Oracle Conference Call: Signed a $29 billion deal, AI infrastructure "does not consume its own cash flow," "we are the disruptors of SaaS"

On March 10, Oracle held its third-quarter earnings call for fiscal year 2026. Previously, due to market concerns that its future capital expenditure plan could reach up to $50 billion and trigger a debt crisis, coupled with the spread of the notion that "AI will end SaaS," Oracle's stock price had significantly retreated from its peak. However, with the latest quarter's performance exceeding expectations (both revenue and non-GAAP earnings per share increased by over 20%), its after-hours stock price surged by more than 8%. During the call, Oracle's management provided direct and powerful responses to Wall Street's most pressing concerns regarding debt pressure, the survival of SaaS, and the commercialization of AI. Leveraging Strength: Decoupling Capital Expenditure from Cash Flow In the face of the AI wave, cloud computing giants often find themselves in a dilemma: not investing will lead to falling behind, while aggressively investing can drain cash flow and increase debt. Oracle's solution is to build its infrastructure with other people's money. Clay Magouyrk, CEO of Oracle Cloud Infrastructure (OCI), pointed out during the call that Oracle has secured over 10 gigawatts (GW) of power and data center capacity for the next three years. Notably, "over 90% of this capacity is fully funded by partners." In addition to having partners fund the construction of data centers, Oracle has also restructured its transaction model with customers. Clay revealed that the company recently signed contracts worth over $29 billion through innovative models. “This model, which combines customers' 'bring your own hardware' (BYOH) and upfront payments, allows us to continue expanding without consuming any negative free cash flow from Oracle,” Clay explained. This strategy directly addresses market pain points. It means that Oracle's remaining performance obligations (RPO) of up to $553 billion do not need to rely entirely on debt for digestion. CFO Doug Kehring also reiterated that the company will maintain an investment-grade rating and that the scale of bond issuance within the year will not exceed the previously announced $50 billion cap. Doomsday for SaaS? "The Disruptors Are Us" Recently, with the evolution of AI programming tools, the market has generally worried that emerging AI companies will completely disrupt traditional SaaS (Software as a Service) giants. Mike Sicilia, CEO of Application Business, refuted this. “I completely disagree with this view.” Mike stated bluntly, “If we do not adopt these AI tools, they will indeed be a threat. But the reality is, we are adopting them very rapidly.” Oracle's logic is straightforward: the moat of enterprise-level SaaS lies in the attraction of "mission-critical data." Oracle is leveraging its internal small engineering teams to not only rapidly develop three new customer experience (CX) applications that Salesforce currently does not have, but also directly embed thousands of AI agents into existing core backend systems. "Customers won't just throw away their core banking systems, retail sales systems, or medical record systems overnight to replace them with a makeshift AI shell." Mike pointed out that because Oracle holds the most critical data of its customers, the AI-generated results are the most accurate. Chairman and CTO Larry Ellison referred to this evolution as "ecological automation." He directly cited an upcoming financial scenario: "In the near future, you will only need to tell the AI agent to close the books, and it will automatically complete the closing without any human involvement." Delivery Cycle Accelerates, Infrastructure in Production Maintains 32% High Gross Margin The market is not only concerned about the size of the infrastructure pie Oracle has drawn but also about how quickly these investments can be monetized. In this quarter, Oracle's AI infrastructure revenue grew by 243% year-on-year. Behind the explosive growth in business scale is a dramatic shortening of the delivery cycle. "The time from rack delivery to generating revenue has shortened by 60% in the past few months," revealed Clay Magouyrk, CEO of Cloud Infrastructure (OCI), at the meeting. In the third quarter, Oracle delivered over 400 megawatts of capacity to customers, with 90% delivered on time or ahead of schedule. In the early stages of heavy asset investment, profits are often significantly eroded by depreciation and construction costs. However, Clay clearly broke down the numbers: The gross margin of AI capacity delivered in the third quarter remained at 32%, firmly holding the company's previously given guidance bottom line of "above 30%." Additionally, Oracle's multi-cloud strategy has completely opened up sales channels (multi-cloud database revenue surged 531% year-on-year). This high-margin (60%-80%) database business synergizes with the AI infrastructure business, further enhancing the overall profitability structure of OCI. Infrastructure Surge Drags Down Current Profits, Additional Issuance "Boot" Still Not Dropped Although Oracle has reassured the market with its innovative model, hidden dangers still exist in this capital-intensive AI infrastructure race. First, the heavy asset expansion has a "backlash" on current profits. In the rapid expansion "hyper phase," the simultaneous construction of numerous data centers has generated significant sunk costs. OCI business CEO Clay Magouyrk admitted: "The reason we haven't achieved higher profitability now is precisely because too many projects are under construction at the same time... These costs are definitely not zero." He acknowledged that the massive expenditures during the construction period are the most direct factor dragging down profit margins. Secondly, constrained by underlying chips, Oracle's delivery capacity still faces a ceiling. Executives repeatedly mentioned at the meeting that "whether it's GPU or CPU, the demand for AI infrastructure continues to exceed supply." This means that even though Oracle holds a massive remaining performance obligation (RPO) of $553 billion, the actual speed of converting this into revenue is still limited by upstream supply chain capacity allocation bottlenecks from companies like NVIDIA and AMD Finally, the market's biggest concern of "equity dilution" risk has not been completely resolved. Although Oracle has just completed an oversubscribed $30 billion financing and has committed not to issue additional bonds within this calendar year, CFO Doug Kehring clearly indicated at the outset: "We have not yet initiated the 'ATM (At-The-Market)' equity financing portion of our financing plan." This also means that within Oracle's total financing plan of up to $50 billion, the "shoe" of directly issuing new shares to the market still hangs in the balance. After the split of TikTok's U.S. data business, Oracle holds 15% stake During the conference call, the CFO also disclosed a change that may affect subsequent non-operating gains and losses: In January of this year, TikTok's U.S. operations completed a split, becoming an independent company, with Oracle holding a 15% stake and obtaining a board seat. Colin stated that this "does not affect" the related revenue generated from Oracle's services as a technology provider; the equity investment will be accounted for using the equity method, and it is expected to be reflected in the income statement's "non-operating income/loss" starting from Q4 (with a two-month reporting lag). Full translation of the conference call: Oracle's (甲骨文) Q3 FY2026 Earnings Call Event Date: 03/10/2026 Company Name: Oracle (甲骨文) Event Description: Q3 FY2026 Earnings Call Source: Oracle Get more event information and transcripts Meeting Statement Operator (Regina): Hello everyone, thank you for your patience. I am Regina, today's meeting operator. Now, welcome to Oracle's Q3 FY2026 earnings call. To prevent background noise, all lines have been muted. After the speakers' remarks, we will enter the Q&A session. Now, I would like to hand the floor over to Investor Relations Director Ken Bond. Please go ahead. Thank you. Ken Bond, Investor Relations Director: Thank you, Regina. Good afternoon, everyone, and welcome to Oracle's Q3 FY2026 earnings call. Joining us today are: Chairman and Chief Technology Officer Larry Ellison, Chief Executive Officer of Cloud Infrastructure Clay Magouyrk, Chief Executive Officer of Application Software Mike Sicilia, and Chief Financial Officer Doug Kehring. You can find copies of the press release and financial statements on our Investor Relations website, which include supplemental financial details for our most recent quarter, guidance for future performance, GAAP and non-GAAP reconciliation tables, and a list of selected customers who have recently purchased Oracle Cloud services or have gone live on Oracle Cloud Just a reminder, today's discussion will include forward-looking statements, and we will also discuss important factors related to our business. These forward-looking statements are subject to risks and uncertainties that may cause actual results to differ materially from those stated today. Therefore, we remind you not to place undue reliance on these forward-looking statements and encourage you to review our latest financial reports, including our 10-K and 10-Q forms and any applicable amendments. Finally, we are not obligated to update our performance or these forward-looking statements in the event of new information or future events. Before we move into the Q&A session, we will first have a few prepared remarks. Now, I will turn it over to Doug. Doug Kehring, Chief Financial Officer: Thank you, Ken. First, I want to emphasize the changes we made in our earnings press release and the format of this conference call. In the press release, we have clearly listed those supplemental financial metrics that would have been provided during the conference call. This way, each of you can have the written information in advance. Therefore, my remarks for this conference call will be very brief, and then I will turn it over to Mike and Clay, who will provide more substantive insights about our business. After that, all of us, including Larry, will answer your questions. Regarding our performance in the third quarter, we had an extremely outstanding quarter, with all metrics exceeding expectations. As we emphasized in the press release, our growth momentum continues to accelerate: the third quarter is the first quarter in over 15 years where organic total revenue and organic non-GAAP earnings per share both achieved growth of 20% or more in dollar terms. I will briefly mention two more things, and then I will turn it over to the two CEOs. The first thing is that in January of this year, TikTok's U.S. company completed the restructuring to separate its U.S. data business into an independent company, and Oracle currently holds a 15% stake in that company and has a seat on the board. In terms of the impact on our financial condition, the revenue generated from the services we provide as its technology supplier is unaffected. As for this equity investment, we will account for it using the equity method and recognize our share of earnings in the new company from the completion of the investment at the end of January to March 31 in our fourth-quarter financial report, as there is a two-month reporting period time lag. This income will be recorded as non-operating income or loss on our income statement, serving as an incremental supplement to our financial performance. The second thing is that in February of this year, we announced plans to raise up to $50 billion in debt and equity financing and stated that aside from this amount, we do not intend to issue any additional bonds in the calendar year 2026. Within days of announcing this news, we raised $30 billion through a combination of investment-grade bonds and mandatory convertible preferred stock, with subscriptions being extremely popular, setting a new record for subscription orders. As we pointed out in the press release, we have not yet initiated the "At-the-Market (ATM)" equity portion of the financing plan Finally, I must remind everyone that despite the increasing scale and complexity of our business, we are still able to release financial results within just 10 days after the end of the quarter. By using Oracle Fusion, we continue to lead in the speed of closing and releasing financial results compared to any company in the S&P 500 index. This provides us with a significant strategic advantage and gives us the opportunity to help our Fusion customers achieve the same level of efficiency in their businesses. Now, I will hand it over to Mike. Mike Sicilia, CEO of Application Software: Thank you, Doug. As Doug just detailed, we have achieved excellent results across the board this quarter and continue to maintain strong execution. I want to talk about our Software as a Service (SaaS) business. Oracle has the fastest-growing and most complete cloud application suite in the market, without a doubt. Our SaaS solutions are the most comprehensive platform in the industry, highly scalable, reliable, secure, and compliant with regulatory requirements, and our customers trust us to run the systems that operate their businesses. At constant currency, cloud application revenue grew 11% this quarter, reaching an annualized revenue scale of $16.1 billion. Among them, Fusion ERP grew 14%, Fusion SCM grew 15%, Fusion HCM grew 15%, Fusion CX grew 6%, and NetSuite grew 11%. Industry-specific SaaS solutions for hospitality, construction, retail, banking, food service, local government, and telecommunications grew a combined 19%. We are undoubtedly very satisfied with the growth of application software this quarter. In this context, I want to talk about the widely reported "SaaS apocalypse" theory. You have all heard the rhetoric: new companies using AI to quickly write code will declare the death sentence for SaaS. I completely disagree with this view. I believe that if we do not adopt AI tools and their programming capabilities, they will indeed become a threat, but we are adopting them, and at a very fast pace. Oracle is leveraging the best AI programming tools and top developers, not only to accelerate our SaaS business but also to provide solutions that can empower entire ecosystems across various industries. Oracle is using AI programming tools internally, enabling smaller engineering teams to deliver more complete solutions to customers faster. We are not only using AI to build entirely new SaaS products but also embedding AI agents directly into our existing application suite. By leveraging small engineering teams embracing AI, we have just built three brand new customer experience (CX) applications: lead generation and qualification, sales orchestration and automated sales, and our new website generator. In fact, we have just used this website generator to build and launch a brand new oracle.com We build these new CX products to help customers sell better, not just to manage sales forecasts or track email open rates. These are three products that Salesforce.com does not have. Of course, Salesforce.com also lacks OCI (Oracle Cloud Infrastructure), AI data platforms, Fusion ERP, and a complete industry suite. The AI-driven, fully automated end-to-end ecosystem platform is a unique advantage of Oracle. In addition, we have directly embedded over 1,000 AI agents into our horizontal back-office applications and industry applications. This does not even include the agents built by our customers themselves or the clusters of agents we use internally. These are AI capabilities directly embedded in our applications and existing processes. I think a great example is in the healthcare sector. Our brand new AI-driven EHR (Electronic Health Record) system has been deployed in the market with very noticeable effects: we are reducing administrative overhead, enabling clinicians to treat more patients, improving access to healthcare services, and increasing satisfaction among healthcare providers. Another example is in the banking sector. We provide an AI-driven comprehensive SaaS platform that covers everything from commercial banking, retail banking, investment banking, anti-money laundering, financial crime and compliance, payments, supply chain financing, to CX, ERP, and HCM. This banking suite alone contains hundreds of embedded AI agents, all of which are provided to our customers for free. In retail, our AI-supported solutions cover merchandise sales, category planning, supply chain management, point-of-sale (POS) commerce, and of course include ERP, CX, and HCM. In summary, these systems cannot be replaced by a handful of features that are haphazardly added under the guise of AI. So yes, some smaller or single-function SaaS vendors may be disrupted, but Oracle will definitely not be one of them. Now let me highlight some key wins in the application software sector for the third quarter. Undoubtedly, this is just a very short list. Memorial Urban Health Systems chose Fusion ERP, SCM, and HCM, which is an order we won by defeating Workday; the University of New South Wales also chose Fusion ERP and HCM, similarly defeating Workday. Gray Media chose Fusion EPM and ERP, beating Workday and SAP; Investec Bank defeated SAP and chose Fusion EPM and ERP; HID Global Corporation also defeated SAP and chose Fusion ERP and SCM Ethiopian shipping and logistics service companies have chosen Fusion ERP, SCM, and HCM, once again defeating SAP; a large bank on Wall Street has decided to fully adopt Fusion ERP across its entire business and all business units, completely replacing SAP. A public school in a certain county has chosen Fusion ERP, EPM, HCM, and SCM; JM Smucker has chosen Fusion ERP and EPM; Westfield Insurance has chosen Fusion ERP, EPM, HCM, and procurement systems. Mitsubishi UFJ Financial Group (MUFG) is one of our existing cloud and database customers who are now migrating to our Fusion ERP and industry SaaS applications; STC Kuwait, as an important existing technology customer, is migrating its EBS (E-Business Suite) to the cloud to support its business growth. This is just a small part of the significant application software wins this quarter. In the third quarter, we had over 2,000 customers go live. Combining industry applications and Fusion applications, over 2,000 customers have successfully gone live; more importantly, we are seeing the average time to go live for customers is shortening. Several representative cases of go-lives this quarter include: a customer expanded EPM and HCM on top of their enterprise ERP; JM Huber has fully gone live with Fusion ERP and SCM; the UAE Health Services Center has gone live with HCM, enabling a comprehensive HR, payroll, and talent suite to enhance its workforce management; Niagara Bottling has successfully gone live on SCM, migrating its on-premises ERP to Fusion; Seadrill is now fully live with ERP, HCM, SCM, and EPM. Among the 2,000 go-live projects, this is just a very short list, but I hope everyone can see that it is not just about business growth momentum, but also about our multi-pillar product expansion momentum among these customers. Similarly, I also have a streamlined list of key technology wins for the third quarter. Lockheed Martin has chosen OCI high-performance computing to efficiently scale its AI capabilities in its environment; some customers have chosen OCI compute, networking, and storage for AI video and security analytics across all workloads; Lucid Motors has chosen OCI core services to gain data and connectivity capabilities, thus expanding into the European market; Japan's Infomart has chosen OCI for its mission-critical B2B platform; Brazil's Clara is using OCI Alloy to build sovereign AI. Air France-KLM has reached a multi-cloud collaboration agreement, focusing on introducing Oracle Database on Azure, which has brought a 13-fold performance improvement for Air France-KLM, with significantly reduced costs; Activision Blizzard's existing Oracle E-Business Suite business has also adopted Oracle Database on Azure Oracle fully embraces AI in strategic application software, triggering broader enterprise-level conversations between us and our clients, involving our full-stack products—OCI, AI data platform, Fusion applications, and industry suites. The core of these conversations is about how to achieve ecosystem automation, rather than a single application. We discussed the simplified go-to-market model in our last quarter's earnings call, which further facilitated this trend. This enables us to combine the powerful capabilities of the Oracle database, our OCI platform, AI tools, and complete application suite to close more multi-product deals with more clients. At constant exchange rates, cloud application deferred revenue grew by 14%, exceeding the 11% growth rate of cloud application revenue this quarter, further validating our argument that our business is accelerating. Clay, I’ll hand it over to you now. Clay Magouyrk, CEO of Cloud Infrastructure: Thank you, Mike. Alright, I will talk about two parts of our business: multi-cloud databases and AI infrastructure. Both of these businesses are growing at an extremely fast pace. Multi-cloud database revenue grew by 531% year-over-year, and AI infrastructure revenue grew by 243% year-over-year. Both of these businesses are also in a state of high demand, and Oracle has a clear execution plan to quickly convert this demand into high-margin recurring revenue. For decades, the Oracle database has been able to run on any hardware and operating system. But until recently, Oracle Database Cloud Service was only available in a single cloud (i.e., OCI). To bring the best database platform to all cloud platforms, we have established multi-cloud partnerships with Microsoft, Google, and ultimately Amazon. These partnerships have unleashed significant pent-up demand—our database customers have long desired to use our database on other cloud platforms. This quarter, we achieved an important milestone: we achieved global regional coverage across all partners' clouds. We now have 33 Microsoft regions online and 14 Google regions online. We also saw significant growth in AWS (Amazon Web Services): we had 2 AWS regions online at the beginning of Q3, which increased to 8 by the end of Q3, and we expect to have 22 AWS regions online by the end of Q4. AI is also accelerating customer adoption of our database cloud services. Rapid advancements in model programming and agent capabilities are prompting customers to migrate their most valuable data to our cloud services. They need access to the latest AI features to support vector embeddings, MCP server access, and advanced security controls. Customers also need their data to be placed alongside the agents themselves, and our multi-cloud database makes all of this simple. Our multi-cloud architecture brings the best features of Oracle Cloud into our partners' regions, ensuring that we can quickly convert vast channel opportunities into high-margin database service revenue Whether it's GPUs or CPUs, the demand for AI infrastructure continues to exceed supply. This is directly reflected in our remaining performance obligations (RPO) of up to $55.3 billion. I would like to share how these RPOs are converted into a high-profit recurring revenue model and provide some operational data as indicators of our early progress. AI infrastructure begins with data centers and power generation. Through our partners, we have secured over 10 gigawatts (GW) of power and data center capacity, which will be put into use over the next three years. Investments in this infrastructure also require funding, and currently, over 90% of the capacity has been fully funded through partners, with the remaining portion expected to complete financing this month. Once data centers are established, many aspects need to be coordinated: data centers and on-site power generation facilities must be built; computing, networking, and storage equipment must be designed, manufactured, delivered, and installed. All computing capacity within the data centers also requires funding support. We are continuously innovating at every stage. We have optimized data center construction through standardized designs; our supply chain has improved, with more suppliers and deeper partnerships. Over the past year, we have doubled the number of manufacturing plants and quadrupled the total output of racks. We have scaled up the installation process to support multiple delivery phases occurring in parallel. In the past few months, the time from rack delivery to revenue generation has been reduced by 60%. We are also continuously innovating in our business model. In the last earnings call, I shared various ideas on how to achieve incremental growth in AI infrastructure without increasing Oracle's debt or issuing stock. Since then, we have signed contracts worth over $29 billion with multiple customers using this new model. This model, which combines "bring your own hardware" and customer prepayments, allows us to continue expanding without consuming any cash flow from Oracle. Of course, this $29 billion is in addition to other deals we signed this quarter. Ultimately, all of this translates into computing capacity delivered to customers and revenue for Oracle. In the third quarter, we delivered over 400 megawatts of capacity to customers. Ninety percent of the committed capacity was delivered on time or ahead of schedule, consistent with our consistently excellent performance over the past few quarters. This is why customers continue to choose Oracle to meet their infrastructure needs. Investing in AI infrastructure is capital-intensive, but our operating model has been optimized to ensure profitability. Flexible infrastructure design, high utilization rates, rapid handover, and a diverse customer base collectively create an incredible business. The expansion of scale will spread our fixed costs over a larger base, thereby improving profitability. It is unprecedented to increase profitability while rapidly scaling a capital-intensive business. From the AI computing capacity we delivered in the third quarter, the gross margin remains stable at 32%, above our previous guidance of 30% If we also consider our other much more profitable OCI business segments (such as our database services), you will understand why Oracle has been able to grow so rapidly and profitably. The numbers speak for themselves: we are exceeding our revenue and profit targets for fiscal year 2026 and continuously raising our forecasts for fiscal year 2027. This is all thanks to Oracle's successful transformation from a primarily seasonal license business to a cloud business with highly predictable recurring revenue. The demand for AI and advanced computing across the entire economic sector will continue to expand widely. Many successful models, agent platforms, and emerging enterprises will emerge in the future. Today, we support hundreds of cutting-edge AI customers, and more and more customers want to collaborate with us. The infrastructure we build is flexible and versatile, capable of supporting a wide range of workloads from the smallest to the largest. We continuously provide the latest accelerators, from the latest NVIDIA and AMD chips to emerging architecture chips from companies like Cerebras. In summary, we firmly believe that our current investments in data centers, computing power, and customer relationships will only become more valuable over time. I will hand it back to Ken and open the floor for questions. Ken Bond, Head of Investor Relations: Thank you, Clay. Regina, please invite the audience to ask questions. Q&A Session Operator: We will now begin the Q&A session. Our first question comes from John DiFucci of Guggenheim Securities. Please go ahead. John DiFucci, Analyst: Thank you. Wow, that's a lot of information. Listen, I’ll leave the questions about AI infrastructure to others. But we heard Doug talk earlier about the “halo effect” that the AI infrastructure business has on your other businesses. This quarter was very strong, and you mentioned that the growth in RPO came from large AI contracts. At the same time, we are now hearing from the front lines that this halo effect is actually translating into business outside of AI infrastructure. It sounds like the number of systems coming online is stable, but there has been substantial growth in business activity from more traditional cloud workloads, especially in the sales pipeline, which even includes dedicated regional clouds, sovereign clouds, and even the Alloy transactions we are starting to hear about. Aside from the application software transactions that Mike just mentioned, which are often associated together, I understand that these types of transactions may not be as large in scale as the AI transactions. But can you talk about the potential momentum that seems to be building in these businesses? Is my thinking correct? Additionally, if possible, could you share with us the outlook for capital expenditures (CapEx) for fiscal year 2027? Mike Sicilia, CEO of Application Software: Okay, John. I’m Mike, and I’ll take that question. Yes, we are absolutely seeing the halo effect, and I’ll add some details In terms of application software business, we have trained so many AI models on OCI, and these models are deployed so close to our applications that it allows us to embed very high-quality AI services directly as features into our applications. Therefore, we are not only serving these customers and providing training computing power for model vendors, but we are also embedding a large amount of output results directly into the applications. Of course, we will engage in prompt engineering and other work to make it relevant to specific businesses. But the key point is that Oracle is the custodian of critical task data for our customers, and our application software business manages a large amount of business data; our resource allocation is very tight, and the proximity to these models allows us to combine the two, enabling customers to derive value from AI very, very quickly. If you have heard any criticism of AI in the world, it usually complains: "I can't quickly derive value." However, when you package AI as a service and expose the private data we manage to the AI system (obviously at the application level), we see excellent results. I just mentioned some relevant vertical industries, but I believe this is universally applicable across all industries. Another very interesting halo effect is that utilizing our infrastructure—pure OCI infrastructure—can serve as a "budget creator" for customers. You have heard us say before that we are faster and cheaper than anyone else. When customers consider these large-scale applications or infrastructure transformations, we can often help them create budgets simply by migrating their workloads to OCI, thereby funding the transformation, as we can run these workloads faster, more efficiently, and at a lower cost than competitors. Finally, before I hand over the capital expenditure question to Doug, another halo effect is around sovereign AI. Our sovereign strategy is not new, nor is it a reflex to what is happening in the world. Combined with our Alloy strategy, we are seeing a growing sales pipeline globally. Our product forms are very different and have a high degree of differentiation: regardless of how many racks are involved (whether it's 3 racks or 500 racks), we can not only provide smaller product forms but also deliver complete OCI services on top of that. We believe this is a significant competitive advantage in the market. So, when you combine application software, OCI's AI services, and sovereign cloud, yes, this creates quite a significant halo effect. Doug Kehring, Chief Financial Officer: Yes, John, first I have to admit that it’s very creative of you to ask two questions at once, and it’s always interesting to see your questions. Regarding capital expenditure, I think we will report to everyone after the end of this fiscal year and discuss next year's capital expenditure at that time. But I can say a few points. **Clearly, from what Clay just introduced, the most important thing you should start focusing on is the "decoupling" between capital expenditure and Oracle's funding needs. It is evident that when we have these additional financing mechanisms, there may be additional capital expenditures, but this does not require Oracle to spend cash out of pocket, which is very interesting ** On this basis, we remain committed to the goals we discussed last quarter, which is to maintain Oracle's investment-grade rating and keep the financing amount within the range we discussed. Clearly, as we announced, our financing amount for this calendar year is $50 billion. So, John, regarding more information on capital expenditures, we will announce that after the end of the next quarter. John DiFuria, Analyst: Thank you very much, Doug, for the detailed background. And Mike, the logic you presented in your prepared remarks about AI and how Oracle is responding is very clear, and everyone should refer to it. Thank you, great job. Operator: The next question comes from Mark Murphy of JP Morgan. Please go ahead. Mark Murphy, Analyst: Thank you, congratulations on achieving accelerated growth. Clay, as Oracle transitions to a deeper level of AI inference, what do you think is the right strategy for optimizing data center locations? For example, if you have these massive centralized data centers in Texas and Wyoming, while they are close to power resources, they are quite far from population centers and many fiber networks on the East Coast. One can't help but wonder if users and devices are too far away. So, as you move into the inference business, do you think it is necessary to shift the locations of these data centers closer to where users and network traffic are? Clay Maguire, CEO of Cloud Infrastructure: Good question, Mark, this is Clay. First, I want to emphasize our view on inference and how it affects data center deployment. The first point I want to make is that, for a while now, it has primarily been about extensive model training. But the demand for inference is rapidly growing everywhere. I believe this is due to the increasing utilization of the models themselves and the emergence of new use cases—anyone who has recently used Claude in the software space knows that these tools are incredible. They are changing the way we do anything. Therefore, inference will generate huge demand. Now, when you talk about data center locations, you mentioned latency. In fact, there are several reasons for choosing a location: it could be for cost, overall availability, or data sovereignty. Therefore, the basis for choosing a location varies. But let's focus on your point about latency. One thing you need to understand is that latency is relative. That is to say, if you are trying to execute ultra-low latency trades in the stock market, waiting for a 100-millisecond round trip between the East and West Coasts of the U.S. is a bad idea. But if you are posing a question for your business, and the AI model needs a few seconds to think before providing an answer, then the additional 40 milliseconds of latency from New York to Wyoming will have no negative impact on you at all. Therefore, when you really talk to customers who need low-latency use cases, you will find that the fundamental cause of latency issues is not actually the location of the hardware, but rather the type of hardware being deployed. That is why you see so much innovation around these AI accelerators. If you look at the roles of Grok 3 or Positron, all these different types of customers are asking, "How can we not only reduce the cost of inference but also significantly reduce its latency?" "I think if you pay attention to NVIDIA's GTC conference next week, you will see their related announcements. But overall, I believe that as an industry, we need to integrate and reduce latency, and we must first start from different inference architectures. Fortunately, the location of data centers actually accounts for only a very small part of this. This allows us to be more flexible in seeking out places with abundant power and ample land to deploy data centers, thereby truly optimizing to meet this growing demand. Mark Murphy, Analyst: Thank you very much. Operator: The next question comes from Siti Panigrahi of Mizuho Securities. Please go ahead. Siti Panigrahi, Analyst: Great, thank you for taking my question. I want to ask a question about your AI database and AI data platform opportunities. With the recent excitement around AI, enterprises are now starting to adopt cutting-edge large language model (LLM) tools. What have you heard from customers about using their private data for training and building private large language models? How confident are you about the "inflection point of AI database growth" mentioned on the analyst day in October? Clay Maguire, CEO of Cloud Infrastructure: Thank you, I'm Clay. I think this question can be divided into two parts. One is how much adoption we are seeing in building private LLMs, and the other is how much demand we are seeing for leveraging AI to process private data. In the early days, many believed that most customers would train their large language models very specifically. But the reality has largely proven otherwise. Instead, I think the approach that is now very popular and increasingly favored is that people want to leverage the best models and combine them with their private data in a private manner. We are seeing huge demand for this approach. If you just listened to Mike's remarks, you would understand how we are embedding these AI models into our applications, which is one use case. But unfortunately, clearly not everything runs on Oracle's applications, and customers have also written many customized applications. Therefore, we have added a lot of features to the Oracle AI database to make it easy to connect through MCP servers (Model Context Protocol) or natural language to SQL, allowing you to use these models. At the same time, we have the AI data platform product, which is designed to address this issue. You have a large amount of data, which may be application data, custom data from different data lakes and lake-house integrations, or data from structured databases. All of this combined provides you with an intelligent platform on which you can quickly build applications and access all the best models from multiple providers. Throughout the technology stack, we are seeing tremendous momentum. That’s why I talked about the growth presented by our multi-cloud database business in my prepared remarks. What we are seeing is that in order for customers to leverage the latest and greatest AI, they must first be in the cloud, but currently, a significant amount of data has not yet moved to the cloud Therefore, we see that customers are accelerating the migration of their most important private data to cloud environments in order to subsequently leverage this data to experience cutting-edge AI technologies. Citi Panigrahi, Analyst: Great, thank you for the background information. Operator: The next question comes from Sanford Bernstein's Mark Moerdler. Please go ahead. Mark Moerdler, Analyst: Thank you very much for taking my question, and congratulations on a truly outstanding quarter. Great job. I would like to slightly shift the topic to discuss financial matters. Since you have completed large-scale debt financing, could you explain how confident you are in the value created by the AI data center business itself, considering the balance between the costs of building AI data centers and the capital costs of financing them? As a related question, if you don't mind, could you elaborate on sovereign cloud? Can you discuss how you plan to transform the AI data center business into a role as a sovereign cloud AI provider, and how this should impact Oracle's value? Clay Maguire, CEO of Cloud Infrastructure: Okay. I think we should break this question into two parts. I'm Clay, and I'll answer the first half, and then I'll ask Mike to talk about sovereign cloud. When you think about the overall profitability of these AI data centers, there are primarily two aspects. First, how profitable the accelerators themselves are. We have previously guided that the gross margin for this segment is expected to be between 30% and 40%. This still applies to us. Moreover, as we get better at operating these data centers, reducing delivery costs, optimizing network and hardware spending, and electricity costs, we expect this number to continue to increase. Therefore, we are very pleased with this. Another thing to understand is that within these AI data centers, whether for inference or training workloads, the procurement needs go beyond just AI accelerators. There are substantial general computing resources, high-performance block storage or large-scale blob storage, load balancing, authentication, security products, and so on. Typically, about 10% to 20% of total spending is allocated to purchasing adjacent services. When you take these factors into account—depending on the service mix, these adjacent services tend to have higher profit margins—the overall profitability will continue to improve. This doesn't even consider the multi-cloud database business I mentioned earlier, which is a much higher-margin (mostly in the 60% to 80% range) and rapidly growing business. So, when you put all these factors together, the overall profitability situation of OCI is continuously strengthening and growing rapidly. One potential issue that I want to address, which may be overlooked, is that the current constraints on profitability are not due to the computing capacity we have delivered. For example, if I am building a data center with four data halls, when I deliver the first hall, that hall is profitable. Despite our EPS and other metrics continuing to grow, the reason our current profitability level has not reached a higher degree is simply because we have too many projects under construction simultaneously, and these ongoing projects incur some costs Of course, we are doing very well in this regard: we have excelled in shortening construction cycles and also in reducing costs during this period, but these costs are ultimately not zero. Therefore, when our business is in this hyper-growth phase, this is the only drag on profitability. Fortunately, we are becoming increasingly adept at delivering capacity. When we deliver this capacity, it has already been contracted at very high profit margins. Considering these factors, we are confident in the delivered computing capacity and the continuously improving profitability of our AI business. Mike, would you like to talk about sovereign cloud? Mike Siciliano, CEO of Application Software: Yes. Regarding sovereignty, as I just mentioned, I believe we are in a very favorable position. A year ago, the talk about "sovereignty" was mainly about data sovereignty, and there were indeed some solutions in the market that achieved sovereignty at the main data level, but not in disaster recovery (DR) or certain other aspects, such as data potentially being backed up in another country. Of course, this is no longer acceptable now. Today's sovereignty refers to data sovereignty, operational sovereignty, and even contract sovereignty. Our Alloy model is fully capable of achieving all three. By delivering a full-stack solution, we have a significant difference from our competitors in the sovereign cloud: we are not just providing a "sovereign zone" at the edge; we are providing a full-stack OCI. It includes all of our OCI services, and as you mentioned about the profit margin mix, this allows us to run all of our application suites and AI data platforms within the sovereign zone. Of course, the profit margins of some of these services differ from those of pure infrastructure profit margins. I believe we are in a very unique position to provide the full suite of Oracle products in the sovereign zone. This sovereign zone can be made as small or as large as the customer needs. On the other hand, we have a high degree of flexibility in defining the "sovereign boundaries." We typically think of sovereignty in terms of national customer boundaries, but we are also exploring with some enterprise customers. For example, they may operate in multiple countries in Europe or Africa and indeed wish to have a sovereign zone—a sovereign zone that they control and operate within their data centers. They serve customers in specific vertical industries, such as healthcare or retail. If their sovereign zone needs to be defined across these countries through our Alloy, we can fully meet that demand. We believe we have the greatest flexibility in contracts and also the greatest flexibility in delivery. Again, the most important point is that what we deliver within these sovereign zones is the full capability of Oracle—not a subset, not a few edge devices, but the entire OCI ecosystem. Mark Moedler, Analyst: Both answers have been very helpful, thank you very much. Congratulations again. Operator: Thank you. Our next question comes from Raimo Lenschow of Barclays. Please go ahead Raimo Lönnroth, Analyst: Great. Thank you. I also want to express my congratulations. I want to ask about a topic that often confuses us when communicating with investors, which is "SaaS software is dead" because AI will replace it. I just want to hear what you have heard when communicating with clients. Is this a fantasy that only those investors have? Do clients discuss this issue as well? How do you explain it? My personal thought is that much of what you do is deterministic rather than probabilistic. That might be an explanation. But I just want to hear your thoughts again. Thank you. Mike Siciliano, CEO of Application Software: Yes, I'm Mike, and I'll answer this question. From the clients I've talked to, I haven't encountered any client telling me they are ready to abandon their existing retail merchandise systems, core banking systems, demand deposit account systems, or electronic health systems to replace all of that overnight with some patchwork, niche AI features. In fact, what I hear from clients is exactly the opposite. They are asking, "How can we consume the AI features you have built into all your applications out of the box? How can we get these systems up and running as quickly as possible because we believe that is the best way to realize value?" Listen, the systems we run at Oracle, as you know, are highly complex, mission-critical systems that embody decades of industry experience and regulatory compliance accumulation. These systems are the core that our clients use to operate businesses, government agencies, healthcare organizations, and more. I am very satisfied with our position in this area. As I said, we are also embracing AI vigorously. Therefore, we have launched 1,000 AI agents in Fusion. Our banking suite alone contains hundreds of AI agents. So, yes, we believe AI is disruptive. Indeed it is. But we believe we are the disruptors because we are actually embedding AI as a complete function directly into our applications without charging any additional fees. These features are included as part of quarterly upgrades and as part of the regular update cadence within the application suite. Therefore, I do not believe AI has pronounced the death sentence for SaaS (at least not for Oracle), but rather I think it has actually enhanced our position in the SaaS space and helped us go to market faster. We are very excited about the results we have achieved so far and look forward to providing more specific details on this in the future. Operator: Okay, thank you. Our last question comes from Brad Zelnick of Deutsche Bank. Please go ahead. Brad Zelnick, Analyst: Great. Thank you very much. I also want to express my congratulations, and I want to say that the message you conveyed is very clear and valuable. My question is directed to Mike, and perhaps Larry can add something as well This can be seen as an extension of the question just posed by Raimo. You have launched the "AI Agent Studio" internally in Fusion. We all know that the most valuable core assets (Crown jewels) of a company are stored in Oracle databases and Oracle applications. But I am curious, in a world where many other vendors are competing to become the "AI interaction layer" across multiple different enterprise systems and workflows, how do you see Oracle's role evolving? Mike Sicilia, CEO of Application Software: Brad, I'm Mike. I'll answer first. I believe data gravity plays a decisive role here. Especially the data gravity of mission-critical data is even more important. As we mentioned, we have announced the integration of the AI Agent Studio within Fusion. Fusion is a system located within enterprise customers, serving as the custodian of their operational data and mission-critical data. If you are going to build a bunch of AI agents, or if your system integrator is going to build a bunch of AI agents, the question I often ask is: where will you start? You will definitely start from within the System of Record, from systems that have data gravity. Because from a reasoning perspective, and from the perspective of Retrieval-Augmented Generation (RAG), this data will be highly relevant and specific, providing a wealth of context for AI. The AI Agent Studio we launched in Fusion is not limited to Fusion data; you can also build AI agents in our industry applications and third-party applications. Third parties can also build AI agents within it. We provide an integrated, best-in-class solution: a full-scale SaaS application, an AI-driven SaaS application, and empower you with the ability to create your own AI agents on top of or alongside it within the quarterly platform release schedule for standard upgrades. I think this will be very attractive because the AI Agent Studio we built in Fusion is part of our quarterly upgrades and regular security patches. Therefore, you will get a win-win experience. You get a packaged SaaS application; you get an agent studio closely tied to the enterprise's most critical and core data; and if you wish, you also gain the ability to create custom, proprietary agents. Larry Ellison, Chairman and CTO: Let me add to that. We provide a large number of pre-built agents for all applications. But beyond that, we also offer a development environment—the AI Data Platform—that enables our customers to easily add their own agents on top of existing content. We do not believe we can build all the application agents for banking systems or healthcare systems. Many of our partners will develop, and many of our customers will also develop. The role of the AI Data Platform is to provide a fully integrated development environment where you can use any AI model in Oracle Cloud to build your own agents, essentially aggregating all popular AI models You can use it to write agent code, allowing it to perform complex multi-step query reasoning. For example, we plan to introduce a complex agent that handles the "Close" work in our Fusion accounting system. When you use Fusion to close in the near future, it will be an autonomous agent that requires no human involvement. You just need to tell the AI agent to close, and then you will get the result. We have built a lot of AI capabilities into the application, but they are also open. Their openness allows our customers and partners to continuously enrich this agent library. What we are building is an entire ecosystem: automated healthcare, automated financial services, automated retail. This is the capability that AI gives us; it expands our vision and broadens the scope of our SaaS software suite, enabling us to automate the entire ecosystem. Let me take the healthcare industry as an example. In the healthcare field, Epic has automated hospitals (especially emergency care hospitals, and in some cases, clinics), but mainly emergency care hospitals. We not only automate emergency care hospitals, but we also automate clinics and laboratories. More importantly, we automate payers, which are insurance companies. We automate HCM systems that are responsible for training nurses and scheduling the appropriate radiologists when an MRI needs to be performed. We automate the financial systems of hospitals. We even automate the processes that the FDA uses to approve new drugs and interface with pharmaceutical companies. This is the vast healthcare ecosystem. Thank God we now have these programming tools that allow us to build a comprehensive, agent-based software suite to automate complete ecosystems like healthcare or financial services. This is what Oracle is doing. This is why we consider ourselves disruptors. This is why we believe that the so-called "SaaS doomsday theory" applies to other companies but absolutely not to us. Our products are truly outstanding. Brad Zelnick, Analyst: Thank you, Larry. Thank you, Mike. Congratulations to you both. Larry Ellison: Thank you, Brad. Ken Bond, Head of Investor Relations: The recording of this conference call will be available on our investor relations website for 24 hours. Thank you all for participating today. Now, I will hand the call back to Regina to conclude. Operator: This concludes today's meeting. Thank you all for participating, and you may now disconnect