--- title: "Oracle (Trans): 2027 capex exceeds 90bn, with one-quarter covered by prepayments" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/41705023.md" description: "Below is Dolphin Research's compiled Trans of $Oracle(ORCL.US) FQ4 2026 earnings call. For the earnings analysis, see 'ORCL Plunge? AI Infra Can't Fix the Core Weakness — High Rates, High Debt + Stalled Software'.Section I: Core recap. 1) FY27 full-year guide: total revenue approx. $90bn; non-GAAP EPS est. at $8.05..." datetime: "2026-06-11T04:24:06.000Z" locales: - [en](https://longbridge.com/en/topics/41705023.md) - [zh-CN](https://longbridge.com/zh-CN/topics/41705023.md) - [zh-HK](https://longbridge.com/zh-HK/topics/41705023.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- # Oracle (Trans): 2027 capex exceeds 90bn, with one-quarter covered by prepayments **Below is Dolphin Research's curated transcript of**$Oracle(ORCL.US) **FY26 (FQ4 2026) earnings call. For the earnings take, cf.** [**'ORCL Plunge? AI Infra Can't Patch the Core Weakness — High Rates, High Leverage + Stalling Apps'**](https://longbridge.com/en/topics/41704595)**.** **I. Core earnings takeaways** 1\. **FY27 full-year guide**: total revenue to reach $90 bn; non-GAAP EPS expected at $8.05, up 18% at cc (ex one-time investment gains from AMP and Bloom Energy in FY26). The company reaffirmed its Analyst Day long-term targets: by FY2030, revenue CAGR of 31% and EPS CAGR of 28%. 2\. **FY27 Q1 guide**: total revenue +27%–29% YoY in USD, with cloud revenue +58%–64%. Non-GAAP EPS of $1.72–$1.76, +17%–20% in USD. **Management expects revenue and earnings to accelerate in 2H as more data center MW come online to meet demand.** 3\. **CapEx and financing plans**: **FY27 net cash CapEx outlay ~$70 bn, plus $20–$25 bn of customer prepayments and timing effects** (from third-party manufacturers, not vendor financing). **The sum equals reported CapEx.** To support the plan, FY27 financing of ~$40 bn via debt and equity is expected (incl. the announced ATM). No additional debt financing is planned in calendar 2026. 4\. **Margin trajectory**: **FY27 GPM will dip due to timing differences as DC projects ramp to full revenue and due to mix**. As DCs reach full contracted revenue, infra margins are expected to improve rapidly. **Opex is guided to decline slightly YoY (USD),** with efficiency actions driving operating leverage. 5\. **Key prints (FQ4 / FY26)**: Q4 revenue of $19.2 bn (+21% USD), non-GAAP OP +22% to $8.6 bn, and non-GAAP EPS of $2.11 (+24%; +20% ex one-time gains). Full-year revenue topped $67 bn for the first time; non-GAAP OP $29 bn (+16%), non-GAAP EPS $7.63 (+27%; $6.83 ex one-time gains), and full-year GPM fell by ~500 bps as guided. Operating cash flow was $32 bn (+54%). **RPO reached approx. $638 bn (+356% YoY),** with ~12% expected to be recognized over the next 12 months and another ~34% in months 13–36; both ratios are expected to accelerate over the coming quarters. **II. Call details** **2.1 Management commentary highlights** 1\. **Cloud infrastructure (OCI / AI infra)** a. Q4 cloud infrastructure revenue grew 93%, reflecting strong demand for AI workloads and database services. Management positions OCI as the 'most secure, highest-performance, most flexible, and lowest-cost' infra, with differentiation from tech innovation (e.g., its in-house Acceleron high-performance, low-cost network), supply chain execution (pairing OCI with Fusion apps for flexible, efficient supply chains), and full-stack capabilities in DC design, power distribution, facility layout, and network architecture. b. Management believes AI infra makes the existing cloud market look small, with a TAM of trillions of dollars annually. Coupled with the previously indicated 30%–40% margin structure, OCI should scale into a very large and highly profitable business. c. **$67 bn of AI infra contracts were signed this quarter, most as BYO hardware or prepay,** bringing 'BYO + prepay' contract value to $75 bn. These deals carry no margin haircut versus others — even when customers bring capital, they choose OCI to deliver infrastructure. d. Deliveries are accelerating: **over 1.2 GW delivered in FY26, and nearly 1 GW in FY27 Q1,** roughly equal to the prior four quarters combined. Four customers each signed for more than $8 bn this quarter, and the customer base is becoming more diversified. e. Utilization remains extremely high: 35,000 GPUs across 59 customers expired in Q4, with 92% renewed by those customers. The remaining 8% were resold in the same quarter; **global GPU utilization reached 97.5%.** f. AI coding is where AI value is most tangible, and Oracle is both a provider and a consumer. Demand from Oracle itself and from customers/partners for related compute has not slowed. g. Five flagship sites update: Abilene, Texas has delivered 42% of total capacity, with another 35% in the next 90 days and the balance the following quarter. Shackleford, Texas (signed Aug 2025) has 115 MW online, over a month ahead of plan, with customer deliveries starting in H1 CY2027; New Mexico (signed Sep 2025) will use Bloom fuel cells for clean, efficient power, delivering in H1 CY2027; Saline, Michigan (signed Oct 2025) is ahead on network core, delivering by year-end, with customer deliveries starting in H2 2027; Port Washington, Wisconsin (signed Sep 2025) begins deliveries in H2 CY2027. 2\. **Cloud database (Database / multi‑cloud)** a. Q4 cloud database revenue grew 29%, with multi‑cloud seeing outsized growth — multi‑cloud revenue +404% YoY and orders +325% YoY. Management stressed multi‑cloud is still 'very early' and will keep unlocking new regions and partners (including competitor clouds), positioning it as an outgrowth engine. b. New database AI features: Oracle AI Agent Memory (a library to help developers build agents that can remember, reason, and act in enterprise context) and Oracle Deep Data Security (database-level access rules to precisely restrict what users and their AI agents can view and act on). These innovations are available on Oracle Cloud, partner clouds, and in customer environments. c. Customer case: Vodafone chose OCI Dedicated Region, multi‑cloud database, partner clouds, and Oracle apps in Q4 to modernize operations, with some processes sped up by as much as 60%. 3\. **Cloud apps (SaaS / Fusion)** a. Q4 apps revenue was $4.1 bn, +10% YoY, and SaaS deferred revenue rose 16% YoY (deferred growing faster than in‑quarter revenue, supporting the outlook). Over the past year, Oracle delivered 1,000+ AI agents across suites that can reason, decide, and execute within business processes. b. Customer activity: thousands of new customers this quarter, including 300+ for Fusion alone. Examples include Acceleron (utilities platform), Bright County (Public Safety suite), and Westfield Insurance (Fusion ERP). The VA's EHR rollout continued, adding 4 centers in Michigan in Q4 and another 4 in Ohio in early Jun; Oracle now supports 14 VA medical centers, serving 29,000 clinicians and covering 500,000 beds. The U.S. OPM today awarded an agency‑wide Fusion HCM contract to Oracle (not in Q4 bookings but a strong FY27 apps start). c. Private data + AI: much enterprise data already sits in Oracle databases, and Oracle's full stack lets customers quickly combine AI with their own datasets. Examples: Claro in LatAm chose OCI, field service apps, and AI data platform to automate customer service for 30 mn users; the UK NHS, a Brazilian retailer, and QXO are also advancing on Oracle infra/database and apps. 4\. **AI monetization and pricing model innovation** a. Introducing agentic pricing to align price with customer value: many AI innovations in core apps remain included at no extra charge, while customers can purchase 'token bundles' to easily and predictably buy more agent compute. b. Rolling out outcome‑based models that tie pricing directly to realized value. For example, a recruiting agent priced by candidates screened, and a hospitality upsell agent priced as a % of consumer upsell transactions. A limited rollout of token bundles began in Q4, with 33 customers including Aon and Liberty Energy purchasing tokens for more advanced reasoning. **2.2 Q&A** **Q: CapEx ran above expectations this quarter. With memory and other component costs surging, how do the contract structures among you, end customers, and suppliers protect margins over the long term?** A: Two parts to this. On CapEx, looking at Q4, **any upside vs. plan was not due to component prices but to timing — part of my job is to accelerate CapEx so revenue ramps sooner,** so I do not see this as component‑price driven. On component prices themselves, memory and storage have indeed spiked. Our approach is simple: when we sell over a 'certain' time horizon — where capacity is already deployed or we have locked the full cost stack from space, power, energy, and labor to components — we sign fixed‑price contracts. When costs are uncertain due to longer horizons or supply chain risk, we avoid fixed pricing and have mechanisms to pass those costs through. I do not like rising costs, customers do not either, and suppliers would love to meet all our asks. But when costs do rise, we have robust mechanisms to ensure Oracle does not suffer margin compression. **Q: As the new CFO, how do you view the previously stated long‑term targets?** A: That is exactly why we showed that slide today. We are reaffirming those long‑term targets — the CAGRs on the slide — and feel good about them. You can see RPO accumulating to levels that give us more confidence in those goals. From my side, the long‑term targets are fully reaffirmed. **Q: Coming from another capital‑intensive industry, how should investors assess Oracle's progress and returns during this heavy investment phase?** A: As I said on the call, we see very strong returns in our infrastructure business (CPU and GPU). At a high level, I look at ROIC to gauge the model's returns. What we see is **project‑level ROIC in the high‑20s once large programs finish ramping and reach steady state,** based solely on our contracted steady state — not factoring potential upside like GPUs lasting longer without replacement. In cases like 'BYO hardware' where we can maintain or improve margins, ROIC is higher. To be clear on my rough math: **ROIC = after‑tax operating profit + depreciation, divided by total investment (project‑level CapEx).** I hope that frames it, and we are happy to discuss more over coming quarters. **Q: With many entrants into AI DCs (incl. Neo Cloud and SpaceX planning a Spain DC), how does Oracle see itself competitively? How will added capacity affect retention, renewals, new wins, and margins?** A: First and foremost, we stay customer‑focused. The good news, visible in our RPO and growing contracts, is that while a lot is happening in the market, we have a large, diverse customer base from very large to smaller accounts. I spend all day thinking about how to make them as happy as possible. Metrics like very high utilization — even when capacity expires and is returned, it is snapped up immediately — show strong relationships, satisfaction with our products, and comfort with our pricing. I expect many to enter this space, and years into the race demand should still far exceed supply. More players will try to meet that demand, but I am not worried; I am focused on satisfying as much demand as possible at a reasonable margin structure, which is why you see us innovating in commercial models. On renewals, what matters most is the relationship we build from now until renewal. We are fundamentally a services company — customers rely on us daily to run and maintain massive clusters, and when we do that extremely well, we earn very positive relationships and smooth renewals. On margins, I have spent 12 years at Oracle, always on OCI. Building an extremely efficient, highly secure, resilient cloud is hard, and customers see and value the flexibility and comprehensive services we provide. As the market matures and we invest more R&D to make it even more efficient, I believe Oracle can achieve rising margins while offering customers lower prices — our mandate for the past decade, which is why the largest, most demanding customers choose us. **Q: You mentioned two new moves — shifting to outcome‑based pricing and launching token bundles. Why now? How exactly does outcome‑based pricing reduce friction, which modules are included (SaaS, ERP, SCM), and how will this affect bookings?** A: Outcome‑based pricing is not brand‑new to us — we have used it in our construction business, pricing based on managed asset value, GC/sub cash flows, and payments to vendors. As noted, in hospitality and even healthcare, with our new AI automation agents (auto doctor notes, auto lab orders), we can measure and price on 'patient throughput' — exactly what providers care about: moving more people through the system, reducing queues, and improving patient service. What is new is extending this across the full product matrix, including Fusion apps. The challenge is that the 'outcome' is not solely created by us, which is a tricky presentation problem. But because we invested full‑stack, pairing best‑in‑class LLM output with our horizontal and industry apps is easy, giving us a very simple way to measure outcomes for customers. We increasingly hear: 'How much will we spend on AI?' and 'How do we get quick ROI?'. Given we operate in infrastructure and own the horizontal and vertical apps, we inherently generate those outcomes for customers. This positions us to help them understand and align AI budgets to value, which is easy to measure. It is a unique product capability enabled by our full‑stack investments, still very early but resonating well with customers who value transparency and value‑aligned AI spend. On tokens, customers want that too — we keep adding AI features in Fusion and industry apps at no extra charge; when customers want more advanced reasoning and more tokens on models, we provide pre‑packaged bundles. We aim to maximize flexibility across the suite and align pricing with value, and I expect this to keep resonating and support growth as we scale across the portfolio. **Q: Beyond AI momentum, there is noise around software. Can you speak to database (incl. OCI@Azure) and overall database momentum? Apps growth slowed a bit but had strong wins — what are you seeing there?** A: On SaaS apps, we think double‑digit growth on $4.1 bn of revenue is solid, and we are happy with sustained double digits. As noted, **deferred revenue grew 16% this quarter, faster than in‑quarter revenue,** which gives us confidence. Some customers did have longer decision cycles a few quarters back, but they have moved past that — especially for mission‑critical systems. Enterprise software, particularly with AI embedded in our SaaS, is a strong and necessary path to modernize and protect the business, so I expect apps to remain a healthy contributor. On database, innovation drove 29% growth this quarter, with **multi‑cloud revenue up 4x and orders up 3.25x.** The good news is that multi‑cloud database is still very early, and we will continue to unlock new regions and partnerships (including with competitor clouds), making it an outgrowth engine. Beyond multi‑cloud, innovations we added — Deep Data Security, Agent Memory, and better database search — align tightly with our strategy: data strategy and architecture matter. With the market early and so much global data already in Oracle databases, we expect database to grow with continued multi‑cloud investment; as the backbone for apps and database‑based workflows globally, the outlook is very positive. **Q: On BYO hardware and prepayments, these are ~12% of RPO. Where could that mix go based on pipeline? How do BYO deals differ in value vs. GPU‑included deals? Also, please clarify CapEx guidance — does the $70 bn exclude the ~$25 bn of prepayments?** A: **I wish I could tell you we know exactly how this evolves, but today I cannot.** The reality is that commercial models are still evolving — many accelerator types, many customers, and many arrangements. **One capability Oracle offers is to front the capital and depreciate over time to help customers finance usage,** but that is not all we provide and not even the most important for many customers. They contract with us because we can build DCs, design them correctly, ensure security, design the internal networks, install the cloud, and wrap the hardware with a full suite of complementary services — a pile of accelerators alone does not make a usable cloud. You also need general compute, general storage, load balancing, security, identity — everything to make it usable — and Oracle provides the whole stack. Operating these environments is not easy; you do not just buy a rack and wheel it in — these are very complex clusters requiring ongoing care and maintenance of both networks and hardware. Given the ecosystem is changing quickly, you will continue to see innovation and evolution across 'BYO vs. prepay vs. we fund', and I cannot give you precise mix targets. A: One addition: as we noted, these structures carry margins that are in line with or better than our prior deals, which is good economics. We introduced the 'net cash CapEx outlay' metric this quarter to clarify our funding needs. **For FY27, we expect a net cash outlay of about $70 bn, which excludes $20–$25 bn of prepayments — these have timing differences and relate to third‑party manufacturers only, not vendor financing,** not vendor financing, purely third‑party manufacturers. The two together equal reported CapEx. From a funding standpoint, these structures lower the cash CapEx we need to plan the business. Economically, because we receive cash upfront, rather than investing CapEx first and billing later, we use that cash to fund CapEx (not necessarily 100%), improving returns on capital. **Risk disclosure and statements:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [BL.US](https://longbridge.com/en/quote/BL.US.md) - [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md) - [AMP.US](https://longbridge.com/en/quote/AMP.US.md) - [OC.US](https://longbridge.com/en/quote/OC.US.md) - [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md)