company that felt it had answered the market's questions before they were asked, and the one soft spot, the decelerating RPO addition, went largely unremarked.

1. The Clean, Gain-Free Beat

Oracle led with the milestone that Q3 was its first quarter in more than 15 years with both organic total revenue and organic non-GAAP EPS growing 20%-plus in USD. Cloud revenue landed at the high end of guidance, total revenue above the USD guide, and non-GAAP EPS above guidance on both bases.

“We had a tremendous quarter that exceeded expectations across the board. 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, Principal Financial Officer

Assessment: After a December print whose beat evaporated once the Ampere gain was stripped out, the market needed a clean quarter, and it got one. The absence of any material non-operating item is the point: this beat is repeatable in a way Q2's was not, and it re-establishes the credibility of the guide that the FY27 $90B target depends on.

2. The Funding Plan Lands: $30B Raised, CY26 Capped at $50B

The single most important development was that Oracle converted the abstract funding fear into a concrete, capped, and partly-executed plan. In February it announced intent to raise up to $50B of debt and equity in calendar 2026, with an explicit commitment to issue no additional bonds beyond that amount, and within days raised $30B via investment-grade bonds and mandatory convertible preferred stock into a heavily oversubscribed order book. The at-the-market equity portion has not yet been tapped.

“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. We have not yet initiated the at-the-market equity portion of the financing program.” — Doug Kehring, Principal Financial Officer

Assessment: This directly answers the question that drove the December sell-off. A capped $50B envelope, an oversubscribed $30B already in hand, and a reaffirmed investment-grade commitment convert an open-ended dilution-and-leverage fear into a bounded, mostly-funded plan. The mandatory convertible preferred is a genuine cost (dilution plus preferred dividends), but pricing an oversubscribed book at this scale is itself evidence the credit market is comfortable underwriting the build.

3. The Capital-Light Model: $29B of BYOH and Prepaid Contracts

Beyond the raise, Magouyrk detailed the structural mechanism that decouples capex growth from Oracle's own cash requirements. Since the last call, Oracle signed more than $29B of contracts using bring-your-own-hardware and upfront customer payments, and said more than 90% of its secured 10-plus-gigawatt capacity pipeline is fully funded through partners.

“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.” — Clay Magouyrk, co-CEO

Assessment: This is the more durable half of the funding answer. If customers pre-fund or supply the GPUs, Oracle earns the operating margin on designing, building, and running the cluster without carrying the hardware capex, which lowers the cash need and improves returns. The press release's note that “most” of the Q3 RPO increase came from such contracts is the reason management can raise capex to $50B while insisting it will not need to raise incremental funds for those deals. The open question is whether the capital-light mix can stay high as the business scales.

4. RPO $553B: A Deceleration Worth Watching

RPO reached $553B, up 325% year-over-year, but the sequential addition was only ~$29B, down sharply from +$68B in Q2 and the +$317B step in Q1, and it came in slightly below the ~$556B Street whisper. Management framed the composition favorably: most of the increase was large-scale AI contracts funded upfront by customers or supplied with customer GPUs.

“Most of the increase in RPO in Q3 related to large-scale AI contracts where Oracle does not expect to have to raise any incremental funds to support these contracts, as most of the equipment needed is either funded upfront via customer prepayments so Oracle can purchase the GPUs, or the customer buys the GPUs and supplies them to Oracle.” — Q3 FY2026 press release

Assessment: The bookings-cadence deceleration is the quarter's quiet caution flag. After two quarters of headline-grabbing RPO steps, a $29B add against a $553B base is a normalization, and the near-miss on the whisper is a reminder that at this scale the backlog cannot compound at triple digits indefinitely. That said, RPO of this magnitude has shifted the modeling problem from demand estimation to delivery execution, and the higher-quality, self-funding composition of the additions is more valuable than the raw sequential number.

5. TikTok US: A 15% Stake and a Board Seat

Kehring disclosed that in January, TikTok US completed the separation of its US data operations from ByteDance into an independent company in which Oracle now holds a 15% equity stake and a board seat. There is no change to the revenue Oracle earns as TikTok's technology vendor; the stake will be accounted for under the equity method and will begin contributing to non-operating income in Q4 (with a two-month reporting lag).

Assessment: Strategically this deepens a long-standing relationship and converts Oracle from pure vendor to part-owner of a major cloud consumer, but the financial impact near term is modest and sits below the operating line. It is worth tracking as a source of non-operating income volatility (and, like the Ampere gain, a reminder to model off operating results), not as a driver of the core thesis.

6. The FCF Trough Still Deepening

Even with the funding plan, the cash math got heavier. In-quarter capex was roughly $18.6B against ~$7.2B of operating cash flow, trailing-twelve-month free cash flow reached −$24.7B, total borrowings climbed to ~$134.6B, and interest expense rose 32% year-over-year to $1.18B. Cash on the balance sheet jumped to $38.5B, but only because of the $30B raise.

Assessment: The trough is the price of the trajectory and it is still getting deeper before it turns. The mitigants are real (capex is largely contracted and increasingly customer-funded, delivered capacity is profitable, and the raise is done), but investors should continue to model multiple years of negative free cash flow and rising interest expense. This is the constraint that keeps conviction short of the Q1 euphoria even as the operating story strengthens.

7. Applications and the “SaaS Apocalypse” Rebuttal

Sicilia used his prepared remarks to confront the thesis that AI-generated software will kill packaged SaaS, arguing Oracle is the disruptor rather than the disrupted because it is embedding AI agents into complete, mission-critical suites at no additional cost while using AI coding tools to build faster with smaller teams.

“Some smaller or single-focused SaaS players may well be disrupted. But Oracle will not be among them ... we have just built three brand-new CX applications ... we have already delivered well over 1,000 agents right inside our horizontal back-office and industry applications.” — Mike Sicilia, co-CEO

Assessment: The argument is coherent and increasingly backed by competitive wins over Workday and SAP, and the AI-code-generation point doubles as a margin story (fewer engineers per product). It remains a narrative ahead of the recognized-revenue numbers, but it is the credible answer to a sector-wide fear and a reason the applications business is a source of optionality rather than a drag.

8. Guidance: A Confident Q4 and an FY27 Raise to $90B

Management guided Q4 FY26 to total revenue growth of 18–20% constant currency (19–21% USD), total cloud of 44–48% constant currency (46–50% USD), and non-GAAP EPS of $1.92–$1.96 constant currency ($1.96–$2.00 USD). It held FY26 at $67B of revenue and $50B of capex, and raised FY27 total revenue guidance to $90B.

Assessment: The FY27 raise to $90B (roughly +34% on the FY26 base) is the headline forward signal and the clearest statement that management sees the backlog converting faster than previously modeled. The Q4 cloud guide of up to +50% implies no slowdown in the engine, and the held $50B FY26 capex against ~$39B already spent implies a lighter Q4 capex quarter, a welcome hint that the in-year spend peak may be passing.

Guidance & Outlook

The guide reads as a company accelerating with more confidence than a quarter ago. Q4 revenue growth of up to +21% USD and cloud of up to +50% signals the conversion is sustaining, and the FY27 raise to $90B is the concrete forward marker. The tension that will define coming quarters is visible in the EPS guide: revenue is guided to accelerate while non-GAAP EPS growth of +15–17% lags it, as rising interest expense and the new preferred dividends offset the revenue surge.

Implied ramp: With ~$48.2B of FY26 revenue booked through nine months, the $67B full-year guide requires only ~$18.8B in Q4, comfortably within the range. On capex, ~$39B spent against a held $50B implies roughly $11B in Q4, a meaningful sequential slowdown from the ~$18.6B Q3 pace.

Street at: Consensus sat near the low end of the Q4 EPS guide ($1.70 area) and near +20% revenue; the $1.92–$1.96 EPS guide is a clear raise. The FY27 lift to $90B moves the multi-year anchor up.

Guidance style: Oracle guided cloud conservatively and cleared it again this quarter. The credibility of the FY27 $90B figure is now underwritten by a $553B backlog and an accelerating OCI line, which is why the raise reads as a down-payment rather than a stretch.

Analyst Q&A Highlights

The halo effect and FY27 capex visibility

The opening question probed whether the AI-infrastructure boom is pulling through the rest of the business (dedicated regions, sovereign clouds, Alloy, applications) and pressed for any read on FY27 capex. Management confirmed a broad halo and, on capex, deferred a number while stressing the theme that capex is decoupling from Oracle's own capital requirements.

Q: “We are hearing from the field that that halo effect is actually turning into business ... the pipeline is up materially from more traditional cloud workloads, including Dedicated Region, sovereign clouds, even Alloy deals ... can you give us any visibility into CapEx for fiscal 2027?”
— John DiFucci, Guggenheim Securities

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

Assessment: The deferral on an FY27 capex number is a modest disclosure gap, but the “uncoupling” framing is the right lens: gross capex may keep climbing while Oracle's own cash outlay is capped by the BYOH and prepaid structures. The halo answer, that OCI serves as a “budget creator” funding customers' broader transformations, is a credible cross-sell mechanism.

Data-center location strategy as inferencing scales

A question on whether the shift toward inferencing would force Oracle to move data centers closer to population centers drew a clear articulation of Oracle's siting philosophy: latency is proportional to the workload, and for AI reasoning the location matters far less than the hardware architecture.

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

A: “If what you are doing is 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 latency problem right now is not the location of the hardware, it is the type of hardware ... it makes it much more flexible for us to put data centers where power is abundant.”
— Clay Magouyrk, co-CEO

Assessment: A useful, non-obvious point: because reasoning workloads tolerate tens of milliseconds of latency, Oracle can keep siting for cheap, abundant power rather than proximity, preserving its cost advantage. It also implicitly de-risks the “stranded remote data center” worry, since the workloads that fill these sites do not need to be near users.

AI Database and private-data adoption

A question on the AI Database and AI Data Platform asked what customers are actually doing (training private LLMs versus using frontier models on private data) and how confident management is in the promised inflection.

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?”
— Siti Panigrahi, Mizuho

A: “In the early days, a lot of people thought most customers would be doing very specific training of their own large language model. That has largely proven to not be the case. Instead ... people [are] taking the best models and wanting to combine that in a private way with their private data ... we see acceleration of moving that most important private data to cloud environments.”
— Clay Magouyrk, co-CEO

Assessment: The admission that private-LLM training “largely proved not to be the case” is a candid recalibration, and it actually strengthens Oracle's position: the winning pattern (frontier models reasoning over private data) is exactly what the AI Database and multicloud strategy are built to serve, and it is a reason enterprise data keeps migrating to Oracle's cloud.

Value creation from the AI data centers after the debt raise

With the major debt raise complete, a question asked how comfortable management is with the value being created by the AI data-center business net of the cost of the build and the cost of capital, and how the sovereign-cloud opportunity fits.

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

A: “Gross margin in the 30% to 40% range on [the accelerators] continues to hold ... adjacent services [are] typically 10% to 20% of the total spend ... that is without our multicloud database business, which is much higher margin, more in the 60% to 80% range ... the only drag on profitability is [that] we have so much under construction at one time.”
— Clay Magouyrk, co-CEO

Assessment: This is the most complete OCI-economics disclosure Oracle has given: a 30–40% accelerator margin, a higher-margin 10–20% services attach, and a 60–80% multicloud database layer, with construction-in-progress as the only near-term drag. Taken together it is a credible bridge from the negative interim free cash flow to an attractive through-cycle return, and it is what the rating leans on.

Will AI kill application software?

Echoing an investor debate, a question asked what Oracle actually hears from customers on whether AI will kill SaaS, and how to explain Oracle's resilience.

Q: “The theme of SaaS 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 [replace it with] some cobbling together of niche AI features ... we think we are the disruptor because we are embedding the AI right into our applications, at no additional cost.”
— Mike Sicilia, co-CEO

Assessment: The customer-level color (no one is ripping out mission-critical systems for bolted-on AI) is the most grounded rebuttal to the SaaS-apocalypse fear, and the “AI at no additional cost inside the suite” posture is a genuine competitive weapon against best-of-breed vendors. It supports treating the apps business as durable optionality.

Oracle's role as the enterprise AI interaction layer

The closing question asked how Oracle's role evolves as many players vie to be the AI interaction layer across enterprise systems, given the AI Agent Studio now inside Fusion.

Q: “How do you see Oracle'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 mission-critical data gravity matters even more ... we provide a development environment, the AI Data Platform, that allows our customers to easily add their own agents ... in our Fusion accounting system, we will have a complex agent that does the close ... no human beings involved.”
— Mike Sicilia, co-CEO, and Larry Ellison, Chairman & CTO

Assessment: The data-gravity argument is Oracle's strongest structural claim: agents are most useful next to the mission-critical system of record, which Oracle owns. Ellison's “autonomous close” example makes the abstraction concrete, and the open, model-agnostic Agent Studio positions Oracle to host the interaction layer rather than be disintermediated by it.

What They're NOT Saying

1. An FY27 capex number. Management explicitly deferred FY27 capex to next quarter. With the “uncoupling” framing, gross capex could keep climbing even as Oracle's cash outlay is capped, but the absence of a figure leaves the single biggest cash-flow variable unquantified for another quarter.
2. Customer concentration in the $553B RPO. Still no hard concentration figure. The oblique reassurance that “some of the largest consumers of AI Cloud capacity have recently strengthened their financial positions quite substantially” hints at the counterparty question without answering it.
3. Why RPO additions decelerated. The sequential RPO add fell to ~$29B from +$68B, and management did not address the cadence directly. Some of this is mechanical (larger revenue recognition against a bigger base), but a clearer explanation would have been welcome.
4. The BYOH / prepaid share of the pipeline. Oracle quantified $29B of capital-light contracts signed since the last call but did not give the share of the total pipeline that runs on those structures, which is the swing variable in the funding math.
5. A permanent CFO. Doug Kehring again presented as Principal Financial Officer with no timeline for a permanent appointment, a persistent gap in the most financing-dependent stretch in the company's history.
6. The blended OCI operating margin. The 32% AI-capacity gross margin is a real disclosure, but Oracle still does not report a segment-level OCI operating margin, so the standalone profitability of the infrastructure business must still be inferred.

Market Reaction

• Pre-print setup: ORCL entered the print at $149.40, down 23.3% year-to-date and having round-tripped from the September all-time closing high near $328 to a February trough close to $135, roughly a 55% drawdown while the business accelerated. The stock was up only 0.4% over the trailing twelve months and had firmed 4.6% over the prior 30 days. The S&P 500 was down 0.9% year-to-date, so this was an Oracle-specific de-rating, not a market one.
• After-hours / reaction move: Shares jumped on the release and closed the March 11 session up 9.2% at $163.12 (a $13.72 gain) on volume of 83.2M shares versus a 29.4M 30-day average (2.8x). The move was a relief rally on a clean beat and a resolved funding question.

The reaction was the inverse of December, and for the inverse reason. In Q2 the metrics pointed up and the cash statement pointed down, and the market paid for the cash; this quarter the metrics pointed up, the beat was clean, and the funding question was answered, so the market paid for the operations. A 9% rally off a deeply de-rated base is a market beginning to re-underwrite the story, though the stock remains far below its 2025 peak, which is precisely why the risk/reward is attractive: the operating trajectory has strengthened while the price sits at less than half its high.

Street Perspective

Debate: Is the funding question resolved?

Bull view: Yes in substance. A capped $50B CY26 envelope, $30B already raised into an oversubscribed book, $29B of capital-light contracts, and 90%-plus partner-funded capacity convert an open-ended fear into a bounded, mostly-executed plan while preserving investment grade.

Bear view: Not really. Total debt is ~$134.6B and climbing, interest expense is up 32%, free cash flow is −$24.7B and deepening, and a mandatory convertible preferred is now diluting holders. Funding it is not the same as it being cheap or safe.

Our take: The bull has the better of it on near-term risk: the raise is done, capped, and oversubscribed, which is exactly what was missing in December. The bear is right that the through-cycle cost of capital is now materially higher, which is why our fair value reflects a higher risk premium than at the Q1 euphoria.

Debate: Does the RPO deceleration matter?

Bull view: No. A $553B backlog is roughly nine times annual revenue; the question is delivery, not demand, and near-term RPO conversion is accelerating as OCI's +84% shows.

Bear view: Yes. A $29B sequential add versus +$68B and +$317B in the prior two quarters, plus a miss on the whisper, is the first sign the mega-bookings cadence is normalizing, and the entire multiple rests on backlog growth.

Our take: It is a caution flag, not a red flag. Triple-digit RPO growth was never sustainable, and the higher-quality, self-funding composition of this quarter's additions matters more than the raw number. But we would watch the next print closely: if RPO fails to grow meaningfully, the concentration and demand-durability questions get louder.

Debate: Is the stock cheap after a 55% drawdown?

Bull view: At ~$163 on a raised FY27 guide of $90B revenue and an ~$8 EPS framework, the stock trades near 20x forward earnings for a business compounding revenue in the 30s with contracted backlog, a rare combination.

Bear view: The EPS multiple flatters the case because it capitalizes near-term earnings while the negative free cash flow, rising leverage, and dilution are the real cost; on cash generation the stock is not cheap.

Our take: We side with the bull while respecting the bear's cash point. A de-rated stock with an accelerating operating line, a resolved near-term funding question, and a raised multi-year guide is the setup the rating framework rewards; the FCF trough is the reason we size the position for volatility rather than avoid it.

Model Update Needed

Valuation impact: We raise our fair-value range to $295–350 from $285–340, reflecting the de-risked funding, the confirmed 32% AI-capacity margin, and the raised FY27 $90B revenue guide (roughly 33–39x an FY27 non-GAAP EPS framework of ~$8.00–$8.50). At $163.12 post-print, ORCL trades near 20x that framework, a wide discount to fair value that persists because the market capitalizes the cash cost of the build while the $553B backlog it serves is recognized over years. The risk to fair value remains asymmetric to the depth and duration of the FCF trough and to still-undisclosed customer concentration.

Thesis Scorecard Post-Earnings

Overall: Thesis strengthened. Management delivered on the two commitments we set after Q2: conversion accelerated (revenue above guide, OCI +84%) and the funding plan was made concrete, capped, and mostly executed. Two long-standing bear points eased in the same quarter (funding de-risked, OCI margin disclosed at 32%). The offsetting caution is the RPO-cadence deceleration and a still-deepening FCF trough, which keep this an Outperform with respect for the risks rather than a return to the Q1 euphoria.

Action: Maintain Outperform; fair value $295–350. A clean, gain-free beat, an accelerating OCI line, a resolved near-term funding question, and a raised FY27 guide, all with the stock at less than half its 2025 peak, is a favorable risk/reward. Conviction improves from the Q2 read. We would upgrade conviction further on a hard customer-concentration disclosure and evidence the RPO cadence has stabilized; we would revisit the rating if OCI growth decelerates sharply, if RPO fails to grow, or if the funding plan expands materially beyond the $50B calendar-2026 cap.