Ladies and gentlemen, thank you for standing by. My name is Abby, and I will be your conference operator today. At this time, I would like to welcome everyone to the Oracle Corporation Third Quarter Fiscal Year 2025 Earnings Conference Call. All lines have been placed on mute to prevent any background noise. After the speaker's remarks, there will be a question and answer session. If you would like to ask a question during that time, simply press the star key followed by the number one on your telephone keypad. If you would like to withdraw your question, press star one a second time. Thank you. I would now like to turn the conference over to Ken Bond, Head of Investor Relations. Mr. Bond, you may begin. Thank you, Abby. Good afternoon, everyone, and welcome to Oracle's third quarter fiscal year 2025 earnings conference call. A copy of the press release and financial tables, which includes a GAAP to non-GAAP reconciliation and other supplemental financial information, can be viewed and downloaded from our Investor Relations website. Additionally, a list of many customers who purchased Oracle Cloud services or went live on Oracle Cloud recently will be available from our Investor Relations website. On the call today are Chairman and Chief Technology Officer Larry Ellison and Chief Executive Officer Safra Catz. As a reminder, today's discussion will include forward-looking statements, including predictions, expectations, estimates, or other information that might be considered forward-looking. Throughout today's discussion, we will present some important factors relating to our business, which may potentially affect these forward-looking statements. These forward-looking statements are also subject to risks and uncertainties that may cause actual results to differ materially from the statements being made today. As a result, we caution you against placing undue reliance on these forward-looking statements, and we encourage you to review our most recent reports, including our 10-K and 10-Q, and any applicable amendments for a complete discussion of these risk factors and other risks that may affect our future results or the market price of our stock. Finally, we are not obligating ourselves to revise our results or these forward-looking statements in light of new information or future events. Before taking any questions, we will begin with a few prepared remarks. With that, I'd like to turn the call over to Safra. Thanks, Ken. Good afternoon, everyone. As you can see, this was our strongest booking quarter ever by a huge margin, as we added $48 billion to our backlog. Our RPO balance is now $130 billion, up from $97 billion last quarter and up from $80 billion last year. That is a growth of 63% year- over- year. This does not include any contracts with Project Stargate. The RPO figure is the leading indicator of demand for our cloud services, while our live data center count and power capacity is the leading indicator of the conversion of RPO to revenue. Speaking of data centers, we marked a milestone this quarter as we crossed into triple digits with our 101st cloud region coming online. It's just a matter of time before we have more cloud regions than all of our competitors combined, reflecting the strategic advantage of our Gen 2 architecture, which offers our customers the most flexibility. From a delivery standpoint, the growth of our power capacity under contract is even higher than the growth in the number of data centers. We expect that our available power capacity will double this calendar year and triple by the end of next fiscal year. As we bring more capacity online, our revenues will clearly accelerate. What we are seeing in the market is that we are the destination of choice for both AI training and inferencing. This is due to the fact that our Gen 2 cloud is faster and therefore cheaper than our competitors, and also due to our ultra-high-speed networking engineering that we started decades ago and that is now highly relevant for AI. Taken together, we have numerous structural engineering advantages that distinguish OCI from our competitors, and as Larry will discuss in more detail. Beyond that, because of the momentum OCI is enjoying, customers are looking at us for many more workloads. Now, shifting to Q3 results, I'll be discussing our financials using constant currency growth rates, as this is how we manage the business. Here it goes. Total cloud revenue at SaaS and IaaS was up 25% at $6.2 billion, with SaaS revenue of $3.6 billion in the quarter, up 10%, and IaaS revenue of $2.7 billion, up 51%, on top of 49%, which we reported last year. As a reminder, the exit from our advertising business last year had the effect of lowering total cloud revenue growth by 2% this quarter. Total cloud services and license support revenue for the quarter was $11 billion, up 12%, driven again by OCI, our strategic cloud applications, and cloud database services. Infrastructure subscription revenues, which includes license support, were $6.2 billion, up 18%. Record-level AI demand drove Oracle Cloud Infrastructure revenue up 51% in Q3, and that's 54% when you exclude our legacy hosting. Both a much higher growth rate than any of our hyperscaler competitors. Our infrastructure cloud services now have an annualized revenue of $10.6 billion. OCI consumption revenue was up 57%. Demand continues to dramatically outstrip supply. Now, we do expect that the component delays that have slowed cloud capacity expansion this year should ease in Q1 FY 2026, so pretty soon. Growth in the AI segment of our infrastructure business was extraordinary. GPU consumption revenue is now nearly 3.5x the size of last year's. Cloud database services, which were up 28%, now have annualized revenue of $2.3 billion. Autonomous Database consumption revenue was up 42%, on top of the 32% growth reported last year. Again, we have acceleration as we get bigger. As on-premise databases migrate to the cloud, either on OCI directly through public cloud, Cloud @C ustomer, or DRCC, or through our Database @ cloud services with Azure, Google, or AWS, we expect that cloud database revenues collectively will be the third driver of revenue growth alongside OCI and strategic SaaS. We are currently live in 18 cloud regions with Database @ cloud services with our partners and have another 40 planned with Azure, Google, and AWS. Finally, database subscription revenues, which includes license support, were up 6%. Application subscription revenues, which again include product support, were $4.8 billion and up 6%, too. Our strategic back office SaaS applications now have annualized revenue of $8.6 billion and were up 18%. Software license revenues were down 8% to $1.1 billion. All in, total revenues for the quarter were $14.1 billion, up 8% from last year. Now, shifting to gross profit and operating income, the gross profit dollars of cloud services and license support grew 10% in Q3. We continue to focus on operating expense discipline, which collectively continued to grow slower. Expense discipline, so expenses continue to grow slower than revenue, a trend that I expect will continue. The Q3 operating income grew 9%, and the operating margin was 44%, up slightly from last year. The non-GAAP tax rate for the quarter was 19.9%, which was higher than my 19% guidance and lowered EPS by $0.02. EPS currency headwind ended up at $0.04, more than I thought would be hurt by currency as currency continued to strengthen. The non-GAAP EPS was $1.47 in US dollars, up 4% in USD, up 7% in constant currency. The GAAP EPS was $1.02 in US dollars, up 20% in USD, up 25% in constant currency. At quarter end, we had $17.8 billion in cash and marketable securities. The short-term deferred revenue balance was $9 billion, up 3%. Operating cash flow for Q3 was $5.9 billion, slightly more than our $5.9 billion in CapEx as we front-loaded some purchases into the quarter. Given the demand that you see in our RPO growth and the additional demand we see in our pipeline, I expect fiscal year 2025 CapEx will be a little more than double what it was last year at around $16 billion. As always, we remain careful to pace and align our CapEx investments appropriately and in line with booking trades. On a trailing 12-month basis, operating cash flow was up 14% at $20.7 billion, and free cash flow was $5.8 billion. As I mentioned, remaining performance obligations, or RPO, is now $130 billion, up 63% in constant currency. It reflects the growing trend of customers wanting larger and longer contracts as they see firsthand how Oracle Cloud Services are benefiting their businesses. Further, our cloud RPO grew over 90% and now represents more than 80% of total RPO. Approximately 31% of that total RPO number is expected to be recognized as revenue over the next 12 months. We are and remain committed to returning value to our shareholders through technical innovations, acquisitions, repurchases, prudent use of debt, and a dividend. This quarter, we repurchased nearly a million shares for a total of $150 million. Over the last 10 years, we've reduced the shares outstanding by more than a third at an average price of $54 a share. In addition, we have paid out dividends of $4.4 billion over the last 12 months. The board of directors increased the quarterly dividend 25% from $0.40 to $0.50 per share today. Before I dive into Q4-specific guidance, I'd like to comment on the financial acceleration we expect to see in the coming years. We now have a clear line of sight to our future revenue growth. We remain very confident and committed to total cloud infrastructure revenue for fiscal year 2025 growing faster than the 50% reported last year, and it will be even faster for fiscal year 2026, likely a lot faster. Our confidence in meeting our $66 billion revenue target for FY 2026 is now stronger than ever and represents around a 15% growth rate. More importantly, I now expect that our fiscal year 2027 growth rate will be around 20%, which is even higher than I previously guided. Let me now turn to my guidance for Q4, which I'll review on a non-GAAP basis. Assuming exchange rates remain the same as they are now, currency should have a $0.01-$0.02 negative effect on EPS and a 1% negative effect on revenue. However, as usual, currency impact may be different, so focus in on constant currency. Total revenues are expected to grow from 9%-11% in constant currency and are expected to grow from 8%-10% in USD at today's exchange rate. Total cloud revenue is expected to grow from 24% - 28% in constant currency and is expected to grow from 25% - 27% in USD. Non-GAAP EPS is expected to grow from 0% - 2% and be between $1.62 and $1.66 in constant currency. Non-GAAP EPS is expected to grow between - 1% and + 1% and be between $1.61 and $1.65 in USD. I should mention that my Q4 EPS guidance is negatively impacted by $0.03 + due to losses recognized from an investment in another company. Lastly, my EPS guidance for Q4 assumes a base rate of 19%. However, as you saw in this quarter, one-time tax events could cause actual tax rates to vary and usually do. Finally, I'm sure this isn't lost on anyone, but we are reporting earnings just 10 days after the close of the quarter, and that's also because there was a weekend. Using Fusion, we continue to file our quarterly and annual financial statements faster than any other company in the S&P 500. With that, I'll turn it over to Larry for his comments. Thank you, Safra. As Safra pointed out, some of our existing businesses, AI training and multicloud database, are experiencing hypergrowth. We are in the process of building a gigantic 64,000 GPU liquid-cooled NVIDIA GB200 cluster for AI training. Our multicloud business at Amazon, Google, and Microsoft grew 200% in the last three months alone. In addition to these rapidly growing existing businesses, new customers and new businesses are migrating to the Oracle Cloud at an unprecedented rate. In Q3, we signed a multi-billion dollar contract with AMD to build a cluster of 30,000 of their latest MI355X GPUs. All four of the leading cloud security companies, CrowdStrike, Cybereason, Newfold Digital, and Palo Alto, they all decided to move to the Oracle Cloud. Perhaps most importantly, Oracle has developed a new product called the AI Data Platform that enables our huge installed base of database customers to use the latest AI models from OpenAI, xAI, and Meta to analyze all of the data they have stored in their millions of existing Oracle databases. By using Oracle version 23ai's vector capabilities, customers can automatically put all of their existing data into the vector format that is understood by AI models. This allows those AI models to learn, understand, and analyze every aspect of your company or government agency, instantly unlocking the value in your data while keeping your data private and secure. This AI inferencing will be another great large new business for Oracle. Back to you, Ken. Thank you, Larry. Abby, please poll the audience for questions. And a reminder, one question per analyst, please. Thank you. Thank you. We will now begin the question and answer session. If you have dialed in and would like to ask a question, please press star one on your telephone keypad to raise your hand and join the queue. If you would like to withdraw your question, press star one a second time. If you are called upon to ask your question and are listening via speakerphone on your device, please pick up your handset and ensure that your phone is not on mute when asking your question. Your first question comes from the line of Brad Zelnick with Deutsche Bank. Your line is open. Great. Thanks so much for taking the question and congrats on just remarkable booking strength. Larry, I'm hoping you can expand more on Stargate because this is just a massive scale American-first venture with Oracle joined by undisputed AI leaders like OpenAI and NVIDIA. They chose you over several other choices in the market. What is Oracle's unique value add here? What can Oracle do that others can't? Thanks. I think it's actually very simple. The capability we have is to build these huge AI clusters with technology that actually runs faster and more economically than our competitors. It really is a technology advantage we have over them. If you run faster and you pay by the hour, you cost less. That technology advantage translates to an economic advantage, which allows us to win a lot of these huge deals. It is not just the Stargate deal, which is in our future, by the way. We got to over $130 billion in RPO without any transactions from Stargate. Stargate looks to be the biggest project, AI training project out there. We expect that will allow us to grow our RPO even higher in the coming quarters. We do expect our first large Stargate contract fairly soon. Thanks for that, Larry. If I could maybe just sneak in a very quick one for Safra. When Stargate does hit, as it is a related party, is there anything that you can share with us as to how we should expect it might flow through the financials? Thanks again. It won't flow through us in any unique way. They will place contracts with us, and they'll come right through. I'll be explaining it to you once it's fully laid out, but it's not going to make your work harder. We're going to be very, very clear on as the contracts come through us. It won't be as much of a change as you think. It's just going to be even larger numbers. Thank you again. Your next question comes from the line of Derek Wood with TD Cowen. Your line is open. Safra and Larry, congrats on a strong booking quarter. I wanted to drill into the growth undercurrents around OCI, especially in light of such a huge RPO number that did not even include Stargate. In some quarters, we hear about particular demand for AI contracts. Certainly seems to be a lot of favorable developments going on there. Other times, we hear about emerging adoption and multicloud and database on hyperscalers. We can hear about strength and dedicated as well with Sovereign Clouds and Alloy. I guess as you look at Q3 bookings and pipeline trends, can you give us a sense as to how demand is unfolding across those three different environments and how you feel about the growth durability and really the infrastructure capacity serviceability of each of these vectors? I mean, the reality is that everything is chugging on all fronts. Multicloud, I gave you some of the numbers. These numbers, I was just looking year over year because we only started really having revenue originally a year ago. It is more than 10 x what it was just a year ago. The numbers are exploding. As I told you, we have 18 live, but 40 coming online. That is going unbelievably well. OCI, public cloud going spectacularly, Cloud @C ustomer really going well. All the pieces around that. We are starting, there are whole parts that are only now rolling out, which are sovereign clouds, disconnected clouds. We have laid out quite a lot of capacity, and it is starting to fill up. Bookings are going very, very well, and it is turning into revenue. Pretty much we are going on all cylinders. For us, we're happy when customers come directly to us with their database workloads, but we're also happy when they come to us through our partners, Azure, AWS, and GCP. For us, we don't care. They get exactly the same capability at the end. It's really ideal. That's great. I think now that, by the way, now the customers can get our database everywhere. They can install an Oracle Cloud region on their premises. They can get Oracle from Azure. They can get Oracle from Google. They can get Oracle from AWS. They obviously can get Oracle from OCI. That Oracle database is becoming more and more capable. It does store most of the world's valuable data. It is by far the largest database installed base in the world with nothing remotely close. Most of those databases are still on-premise, but now they're beginning to migrate to the cloud. One of the big drivers of them migrating to the cloud is the autonomous version of the database. Now, perhaps just as importantly, the AI data platform, which allows you to take all of your existing data, all of your existing data, and make it available to any of the leading AI models. As I say, Grok, ChatGPT, Llama, all of them can immediately take advantage of your existing data and your existing database and turn it into insights and actions and agents directly, again, on your private data while keeping that private data private. That has been the missing link in companies and government agencies taking fully trained on all of the public data that's available on the internet. Now, that's a huge amount of information and makes for brilliant AI, but the AI does not have a lot of information about your company or your government agency because that data is not available on the internet. That is not data that the AI model was trained on. Now, with the Oracle Database 23AI, the AI model can look at your data, train itself on that data, and provide you with, again, insights, actions on your existing data while keeping it private. You do not have to share it with anybody. Fantastic. Thank you. Your next question comes from the line of Alex Zukin with Wolf Research. Your line is open. Hey, thank you, guys. Echoing your congratulations for a truly unbelievable bookings number. I guess maybe, Larry, could you opine on the current kind of state of the AI training versus inferencing opportunity? You have potentially investors worried about diminishing returns to training, even some of your hyperscaler peers seemingly walking back on some of their CapEx commitments. What are you seeing out there with respect to training versus inference, both in the incremental bookings that you're adding into the pipeline and maybe just how Oracle is differentiated on the inferencing side versus both hyperscalers and neoclouds, particularly with this AI Data Platform product that you just announced? Okay. Obviously, our training business is getting bigger and bigger and bigger very rapidly, as evidenced by our RPO. It was not just AI training that drove the RPO up. The AI inferencing, the potential of AI inferencing, and think about all of the Oracle databases out there, that data in those databases are going to train AI models. The AI models are only useful if they're familiar with your data. The AI models have to understand your products, your customers, your service requests, your financials. You have to make all that data available to the AI models in those databases. This is, we are right at the beginning of that. Again, on top of those Oracle databases now, we ourselves, because we are in the application-built business, have built lots and lots of agents on top of the Oracle databases and made those agents a part of our applications, modernizing and automating our applications. Customers need to do the exact same thing as they build software inside of their government agency or they build software, again, AI agents inside of their company. How are they going to, how do they go about doing that? They go about doing that is training the AI models on their data, on their data that is currently in an Oracle database. We make that very easy. You push a button. The new version of the Oracle database, 23AI, with vector capabilities, allows you to convert your data into vector format that's understood by the AI models. Nobody else has that. Nobody else has that. You can easily now train the AI models on your data for inferencing, obviously, and for building of agents. You can do that automatically with the Oracle database. We are the only one with that capability. We think inferencing in the end is a much bigger opportunity than AI training. There are literally millions of Oracle databases all over the world that will, and that data, all of those millions of databases, all of that data will be used to train AI models. On top of that, they'll build agents and applications. We think, again, that, I mean, we don't have these one or two or ten huge contracts for training because there aren't that many people building frontier models. There are hundreds of thousands of our customers that will be consuming those AI models and training those same AI models on their private data and then running agents and applications on top of all of that. That's a much bigger market than AI training for us. Super helpful. Maybe just, Safra, how to think about RPO trends and trending over the course of the next few quarters in light of that. I think there are going to be lumpiness, as you can see, but we actually expect some extremely significant numbers coming within the next few months also. There is just a lot of demand, folks. There is a lot of demand where people want to lock in and schedule in to our cloud. We are going to see increases in RPO. Remember, our remaining performance obligation, we also burn down some of it through the quarter as capacity goes online. I expect that number to be extremely large. This is enormous, but I expect it actually to continue to be very large, amazingly. Thank you so much. Your next question comes from the line of John DiFucci with Guggenheim. Your line is open. Thank you. You said you were live on 33 cloud regions and another 40 planned with Azure AWS. 18. 18. Okay. I am sorry. 18 multiclouds. Yeah. I I got you way ahead of yourself. I'm sorry about that, Safra. Anyway, I know you said you had another 40 planned with Azure, AWS, and Google. This quarter, we saw a big uptick in discussions between partners and large enterprises regarding Oracle database at Pick Your Hyperscaler. The discussions seem to be happening in mass, but not the deals yet. One of the things, a limiting factor in talking to partners, was that some of these global enterprises needed it to be global because they're global enterprises. Those 40, I guess, does that get you there for that? When do those 40s get deployed? Are they going to be deployed over the next year, or is it going to be over the next I don't know. About how long does that take? Of course, we're not in full control of that, but everyone's motivated to get them as quickly as possible because, as you know, the way the revenue flows through, it flows through to our partner or our host, and then they pay us. They are very, very motivated to get it as quickly as possible because that is holding up revenue for them until they can deploy it. It has been moving very quickly. They have accelerated recently. As you can imagine, there is some significant competition between those three hyperscalers to grab those workloads before their competitor takes them. It is moving quickly, but there is enormous demand. Again, enormous demand, but that capacity is not always within our control. They are very, very motivated to get it going because when those customers move, they often bring a lot of additional workload connected to the database into their cloud, but often directly into OCI also. Of course, we are the fourth competitor for them at the same time. I do not have the exact dates, but someone in my organization no doubt does. We do expect it in short order to be a lot more deployed very quickly, and demand is extremely high. Can I help? I think I can help on this. Everyone needs a primary and a backup data center in North America. Everyone needs a primary and a backup data center in Europe, typically Western Europe. Everyone needs a primary, and not everyone, and a lot of people need primary and backups in Asia. That is six data centers. Once we get to a primary and a backup in North America and Europe and in Asia for, let's just say, Google, we're ready to roll. Most of the impediments are out of the way. It is just a matter of building the people that want to that are Japanese national companies. That is for the multinationals. For the multinationals, that works. There is obviously the Japanese domestic market and the German domestic market and so on and so forth. I think by the time we get to 40, which is around close enough to say 12 months from now, and Safra said it is not entirely in our control because AWS has to provide us with the space because we are literally embedding the OCI data centers inside of AWS, inside of Google, and inside of Azure. We expect this is growing extraordinarily fast, and we think we'll be able to meet most of the needs of customers that need for primary and backup around the world, certainly in the coming months. It is just going to be adding capacity country by country. Your next question comes from the line of Kirk Mattern with Evercore ISI. Your line is open. Thanks very much for taking the question, and I'll add my congrats on the RPO and booking strength this quarter. Larry, you mentioned some of the agents you've been building out on top of your application platform earlier. I was just wondering, are you now starting to see the demand for those technologies or those functionality that agents are bringing starting to have an impact on your strategic SaaS business and perhaps the pace of conversions from any legacy on-prem systems that still might be out there? Thanks. Okay. I would say, I mean, our biggest differentiator is when we're competing in the healthcare field, our biggest differentiator is the quality of our AI agents, that we have a lot of AI agents in healthcare. One, we listen when a doctor consults with a patient, we listen to that consultation, and we record all the prescriptions, all the doctor's orders, all the doctor's notes. We automatically provide the doctor, prior meeting with the patient, a summary of the patient's, their latest lab tests and vitals. When the doctor is finished meeting with the patient and has given all his orders, prescriptions, and come to diagnostic conclusions, we automatically update the electronic health records. They're taking a huge burden off the physician. They don't have to type any of this stuff in. We just listen. The interface to our system is primarily voice, which is all AI. The whole system is made up of AI agents. Let me give you another one, another AI agent that we're in the process of building. The doctors have to, the hospitals have to get permission to prescribe an expensive cancer drug or to do a heart transplant or something like that or a knee transplant. They have to get prior authorization from the payer, the insurance company in the United States or the government, like the NHS in the U.K. There is this negotiation. You send your patient's electronic health records to the insurer. The insurer looks at that, analyzes that, and says, "Yes, this person is authorized to use Herceptin or a cancer drug or not." That is all done manually on the phone. Our AI agents, we read the insurance policies. We read all the electronic health records. We actually make the determination whether this is a reimbursable drug or this is a reimbursable surgery. We do that. We automate that entirely, make it completely electronic. This saves billions of dollars in the healthcare field and makes a huge difference and determines whether they're going to buy our system or buy some alternative system. I know people are saying, "Well, what's the dollar impact when you're selling the agents?" The dollar impact is we sell an entire country, an entire health system based on whether our agents and our health software is better than our competitors and saves them money. It's not just money really attributable to AI agents in healthcare. It's the fact we're selling more and more healthcare systems because we have a lot of AI agents embedded in them, which helps doctors, which produces better outcomes for patients and saves governments and payers money. This is applied to our Fusion applications, which have dozens of embedded agents, whether it's in supply chain, financials, HCM. We have literally dozens of agents already embedded in our applications. Unlike our competitors who are talking about it, we actually have them already built and deployed at customers. If I can make, let me make one more statement. It's going to be, since our applications are going to be primarily AI agents, again, I'm going to say it again, the applications themselves will migrate to be basically a bunch of connected AI agents. You're not going to be able to separate how much of this, how much did you make on the AI agents and how much did you make on the rest of the applications. All of our applications are becoming AI agents. Safra, I'm going to ask you a really quick follow-up. Is this the tipping point for any customer that hasn't moved to the cloud to have to get there now to get this functionality? Talking to some partners, it feels like anybody that's been holding out is now ready to go because they can't get any of this functionality if they're still on-prem. Thanks. Yes. I mean, this is the motivator, the ability to have your system do so much of your work. As I was signing off with my finance and audit committee and talking with my Chief Accounting Officer, they have the entire description of the balance sheet issue, all the different balance sheet parts, all done because of our products. That is such a time saver and so much incredible productivity and insight that you're at a disadvantage if you do not use this. That is why I know every quarter, I mention again, we're announcing. Of course, it's Monday. I could not announce Sunday or Saturday or, of course, Friday, something you do only if you have bad news. I mean, I had to announce the 10th. Imagine so much of the work that my teams do is enabled because of these advanced technologies. It's your way to get there. Ultimately, everyone's going to come to it, and it should be motivating them because it is such a massive, not only productivity improvement and as a result, lower cost, but it also gives you incredible insight into your business. It's really amazing. Your final question comes from the line of Mark Moerdler with Bernstein Research. Your line is open. Thank you very much for taking my question and really congratulations on how incredibly well this is being executed. We know Oracle spends less on CapEx per dollar of IaaS PaaS revenue than your larger hyperscale cloud provider peers. How should we understand why CapEx is lower? How should we think about the trajectory of CapEx given the strength of RPO and especially the strength of OCI and OCI AI? Thank you. Okay. I'll take a crack at that. We can start our data centers smaller than our competitors, and then we grow based on demand. Building these data centers is expensive, and they're really expensive if they're not full or they're not at least half full. We tend to start small and then add capacity as demand arises. That allows us to have higher utilization. That's one thing. The other thing is we have a high degree of standardization and automation inside of our cloud. Operating the cloud also gives us better margins. You will not see that on CapEx. You'll not see that on CapEx. That will be on operating profit. It's really the combination of starting smaller and growing with demand that affects CapEx and then overall margins, the fact that we have a high degree of automation, which lowers our labor costs dramatically. By the way, more important than lowering our labor costs, with no labor, there's no human error. There's no human mischief. So we're much more reliable and much more secure because we don't have a lot of human beings in our data centers. Makes a lot of sense. Thank you. Thank you, Safra. Thank you, Larry. Thank you, Mark. A telephonic replay of this conference call will be available for 24 hours on our investor relations website. Thank you for joining us today. With that, I'll now turn the call back to Abby for closing. Ladies and gentlemen, this concludes today's call, and we thank you for your participation. You may now disconnect.