The Zhitong Finance App learned that cloud computing giant Oracle (ORCL.US) announced the results for the first quarter of the 2027 fiscal year after the US stock market on Thursday. According to financial reports, Oracle's revenue for the first quarter increased 30% year over year to US$19.35 billion, better than analysts' average expectations of US$19.13 billion. The company attributed revenue growth to trends such as strong cloud infrastructure business performance and increased data center capacity. Operating profit was $6.73 billion, up 57% year over year; net profit attributable to common shareholders was $4.68 billion, up 60% year over year; adjusted earnings per share was $1.92, better than analysts' average forecast of $1.75.
By business segment, cloud business revenue increased 62% year over year to US$11.61 billion, a record high, better than analysts' average forecast of US$11.52 billion. Among them, cloud infrastructure business revenue increased 121% year over year to 7.4 billion US dollars, better than analysts' average forecast of 7.19 billion US dollars; cloud application revenue increased 10% year over year to 4.2 billion US dollars. Some analysts previously predicted that as the market demand for AI computing power continues to rise, the increasingly competitive pricing environment will benefit the Oracle Cloud business.
At the same time, software business revenue fell 3% year over year to US$5.55 billion, falling short of analysts' average expectations of US$5.67 billion; hardware business revenue increased 15% year over year to US$770 million, better than analysts' average expectations of US$670 million; service business revenue increased 5% year over year to US$1.41 billion, better than analysts' average expectations of US$1.38 billion. Regarding the decline in software business revenue, Oracle said that this reflects the continuous migration of customers from locally deployed products to cloud products.
Customer demand for AI cloud training and inference services continues to grow, outpacing supply. Oracle added more than $30 billion in AI cloud contracts in the first quarter, increasing its remaining performance obligations (RPO) by 4% month-on-month to $664 billion, higher than analysts' average expectations of $618 billion. Based on the structure of these new contracts, the company confirmed that there would be no additional impact on its financing plans. Since the end of the fourth quarter of fiscal year 2026, the company has also delivered more than 300,000 GPUs to its AI cloud customers, almost three times the delivery capacity of the fourth quarter of fiscal year 2026.
Operating cash flow for the first quarter reached US$23.1 billion, up 184% year on year; free cash flow for the first quarter was negative US$5.4 billion, and Oracle has yet to give a specific timeline for recovering positive free cash flow. Capital expenditure for the three months ended August 31 was $28.5 billion, most of which was for data center equipment. This figure is far higher than analysts' estimate of $19.23 billion. The company's executives confirmed that the company expects capital expenditure for the fiscal year ending May 2027 to reach $70 billion. Additionally, since some parts require advance payments, an additional $20 billion to $25 billion will also be spent.
Looking ahead, for the second quarter, Oracle expects revenue to increase 30% to 34% year over year, slightly better than analysts' average forecast of 31.9%; the company also expects adjusted earnings per share of $1.85 to $1.93, which is in line with analysts' average forecast of $1.89. For fiscal year 2027, Oracle expects revenue to “reach at least $90 billion,” up from $90 billion; the company raised its adjusted earnings per share forecast to $8.10 billion from $8.05 previously.
Here are the details of Oracle's first quarter earnings call.
I. Management statement
Hilary Maxson — Chief Financial Officer
If I were to describe this quarter in one word, I think the term would be “acceleration,” because we are seeing an acceleration in the execution of the entire company, which translates into results on the revenue and profit side. The first quarter was another record quarter, mainly due to strong performance in our cloud infrastructure and cloud application business.
Total revenue for the first quarter reached a record high of US$19.3 billion, up 30% year over year in US dollars. Furthermore, this is the first time that total revenue increased sequentially in the first quarter, which is an important sign that we are continuing to make progress in large-scale infrastructure construction. Historically, a record fourth quarter is usually followed by a weaker first quarter. But as our entire technology stack progresses at an accelerated pace, from infrastructure to databases to software, this is no longer the case.
Cloud infrastructure revenue for the first quarter was $7.4 billion, up 121% year over year, reflecting our strong execution capabilities. Our new megawatt capacity online has reached a record level, while the demand environment for computing resources and database services remains strong. Cloud application revenue increased 10%, with Fusion and our industry application growth rate significantly higher than this level. Mike and Clay will go into more detail about these businesses later.
Our non-GAAP operating profit increased 31% in US dollars to $8.2 billion, mainly due to strong revenue growth and operating leverage. As expected, our gross margin declined due to the impact of data center commissioning and accelerated infrastructure revenue growth. However, lower operating costs this quarter, as well as strong operating leverage associated with simplification and efficiency measures, offset this impact.
Overall, the non-GAAP operating margin remained flat at around 42% for the quarter. These factors ultimately drove our strong 30% increase in non-GAAP earnings per share in dollars to $1.92 this quarter.
The final financial highlight I want to talk about is that our remaining performance obligation, or RPO, increased by $26 billion from the fourth quarter. There are two things going on here.
First, we continued to increase our RPO this quarter to support future revenue. The vast majority of these additional contracts use prepayment, client-owned hardware, or similar mechanisms, and therefore do not require additional capital from Oracle. Additionally, these additional RPOs won't impact our capital expenditure or revenue until fiscal year 2028 or later.
Second, we began to see a strong shift from RPO to revenue this quarter, which drove cloud infrastructure performance. We've added a few slides here, and you can see the strong inflection point of RPO turning into revenue and operating profit.
The first is cloud infrastructure revenue. As I mentioned earlier, the quarter's growth was 121%, compared to the fourth quarter's growth of 93%. As we turn more of our RPO into revenue, we expect this acceleration trend to continue throughout the rest of the 2027 fiscal year. We currently expect approximately half of our RPO to be converted into sales within the next 36 months.
Second, unsurprisingly, you can see that the company's total revenue is growing at an accelerated pace, which is shown here based on the past 12 months. As the cloud infrastructure business accelerated and became an increasingly important contributor, the growth rate in the first quarter was 5 percentage points higher than in the fourth quarter. Finally, our operating profit growth, also based on the past 12 months, showed a similar strong acceleration, rising from 16% in the fourth quarter to 21% in the first quarter.
Now let's talk about our balance sheet and cash flow. Our operating cash flow in the first quarter reached a record of $23 billion, which once again reflects our ability to execute in an environment of strong demand and the collection of advance payments from customers. Capital expenditure for the quarter was $28 billion, resulting in negative free cash flow of $5 billion. Our net cash capital expenditure, after deducting advance payments, was $18 billion for the quarter.
It should be noted that our capital expenditure will not be distributed linearly over the year. We continue to forecast full year capital expenditure of $90 billion to $95 billion, of which net cash capital expenditure is no more than $70 billion. Finally, we are pleased to announce that we have fully completed our previously disclosed $20 billion share issuance plan at market price in the first quarter.
Next, I handed over the phone to Mike, then Clay, and asked them to further explain our cloud applications and infrastructure business.
Mike Sicilia — Co-CEO and Director
Thanks, Hilary. Let me first tell you more about the app business.
We continue to see the strong value of application suites in the minds of our customers. They're investing in complete, trusted solutions that can now seamlessly combine intelligence and applications to run their business.
The introduction of AI is an accelerator rather than a replacement for packaged apps. Therefore, our decades of experience and expertise in running business processes across industries, regions, and organizations of all sizes allow us to understand how we can help our customers succeed.
Prior to the advent of AI, the suite of applications had proven its effectiveness. Companies are able to increase profit margins because standardized, efficient, and end-to-end automated business processes have proven to be far more effective than one-off customized solutions. But this requires organizations to follow workflows and processes the way the system is designed, and many businesses struggle to do this consistently across functions, teams, and regions.
AI is changing this dynamic. Instead of requiring every employee to accurately handle and execute processes in the manner required by the system, AI agents can use the organization's established workflows and business rules to perform tasks. Employees then turned to monitoring intelligence, resolving anomalies, and applying human judgment where it was needed most.
By combining applied AI with complex business rules, regulatory compliance systems, security models, data models, and customer configurations accumulated over decades, we enable customers to continuously realize the value of AI while ensuring data security and maintaining existing operational barriers. This makes it easier than ever for organizations to harness the power of our suite of apps. We are confident in the potential of this new paradigm to provide customers with a faster return on investment.
At the AI World conference in October, we will launch a new intelligent AI accelerator, which is expected to redefine the way customers deploy Oracle applications, making deployment faster, simpler, and significantly lower costs. AI agents will work collaboratively with Oracle and customer teams to automate and orchestrate implementation processes on an unprecedented scale, reducing the SaaS deployment cycle from years to months and then from months to weeks. What really strengthens my confidence in the growth prospects of the app business is the combined power of these capabilities.
Many of our SaaS products performed strongly in the first quarter, thanks to the demand environment I just described. Overall, our SaaS business grew 10%, with Fusion growing 14%. Our Oracle Health business continues to accelerate. Although we don't separately list this data, our industry applications grew by more than 20% in the first quarter.
As I mentioned last quarter, NetSuite's decision-making cycle slowed down in the previous fiscal year, so its growth rate was slightly lower than other businesses, but we now have an exciting new product fully launched, which I'll cover later.
Now let's take a look at some of the customer side of the quarter, which is just a small part of the longer customer list. Uber, Stanford University, and Japan's Mitsubishi UFJ Bank have all launched Fusion and/or accelerated the use of Fusion. Pye-Barker Fire & Safety chose Oracle's complete suite of applications, from industry applications to Fusion, including Fusion Smart Asana applications. Johnson Controls, Saudi National Bank, GuideWell Mutual Holding Corporation (a health solutions company serving more than 45 million people), and Malaysia's national energy company Petronas all added Fusion smart applications this quarter to drive better business results.
Let me share some data on embedded AI usage and progress this quarter. Customers used our embedded AI features more than 150 million times this quarter, an increase of 42% over the previous quarter. Our AI agents executed more than 3.5 million times in production environments this quarter, almost doubling from month to month. Customers currently have more than 2,300 AI agents in production environments, an increase of 90% over the previous month. Overall, Fusion's AI production environment usage this quarter alone consumed 900 billion tokens. I think it's fair to say that customers are using our AI built into Fusion apps and across the entire application technology stack every day.
Now let's talk about NetSuite. We are announcing the official launch of our new AI-powered product, NetSuite Next. This product provides a simpler, more powerful intelligent asana experience, and incorporates AI into the workflows that customers rely on every day. It's easier to adopt, increases productivity from day one, and creates more value as customers grow.
Additionally, the NetSuite AI Connector Service enables customers to securely connect NetSuite data to leading AI assistants of their choice, including ChatGPT and Claude. The service has become one of the fastest adopted features in NetSuite's entire history, and is currently being used by over 10,000 customers. Personal care company Every Man Jack estimates that this service alone could save $350,000 and nearly 5,000 hours of work each year.
This month, at our Oracle Health and Life Sciences customer event, we're launching a new intelligent asana care management system with a lineup of world-class external speakers. It's more than just an EHR. The system links clinical research with medical services, reduces the burden of managing patients' medical records, and enables medical service providers to maximize their professional qualifications by using AI as a user interface. Few AI's mission is more important than this—helping deliver better care while enabling healthcare providers to focus on what matters most: patients, not computer systems.
These verification points surrounding Oracle's AI products are clear, and we are still confident about future opportunities. Let's hand over the phone to Clay.
Clay Magouyrk — Co-CEO and Director
OK, thanks, Mike.
The Cloud Infrastructure Business (OCI) continues to grow rapidly as we continue to deliver the capacity our customers require. Since the end of the fourth quarter, we've delivered 850 megawatts of AI capacity to customers, including over 300,000 GPUs. The delivery volume in the first quarter was almost three times the total delivery volume in the fourth quarter, which is equivalent to 73% of our total delivery capacity in the previous fiscal year. This reflects our investments over the years in every aspect of our infrastructure, from data center design, to supply chain and manufacturing, to installation and operation. Customer demand continues to support these investments.
In the first quarter, we signed more than $30 billion in new AI contracts without additional capital from Oracle. Our ability to operate large multi-tenant GPU clusters remains an important advantage. GPU utilization remained extremely high in the first quarter, reaching 97.9%.
The longevity and value of GPUs continues to be impressive. All GPUs that entered the renewal period in the first quarter had their capacity renewed or resold at a 20% premium over previous contracts. Most of these GPUs have been around for 4 years or more. We have seen that the AI capacity we have deployed has a long lifespan, and its value continues to increase.
The Abilene, Texas data center project continues to advance at an extraordinary pace. In the first quarter, we delivered 131,000 GPUs there, 1.9 times the volume delivered in the fourth quarter. Six of the eight park buildings have been delivered to customers, representing 618 megawatts of capacity, accounting for 75% of the total capacity. The customer acceptance time has been reduced to just 24 hours, which indicates that the system arrived ready to carry the customer workload. The recently released GPT-6 Astra is being trained at our Abilene campus.
Shackelford County, Texas is our next gigawatt campus and is progressing well. The Nvidia Vera Rubin system performed better than expected in terms of hardware quality, manufacturing yield, and performance. We will be delivering the first Vera Rubin systems to customers in the second quarter.
Our database cloud business is also growing rapidly. Multi-cloud database revenue increased 353% year over year, and the number of multi-cloud customers increased 180% year over year. We've completed the planned Azure and AWS regional coverage expansion to 70 multi-cloud database regions and 119 availability zones. This provides customers with a consistent way to deploy the Oracle AI Database with their applications and data in the cloud environment of their choice.
We also officially launched Oracle Interconnect for AWS. With the launch of this product, OCI has now established private high-speed connections to all hyperscale cloud service providers, and there are no data transmission charges. This makes it easier for customers to build distributed applications and migrate workloads between clouds. The pace of AI innovation across the entire ecosystem is accelerating.
We've expanded our partnership with OpenAI to provide OpenAI API access, ChatGPT for Work, and Codex, including GPT-6 Astra, through Oracle Marketplace. We are introducing the Gemini model to Oracle enterprise applications and are releasing new Grok inference models, multi-modal models, and text-to-speech models. We're also continuing to expand our catalog of open source models, including new models from Nvidia, Qwen, Google, DeepSeek, and others.
We are announcing a multi-year partnership with Quantinium to provide its Helios quantum computer through OCI. Helios will operate at OCI AI data centers in the US, providing hybrid quantum computing and AI workload capabilities for enterprise, research, and AI customers.
Oracle APEX 26.1 integrates these infrastructure, database, and model capabilities to provide services for application developers. APEX currently runs more than 2 million active apps, and thousands of new apps are added every day.
ApexLang is a new technology that can represent APEX applications in a structured, readable form of human language. These application definitions can be stored in a source control system and verified and governed. AI programming agents can generate and modify these definitions, while the APEX engine continues to provide the security, reliability, and operational control required for enterprise applications. As a result, developers can gain the speed of generative development while avoiding the drawbacks of application code that is difficult to maintain and lacks transparency.
We're taking the same approach with Oracle AI Data Platform. The AI Data Platform is now integrated with Codex and Claude Code, enabling developers to directly use the AI Data Platform's data, knowledge, and capabilities in their preferred programming environment. We're also adding advanced MLOps capabilities and working with early release clients to develop business knowledge models, semantic reasoning engines, and deep insight agents.
These capabilities enable AI to be built on the true meaning of an enterprise, with context, semantics, and governance capabilities. They also bring AI-driven analytical capabilities to Fusion Data Intelligence and Oracle Analytics Cloud, where many customers already manage their most important business data.
Taken together, the first quarter showed how the components reinforce each other. We're delivering data centers and GPU capacity at a rate that seemed nearly impossible a year ago. Customers are signing new contracts to renew capacity at a higher price and keep the entire GPU cluster running at close to full capacity. We're deploying the Oracle AI Database to every major cloud platform and providing more proprietary and open source models on OCI. We then connected these models to enterprise data, apps, and developer workflows.
This combination is why demand continues to grow, and why we remain confident in the long-term value of the technology we are building. Now return the phone to Hilary.
Hilary Maxson — Chief Financial Officer
Thanks, Clay.
Finally, let me introduce the results guidelines for the second quarter and full year of fiscal year 2027. In the second quarter, we expect total revenue growth of 30% to 34% in US dollars. Among them, we expect cloud revenue in dollars to grow 65% to 71%. Non-GAAP earnings per share are expected to be $1.85 to $1.93, an increase of 21% to 25% in dollar terms. This guidance does not include Ampere earnings confirmed in the second quarter of last year.
For the full year, given the strong execution and acceleration shown in the first quarter, we raised our performance guidance to total revenue of at least US$90 billion, an increase of 34% year over year, while also raising the non-GAAP earnings per share guideline to $8.10.
Finally, I have a small reminder. Please keep in mind the schedule. Our Investor Day will be held in October.
Let's return the phone to Ken and move on to the question and answer session.
II. Q & A session
1. Brad Zelnick - Deutsche Bank Analyst
I fully expect Oracle to continue to be one of the few market leaders in AI infrastructure. You told us that the 2027 and 2028 fiscal years will be peak capital spending years. But at the same time, other companies in the market are investing hundreds of billions of dollars to build production capacity, and there seems to be no end. If other companies aren't slowing down, how should we understand the possibility that Oracle may slow down spending after the next two years? Is there a scenario where we see a higher peak in capital expenditure after FY2028? What factors will guide your investment decisions? Also, given the current backlog of orders, when should we expect the company to return to positive free cash flow?
Clay Magouyrk — Co-CEO and Director
Thanks, Brad. I'm Clay. I think you're actually asking about four questions here. So, I'll answer the part I wanted to answer and then let Hilary handle the harder part.
You've been hearing us talk for the past few quarters, and we're constantly looking for fun ways to fund our business. Of course, one of the mechanisms we use to finance our business is to invest using our own capital. However, we have invested heavily in building relationships with different suppliers and vendors, and created new business models, including customer-owned hardware models, all of which can allocate capital investment in different ways.
Therefore, I think we need to think about Oracle as an investment in capital expenditure and separate these two things from how we look at how the business grows. Because from our perspective, I think we've seen some ways — obviously capital is still needed to do this work, but that capital isn't necessarily all from Oracle. It doesn't necessarily have to be Oracle's capital expenditure. So, I don't think this is a limiting factor for our business growth. I think this is just a reflection of the continuous evolution of the business model we are developing as AI continues to expand.
Hilary, would you like to talk about something else?
Hilary Maxson — Chief Financial Officer
Of course. So, I think the other question you raised is free cash flow. We haven't given a specific schedule yet, and we don't plan to do so today. What I want to say, however, is that the projects we are working on are essentially projects that generate strong free cash flow. As a result, once these projects start climbing, they will soon be able to achieve a free cash flow conversion rate equivalent to approximately 100% of EBITDA after tax.
In fact, from a certain stage, the business itself will “self-finance” to a certain extent because it can generate large amounts of free cash flow. So, we haven't given a specific timeline yet, but as we continue to climb production capacity over the next few quarters, you'll see what this situation might look like.
Once again, we expect this to happen fairly quickly after the large-scale climb that is currently underway. The only deciding factor here is the increased growth capital expenditure you have indicated, and this is the part where we want to continue investing at the right level to drive business growth.
2. Siti Panigrahi - Mizuho Securities Analyst
There has been a lot of speculation recently about potential data center delays in New Mexico and Wisconsin. But you've met your first-quarter goals, and have now even raised your full-year revenue guidance for FY2027 to at least $90 billion. Can you tell us how far these two projects have actually progressed? Are there any risks to the delivery schedule for these data centers that could risk your fiscal year 2027 revenue expectations? As you continue to increase your RPO, how confident are you in your ability to secure and launch sufficient capacity in the future to support this growth?
Clay Magouyrk — Co-CEO and Director
OK, thanks, Siti. OK, I'll answer the specific parts, but I think it's important to provide some context before continuing.
Both New Mexico and Wisconsin are big projects that are very important to us. But I think it's important to understand the context. When we talk about these projects, we usually say they are about 1 gigawatt each. We just delivered 850 megawatts in the first quarter.
This means — as you can see — none of the Shackelford, New Mexico, Wisconsin, and Michigan projects were delivered in the first quarter. As a result, we have numerous, diverse, and widely distributed data center development projects across the US and around the world, and are delivering capacity to our customers.
Now, projects like New Mexico and Wisconsin are clearly getting a lot of attention. There's also a lot of discussion about them. But I think people need to be aware that we don't put all our eggs in one basket. Next, I want to make sure everyone understands that when these big projects are completed, they won't go online at the same time.
Let's say you have a 1 gigawatt data center project. It is scheduled to begin delivery in January of a certain year. That doesn't mean you'll get 1 gigawatt of capacity by January. Capacity will be rolled out in stages over several quarters. Therefore, if a project is delayed relative to the plan, it doesn't mean that a single quarter will experience a huge impact.
Also, I'd like to say that anyone who has been involved in construction or major infrastructure development knows that if your plan depends on every deliverable being completed 100% on time, we have one word to describe this kind of plan. That's a “bad plan.” So we try very hard to avoid making bad plans.
We clearly know how difficult and complicated what we're doing is, so we don't assume anything will always be 100% successful. We have prepared alternatives for these situations, and we don't assume that everything will always be done exactly according to the schedule.
Now let's answer questions specifically about New Mexico and Wisconsin. New Mexico is an interesting place. We are progressing very well. As far as construction is concerned, the data center project is definitely on the right track. We are proceeding with the process of obtaining air emission permits. The technology we plan to deploy there is Bloom fuel cells. This is the most environmentally friendly on-site power generation method we can use so far, with extremely low water consumption and extremely low emissions compared to almost any other on-site power generation method.
Therefore, we are very confident that as this process continues, we will work with local regulators and community residents in Doña Ana County, as well as other relevant parties in New Mexico. However, we are currently only in the normal approval process. For a major project like this, I don't think it's unusual.
In Wisconsin, we don't generate electricity on-site. We are actually working with various partners to design and deliver the required energy capacity through the grid. But again, we are moving forward with the relevant process. These are complex projects. Once again, in Wisconsin, the data center delivery was actually very much in line with the plan and progressed smoothly.
We're working with the Public Service Commission (Public Service Commission), ATC, and We Energies. We are constantly adjusting different aspects of our energy design and delivery plans, but we are very confident in both projects.
As for the impact of these two projects on FY2027 revenue, neither of these projects will have any impact on our previously announced FY2027 revenue or profit guidelines. So what I want to say is that some projects are ahead of schedule, while others are more difficult than others. We handle all projects one by one.
We have developed relevant plans, including arrangements for risk management. At the same time, we also take these factors into account when making predictions and formulating performance expectations for ourselves and investors.
And then the last question is, OK, what do we think about bringing more capacity online? I think we're really excited about it. Clearly, the environment is changing all the time. The factors that limited us in the past were GPUs and fabs, then the limiting factor turned to power generation capacity. Now there are data center restrictions again. But the world is big. There is significant demand for this capacity. We're driving these capacities online through a variety of different channels. As a result, we are still very excited and very confident that we will continue to meet our current RPO and the RPO growth we expect to see in the future.
3. Raimo Lenschow - Barclays Analyst
My question is also about data centers. Obviously, there is a lot of discussion on the market now about issues such as rising parts prices. Clay, can you talk about the price change between contract pricing and spot pricing? Last year you gave a gross margin framework. Does this framework still hold true today? In the current environment where parts prices continue to change dynamically, how should we understand this?
Clay Magouyrk — Co-CEO and Director
Of course. In a world where demand exceeds supply, prices generally don't fall. Prices will go up. So, I think the end effect is obviously that things get more expensive, but then we also have to charge a higher price for these things to be compensated accordingly. We're doing this in all of these different businesses. We don't expect this to have an impact on our gross margin. So, I think the guidelines we gave earlier are still valid.
What I'm saying is that, as I mentioned in my prepared statement, I think many people have always been very concerned about how long the different hardware assets will last, and how long the demand for data centers and the power capacity associated with them will last. I've basically spent my entire career in the infrastructure industry. I've been doing this at Oracle for 12 years. Until now, one constant has been true: demand for server-side computing and data centers will only increase.
As it turns out, the AI usage scenarios we've seen so far are the same, except that the process is more advanced. When we launched capacity that entered the renewal period, we were actually able to achieve a higher price of around 20%, which is a very positive sign. This shows that the continued demand, growth, and profitability of this business are very strong.
4. Mark Moerdler — Bernstein Analyst
I want to focus on RPO. Can you further explain the drivers of RPO growth that do not require additional capital expenditure, namely models such as prepayment and customer-owned chip? Are these customers AI labs? Is it a semiconductor company? Or is it a sovereign customer? In fact, we haven't discussed sovereign clouds in a while, and we haven't even discussed sovereign AI at all. Can you explain how this aspect of OCI is currently progressing and how it will impact capital expenditure and profit margins?
Clay Magouyrk — Co-CEO and Director
Of course. First, I'd like to clarify one thing before going deeper into each section. I didn't say — I don't think Hilary or I said — these projects don't require additional capital expenses. What we're saying is that they don't require additional cash from Oracle. I'm not an accountant, but sometimes I can play as an accountant on TV.
However, our intention is that while there is clearly capital expenditure here, this doesn't require Oracle to go out and find additional cash to complete these investments. So the question becomes, how did you guys do this?
We have many different models. Sometimes we work with suppliers through different financing arrangements that enable us to pay for capacity while our customers pay us. It's a mechanism.
Another mechanism is that the customer might say, “Hello, I want to pay for the hardware, but I want to use your operational capabilities and cloud infrastructure and technical assets in the data center to actually transform this hardware into an AI cluster.” This is another option.
The third option is that the customer is already able to raise capital. It could be a startup or a mature company. They'll say, “Hello, I'd like to pay you guys in advance.” In return, this does not require you to first take out cash and use your own funds to pay for these capital expenses.
So we have different models to achieve this goal. As for the types of customers and where we see this kind of demand, it's actually pretty broad. This is the case whether the client is a startup or the most valuable investment-grade company. The industry has now recognized that access to capital and different financing methods are a constraint, and the entire industry is making adjustments to allocate this capital in the most efficient way possible.
Now let's specifically answer your question about sovereign clouds. Our Alloy business is doing well. We have many partners in Japan. We also have great partners in the Middle East. Our partners range from expanding commercial business in a more sovereign way to sovereign cloud businesses that are more focused on the government sector. This business is still expanding widely. We're seeing a lot of demand. At the same time, this is actually related to AI, because we provide GPU capabilities for many of these customers, and they will deploy these GPUs in sovereign workloads.
Obviously, we also have a very large, very fast growing general cloud business, and we don't talk about it that often. However, this business is growing rapidly, and the growth rate is very high, and the profit margin is also very good. It does require a certain amount of capital, but it doesn't require as much capital as these large AI clusters.
5. John DiFucci — Analyst at Guggenheim Securities
Hilary, gross margin declined markedly this quarter, but you have brought so much production capacity online this quarter. My understanding is that prior to the launch of these production capacities, related transactions had already generated significant costs, but there was not much revenue at the time. Despite this, gross profit margins have declined significantly. When we talk to investors, this is a topic that people often talk about during this conference call. However, as you pointed out, the operating profit margin was basically flat, with a slight increase over the previous year.
I know that operating margin is a “North Star (key)” indicator because it is closest to final profit and final cash flow. But in addition to operating profit margins, how should we view gross margins and the entire ecosystem in the future?
Hilary Maxson — Chief Financial Officer
For me, gross margin is clearly an important indicator that requires attention both internally and externally. Especially in reality, in the two or all three business models we have, gross margin is an indicator for measuring the health of the business. Are you capable of pricing? Is your pricing at the right level? Are you able to manage the investment costs? Therefore, gross margin has always been an important indicator for me.
Internally, it's probably even more important because we need to keep checking. I think investors clearly want such checks too. For example, if our pricing model isn't right, gross margins can change very quickly. However, when we talk about how to drive value and how to create value for the business, for me, operating margin is probably the metric we ultimately want to focus on.
So, as you mentioned, there are many factors at play in current gross margin. We have both data center climbing and adjustments between the two business models in our business.
The gross margin of the software business is much higher, but it bears higher R&D and sales costs below the gross margin. The gross margin of the infrastructure business is low. We've talked about this before, and we've also given relevant figures. Clay has given a fair share of expectations for this business. Of course, databases aren't included here, but more AI infrastructure and cloud businesses are included. However, judging from the nature of the business, the R&D and sales costs associated with this business are much lower. At least for a company like Oracle, we can effectively utilize the R&D work that has already been completed and is ongoing throughout the company's other businesses.
So, in the long run, how should we observe the creation of business value? I think operating margin is really a key metric we should focus on. Once again, we must not ignore gross profit margin; it is also an indicator that needs attention. But operating margin is a very, very important value indicator.
John DiFucci
That's pretty clear. This is something we'll focus on. But I'm wondering how we should understand this because most people — I think Wall Street usually overestimates your future gross profit margin while underestimating its impact on operating margins. So I'm not sure if you guys are ready to say whether gross margin should continue to decline in the future? Or should it stabilize around current levels for the rest of the year? How should we understand it?
Hilary Maxson — Chief Financial Officer
Yes. As we've mentioned, and I mentioned it in the fourth quarter, we expect a step-down in gross margin this year. However, you can see our earnings per share guide, so you can also understand our operating margin expectations based on this.
Over the next few years, as we complete the climb, you can reasonably expect gross margins to stabilize. But we haven't given specific guidance today. However, we'll be discussing some of these further during the upcoming Investor Day in October.
6. Brent Thill - Jefferies Analyst
Mike, I'm excited to see SaaS achieve double-digit growth. If so, could you share your thoughts on the next few years? I think the entire industry has always had a concern that the SaaS business may be impacted by AI. But that doesn't seem to be your opinion. Please share your thoughts on how to maintain sustainable double-digit growth.
Mike Sicilia — Co-CEO and Director
Of course. Thanks for the question. This is how I view our SaaS business. I'm going to expand the scope a bit because I think there are a few key neighboring factors that are important in the SaaS field.
First, we have highly differentiated products because we provide an end-to-end suite that automates complete industries, including healthcare, retail, telecommunications, construction, etc. There are also many such industries, and we have horizontal and vertical applications. If you look at our two very large categories this quarter, both Fusion and industry applications, saw very good growth.
The ability for customers to obtain these products in the form of complete kits and complete software packages has been a differentiating advantage of Oracle for many years. Now, the AI embedded in these workflows is superimposed, AI is used as part of the service, and AI capabilities are continuously provided as customers receive regular application updates. We think Brent is a great way for customers to get value and return on investment from AI or embedded AI very, very quickly, without having to install anything extra like add-ons.
The next level is our intelligent asana ability, the Fusion Agentic AI Studio. It allows customers and/or partners to build their own AI agents directly within the same platform. It's not a different platform. Nor is it a different control plane. It runs on the same control plane and platform as our app.
This means Agentic Studio allows customers or partners to build their own agents and receive the same quarterly updates, the same security patches, and deliver them as a complete service to customers. This means that customers can use AI as a user interface, superimposed on a very complex set of business rules, highly differentiated security models, and of course data models that have evolved over the years. So when you combine the horizontal package — that is, horizontal application, vertical application, and Agentic Studio — we think this is a very attractive combination.
But if that's not enough, the next part is the AI-assisted launch capability I just mentioned, that we're introducing. In many cases, we're taking on the difficult job of helping our customers solve highly critical, industry-specific issues. Sometimes, it can take a long time for these systems to go live. They are complex and also involve data migration. And AI has given us tremendous capabilities to speed up the launch of these systems. The tools we have launched so far already have some verification points.
At the AI World conference, we will also launch more related tools. For example, we have seen that some complex system launch projects in medical applications, which originally required a high double-digit number of months, can now be reduced to a single digit of several months. In NetSuite applications, we've seen early customers use these AI tools to reduce the system launch cycle from double-digit months to single-digit weeks, so it can be put into production and operation. Therefore, we think this will have two effects.
First, it helps — or more accurately, customers — to reap the value of AI faster than ever before, and at a lower cost. Second, in some very complex industries, the launch of these systems itself has a climbing period. As a result, this allows us to lift restrictions on online hill climbing and confirm revenue more quickly than with traditional manual implementation.
But I think another very important part is that, as we mentioned in the press release, it's still part of the same platform we run our apps on, the AI Data Platform. It is capable of automating the creation of the enterprise itself. While we are honored to serve our customers with a very broad and large suite of applications, our customers have other applications. It is possible to obtain these applications that customers are running, automatically create enterprise services, and then superimpose these AI intelligent body studios on them to combine all of these capabilities.
I just don't think there is any other company in the market right now that can provide all of these capabilities in the form of cloud services. Especially in a highly regulated industry, no other company can deliver on our scale.
Then I'd like to talk about another aspect of the SaaS business. Our SaaS business is also an excellent lead source business for the IaaS business. Our SaaS customers are also buying OCI. They also have other workloads, including non-Oracle workloads, and they're happy to take advantage of OCI. As SaaS customers, they are able to test OCI's performance and scalability very well because at this stage they are already OCI consumers themselves. Therefore, we look at the SaaS business as part of the overall solution portfolio we provide to our customers.
We continue to invest in this solution portfolio. We continue to make it easier to bring systems online and add more and more tools to enable customers to configure and utilize AI. For all of these reasons, we're very optimistic about the future of the app business.
7. Kirk Materne - Evercore ISI Analyst
You mentioned AI Data Platform in your press release. So I wanted to dive a little deeper into this. This seems like a very important way to help customers deploy agents for proprietary data. Can you help us understand the business model surrounding this business? Will it drive incremental consumption of Oracle Database and OCI? Or do you think it will become an independent software revenue opportunity? I'm just trying to understand how should we understand the impact of this business on financial results over time?
Mike Sicilia — Co-CEO and Director
Yes. I think you've actually touched the answer. The business model is all of these situations you mentioned. I mean, of course, we can adopt a model where we handle 100% of non-Oracle workloads. This is by no means an exclusive product for Oracle Database or Oracle applications.
The AI Data Platform is independent of a specific data source; it can extract ontology from any data source and automate it. In fact, we're currently doing this automation for hundreds of data sources. So, whether it's pure consumer AI Data Platform, or using it in combination with our apps, or OCI, we're more concerned about letting customers make the choices that work best for them, or let partners make the choices that work best for them. So I think it can actually help all of our businesses.
I'd also like to say that Oracle Database remains the world's most important — and probably the most valued custodian of mission-critical data. Of course, as we can automate the ontology and use Oracle Database as one of our primary data sources, at least for the mission-critical data in it, it will also help us further unlock our growing multi-cloud database business.
Customers clearly want to be able to easily migrate these on-premise Oracle Database workloads to the cloud. As we continue to invest in this capability across every cloud platform and every region, we believe AI Data Platform will also help drive this growth.
Kirk Materne
I know you have a huge ecosystem of services. But will you deploy field engineers around this business to help customers understand some of these opportunities and capabilities they may have discovered through databases that they could not have discovered before? Or is there some kind of market entry strategy here too?
Mike Sicilia — Co-CEO and Director
Yes, of course. In fact, we're already investing in deploying field engineers to our customers. This is the case with AI Data Platforms. The same goes for Fusion Agentic Studio, because we really think of them as a combined platform running on OCI's single control plane. So yes, there's no doubt that we think this service is necessary.
In fact, I think it will help customers get online as fast as possible. As I said, we're currently using “weeks” to measure some of these success stories, and in a highly regulated industry, this speed was probably just a year ago, completely unimaginable.