RPO Inflects to $455B, OCI Guide Laid Out at $144B by FY30 — the OpenAI / Stargate Reset: Maintaining Outperform, FV $310–360

Rating Action

This print maintains the Outperform rating we moved to at Q3 FY25, with the underlying thesis now decisively validated and FV expanded.

• Q2 FY25 (Initiating at Hold): Started at Hold with a constructive bias on the cloud and multi-cloud strategy, but flagged capacity-binding revenue conversion, FY25 capex doubling without an FY26 framework, and a +69% YTD stock as reasons to wait.
• Q3 FY25 (Upgrading to Outperform): All four upgrade triggers cleared. $48B added to RPO, capacity to double in CY25 and triple by end-FY26, FY26 reaffirmed at $66B, FY27 raised to ~20%. Upgraded.
• Q4 FY25 (Maintaining Outperform): FY26 raised to >$67B (+16%) with cloud +40% and OCI +70%. RPO >$138B; FY26 RPO guided to >+100%. Stargate framed as upside optionality. Maintained.
• Q1 FY26 (Maintaining Outperform): RPO inflected to $455B (vs. our +100% trigger of ~$275B exit-FY26); OCI 5-year framework to $144B; FY26 capex raised to ~$35B; OpenAI / Stargate-related $300B contract effectively in the books. The bookings step-up is materially above what an aggressive maintenance underwrite would have priced in. Maintaining at expanded FV $310–360.

Results vs. Consensus

The Q1 print landed in line with guide on revenue and EPS. The story is not the print — it is the bookings line and the forward framework.

Segment Performance

Cloud Infrastructure (OCI) — the Inflection Engine

• Revenue: $3.3B (+54% cc), on top of +46% LY. Annualized run-rate ~$13.2B exiting Q1 FY26.
• Consumption +57% cc; demand “dramatically” outstrips supply for the fifth consecutive quarter.
• FY26 OCI guide raised to $18B (~+77% growth) from prior +70% framework.
• 5-year framework laid out: $18B (FY26) → $32B (FY27) → $73B (FY28) → $114B (FY29) → $144B (FY30). Implied ~52% CAGR over the period.
• Catz: “Much of this revenue is already booked in our $455 billion RPO number.” This is the cleanest visibility-to-revenue framework any infrastructure cloud has ever provided publicly.

Cloud Database & Multi-Cloud — the Second Leg

• Cloud database services: $2.8B annualized (+32%); Autonomous Database revenue +43% on top of +26% LY.
• Multi-cloud database revenue +1,529% (off small base) — 34 Oracle data centers now live inside Azure / Google Cloud / AWS; another 37 in build for 71 total.
• Ellison’s $50B reframing: 10% of database support migration to cloud equals ~$50B of cloud revenue (because it includes compute, networking, storage). Multi-cloud removes the cloud-of-choice friction.

Cloud Applications & Strategic SaaS

• Cloud application revenue $3.8B (+10% cc); strategic back-office $2.4B (+16%) — the strategic-SaaS-only metric, cleanly stripped of legacy noise.
• The 100+ embedded AI agents in Fusion are now in customer rollouts; Ellison referenced AI-generated applications as a structural cost advantage that doesn’t exist at Workday / SAP / ServiceNow.

Total Software (Legacy)

Total software revenue $5.7B, −2% — flattening as cloud-substitution accelerates. License sales remain a cloud-migration leading indicator (BYOL pricing).

Key Topics & Management Commentary

The OpenAI / Stargate-Related Contract: Effectively in the Books

Catz did not name OpenAI directly in the $317B QoQ RPO addition, but the public OpenAI commitment of ~$300B over five years is consistent with the magnitude. Catz’s framing was practical:

“We have signed significant cloud contracts with the who’s who of AI, including OpenAI, xAI, Meta, NVIDIA, AMD, and many others. At the end of Q1, remaining performance obligations or RPO now top $455 billion. ... I expect we will sign additional multibillion-dollar customers and that RPO will likely grow to exceed half a trillion dollars.”
— Safra Catz, CEO

Read-through: The Stargate optionality we treated as upside through Q2-Q4 FY25 is now base case. The forward framework explicitly assumes the OpenAI commitment converts to revenue in the FY27-FY30 window. Stargate-additional contracts (for example with SoftBank or other consortium members) remain upside on top of the $455B base.

Inferencing > Training: Now the Operating Hypothesis

Ellison’s prepared remarks were the cleanest articulation we have heard of why Oracle could compound beyond the current AI training upcycle:

“The AI inferencing market will be much, much larger than the AI training market. AI inferencing will be used to run robotic factories, robotic cars, robotic greenhouses, biomolecular simulations for drug design, interpreting medical diagnostic images and laboratory results, automating laboratories, placing bets in financial markets, automating legal processes, automating financial processes, automating sales processes.”
— Larry Ellison, Chairman & CTO

The Oracle 23ai vector database + bundled access to all major LLMs (ChatGPT, Gemini, Grok, Llama) inside the Oracle Cloud is the inferencing offering. The Oracle AI World demo (October) of a ChatGPT-equivalent reasoning experience over a customer’s private database is the catalyst. We treat this as a free option in the FY29-FY30 framework: the $144B OCI guide does not appear to embed material inferencing-market revenue beyond what the existing AI training commitments imply.

Capex Mechanics: Asset-Light, Customer-Accepted Quickly

Derrick Wood’s capex / ROI question drew the most useful operational disclosure of the call. Catz:

“We do not own the property. We do not own the buildings. What we do own and what we engineer is the equipment. ... We put in that equipment only when it’s time, and very quickly assuming that our customer accepts it, we’re already generating revenue right away. ... It’s, in some ways, asset pretty light.”
— Safra Catz, CEO

Ellison’s extension was the operational kicker: a recently-delivered “giant data hall” was customer-accepted in one week versus what could have been “a couple of months.” That compresses the equipment-to-revenue conversion lag dramatically and is meaningful for the FCF trajectory through FY27.

Vendor-financing leverage also flagged: as one of the largest GPU and networking buyers in the world, Oracle is securing better financing terms than smaller competitors — another mitigating factor on the cash-flow trough thesis.

Butterfly: $6M Cloud Region in Three Racks

Ellison disclosed a new product: “Butterfly” — a complete OCI cloud region in three racks for $6M. He claimed competitor equivalents would cost >100x. This is a genuinely new disclosure and a structural advantage for the cloud@customer / dedicated-region market:

• Vodafone referenced as a customer; multiple large enterprises buying their own private OCI regions because they don’t want shared-cloud neighbors.
• Pay-by-consumption pricing while dedicating the region; eliminates the capex burden for the customer.
• 1% of competitor entry-level pricing for equivalent functionality.

This pulls forward TAM that hyperscalers can’t serve without dedicating massive footprints. We model Butterfly as a small but high-margin contributor to FY27-FY28 revenue, with optionality to expand the dedicated-region installed base.

AI-Generated Applications: The Cost-Side Story

Ellison’s strategic comment on application-generation deserves attention because it’s the cost-side complement to the demand-side narrative:

“The latest applications that we are building — we’re not building them. They’re being generated by AI. ... We don’t charge separately for our AI and our applications because our applications are AI. They’re entirely AI.”
— Larry Ellison, Chairman & CTO

Read-through: If Oracle is materially shrinking the human-engineering cost of building applications via its own application generators, the strategic SaaS gross margin trajectory looks better than the Workday/SAP comparable set, where engineering headcount continues to scale with R&D as a percentage of revenue. We do not yet model this in our framework; flag as upside lever to FY27-FY28 strategic SaaS gross margins.

Q2 FY26 Guidance & Forward Framework

Forward framework:

• FY26: Total revenue +16% cc reaffirmed; OCI raised to $18B (+77%); operating income +mid-teens.
• OCI 5-year: $18B (FY26) → $32B (FY27) → $73B (FY28) → $114B (FY29) → $144B (FY30).
• FY26 capex: ~$35B (raised from $25B), “may be a little higher.”
• RPO: “Likely to grow to exceed half a trillion dollars” within FY26.
• Long-range financial targets update: Oracle AI World, Las Vegas, October.

Analyst Q&A — Notable Exchanges

Q&A was the shortest in recent memory (only five questions before the call closed) and the tone across all questioners was unmistakable: this was treated as a generational print. Notable threads:

• John DiFucci (Guggenheim) opened with an appropriately wide-open question on what is driving the forecast. Ellison’s “running out of inferencing capacity” framing on a customer that asked for “all the capacity you have, anywhere in the world” for inferencing was the most disclosure-rich data point on demand strength. DiFucci’s closing “career event” comment was unusually candid for an analyst on a call — consistent with the buyside framing.
• Brad Zelnick (Deutsche Bank) asked if the AI shift is an “extinction event” for app vendors without the database stack. Ellison’s “application-generation generates better applications than we can hand-build” framing plus the integrated-suite argument was the cleanest competitive-positioning answer of the call. The mention of going head-to-head with Epic in healthcare ($) is a verticals-stack disclosure worth tracking.
• Derrick Wood (TD Cowen) probed the $300B+ RPO addition’s capex and cost structure implications. Catz’s asset-pretty-light framing plus Ellison’s “one week customer acceptance” data point was the cleanest pushback on the FCF-trough bear narrative. Vendor-financing edge was a useful incremental disclosure.
• Mark Moerdler (Bernstein) asked the AI training moat-vs-commoditization question. Ellison’s one-sentence answer — “our networks move data very, very fast… if we’re twice as fast, we’re half the cost” — framed the network-fabric-as-moat argument concisely. Catz’s “hard to beat that” was uncharacteristically casual but reinforced the point.
• Alex Zukin (Wolfe Research) closed on the AI Database / inferencing pacing question. Ellison’s framing of CEOs and heads of state asking specifically about secure-private-data inferencing was a useful customer-conversation-level data point.

What They’re NOT Saying

• OpenAI not named directly in the $317B addition. Catz listed OpenAI among the “who’s who” AI customers but did not specifically tie the $317B QoQ RPO move to the OpenAI contract. We treat this as customer-confidentiality discipline rather than a thesis hole. The public press around the OpenAI / Stargate $300B commitment plus the magnitude of the RPO move makes this functionally confirmed.
• No FY27-FY30 capex framework. The OCI 5-year revenue framework was published; the matching capex framework was not. With ~$35B FY26 and a $144B FY30 OCI run-rate, FY27-FY30 capex is plausibly $40B-$60B per year through FY28 before normalization. The cash-flow trough is undefined past FY26; this is the principal modeling overhang.
• No FY26 FCF framework. Catz committed to mid-teens operating income growth and accelerating into FY27, but no FCF guide. With $35B+ capex and $80B+ FY26 revenue, FY26 FCF could plausibly run −$5B to −$10B before recovery in FY27. The financing plan is not addressed.
• No specific debt issuance plan. Cash $11B at quarter end; FY26 capex $35B+; FCF trough deepening. Material debt issuance is implicit in the framework. The mechanics — tenor, rate, market access — have not been disclosed.
• No commentary on Stargate consortium dynamics. The $300B OpenAI commitment may or may not flow through Oracle exclusively; SoftBank, Crusoe, and other Stargate members are referenced in press reports. Oracle’s share-of-Stargate has not been quantified beyond the OpenAI direct contract.
• No disclosure of cloud GP percentage. Now five consecutive calls without a specific OCI gross margin number. The Q4 framework arguments (autonomous-database-as-cloud-controller, equipment-only spend) make a defensible structural case but the absence of a number remains the cleanest single bear hook.
• No commentary on customer concentration on OpenAI. If $300B / $455B RPO is OpenAI, that is ~66% of total RPO concentrated on a single customer with significant operational and financial uncertainties of its own. The risk is not addressed.

Market Reaction

Shares surged approximately +40% on September 10, hitting an all-time intraday high of $345.72 — one of the largest single-session moves for a mega-cap tech name in years. The move was attributed to the combination of (a) the $317B QoQ RPO inflection, (b) the OCI 5-year framework laid out at $144B FY30, and (c) the publicly reported $300B OpenAI commitment that gives the bookings figure a tangible counterparty.

The move was historic but in our view not unwarranted. The market is repricing Oracle from a software incumbent transitioning to cloud (~25-30x EPS multiple) toward an infrastructure-cloud peer with multi-year visibility (potentially 35-45x on the FY27 EPS framework). Our FV range of $310–360 reflects the upper-mid of that re-rating; we do not chase the move past $360 absent further bookings catalysts but we do not exit Outperform — the framework was raised faster than the price.

Street Perspective

The bull case being made on the Street post-print converges on three planks: (1) the $455B RPO with cloud RPO +500% gives Oracle the cleanest multi-year revenue visibility of any large-cap tech name — better than any hyperscaler, better than any AI infrastructure peer; (2) the OCI 5-year framework to $144B by FY30 implies a ~52% CAGR that is supported by booked-and-non-cancelable contracts, not just demand projections; (3) the inferencing market opportunity, which Ellison frames as “much, much larger” than training, is largely free option in the framework given Oracle’s database-installed-base advantage.

The bear case being articulated centers on: (1) ~66% of RPO is plausibly concentrated in a single customer (OpenAI / Stargate) with significant capital and operational uncertainties, creating concentration risk that is not visible in aggregate metrics; (2) FY26 capex of ~$35B and an unbounded FY27-FY30 capex framework drives FCF deeply negative for multiple years, requiring sustained debt issuance and creating equity-dilution-via-leverage risk if AI demand normalizes; (3) the $300B OpenAI commitment implicitly assumes OpenAI’s revenue trajectory supports the contracted spend — a multi-step counterparty risk; (4) Oracle’s capex-to-revenue ratio is now tracking above the hyperscalers’ (~50% at the $35B / $67B ratio), undermining the architectural-cost-advantage thesis; (5) Stargate-related related-party considerations require careful disclosure; the SEC and auditor scrutiny will track.

Our read sides with the bull framing on (1) and (3). Bear framing on (1) is the most material — we explicitly flag concentration on OpenAI as the principal latent risk in the rating. Bear framing on (2) is real but bounded by Catz’s pretty-asset-light framing and the one-week customer-acceptance data point. Bear framing on (4) requires the FY27-FY28 capex framework to test — we will revisit at Oracle AI World.

Model Implications

• FY26 revenue: $67B-$68B (+16%); OCI $18B (+77%); cloud applications ~$15B (+5%-7%); rest of business flat.
• FY27 revenue: $80B-$85B base case; OCI $32B per company framework; software businesses flat-to-modestly-up.
• FY28-FY30 revenue: Inflecting steeply as OCI hits $73B / $114B / $144B per Catz’s framework. FY30 total revenue plausibly $180B-$200B.
• FY26 EPS: $6.40–$6.70 (mid-teens operating income growth less D&A drag); FY27 $7.50-$8.50 base case.
• FY26 capex: $35B+; we model $36B-$40B with upside if Catz’s “may be higher” framing materializes.
• FY26 FCF: −$5B to −$10B base case; recovery to positive in late FY27 as capex/revenue ratio normalizes.
• RPO trajectory: Q2 FY26 expected $475B-$500B base case; Catz’s “exceed half a trillion” comment implies this will land in FY26.
• Capital structure: Net debt expected to materially increase through FY27 as capex and dividends are funded; equity buyback at minimum cadence ($95M Q1 FY26).

Thesis Scorecard

Overall: Thesis decisively confirmed. Six bull pillars now confirmed (with Butterfly as a genuinely new sixth pillar); the inferencing market is the principal upside lever. Bears: cash-flow compression is real but bounded; counterparty concentration on OpenAI is the most material new risk and warrants explicit acknowledgment in the rating. The principal debate has shifted from “is the demand real?” (yes) to “is OpenAI’s revenue trajectory sufficient to underwrite the contracted spend?” — a counterparty-credit question that we model as a tail risk, not a base case.

Action: Maintaining Outperform; fair value $310–360 (~37–43x FY27 EPS framework of $7.50–$8.50; the elevated multiple reflects multi-year visibility from $455B RPO and the OCI 5-year framework). The next concrete catalyst is Oracle AI World (October), which will quantify the inferencing demo and provide the long-range financial-targets update Catz committed to. Holders should reassess sizing if the stock breaks meaningfully through $360 without thesis upgrade. Net: a generational repricing event for ORCL; we sized into Outperform at Q3 FY25 and the framework has been raised faster than the share price has run; the rating compounds.