Q4 2024 Oracle Corp Earnings Call

Participants

Ken Bond; Senior Vice President, Investor Relations; Oracle Corp

Safra Catz; Chief Executive Officer; Oracle Corp

Lawrence Ellison; Chairman of the Board and Chief Technology Officer; Oracle Corp

Raimo Lenschow; Analyst; Barclays

Brad Zelnick; Analyst; Deutsche Bank

Siti Panigrahi; Analyst; Mizuho Securities

Alex Zukin; Analyst; Wolfe Research

Kirk Materne; Analyst; Evercore ISI

John DiFucci; Analyst; Guggenheim Securities LLC

Presentation

Operator

Good day, everyone, and welcome to Oracle's fourth-quarter 2024 earnings call. Today's call is being recorded. And now, I would like to turn the conference over to Ken Bond. Please go ahead.

Ken Bond

Thank you, Krista. Good afternoon, everyone, and welcome to Oracle's fourth-quarter and fiscal year 2024 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 the 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 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 factors and other risks that may affect our future results or the market price of our stock.
And 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 questions, we'll begin with a few prepared remarks. And with that, I'd like to turn the call over to Safra.

Safra Catz

Thanks, Ken, and good afternoon, everyone. Clearly, we had an absolutely incredible quarter. As you know, Oracle's Q4 is known for customers purchasing large software license contracts to power their businesses. But because the pivot to the cloud, this Q4 was power powered by the enormous demand for our cloud services. And they showed up in RPO or remaining performance obligations.
In Q4, Oracle signed the largest sales contracts in our history, led by huge demand for training large language models as well as record levels of sales for OCI, Autonomous, Fusion, and NetSuite. RPO was $98 billion, up $18 billion from Q3 and up 44% year over year from $68 billion last year. And we are trading one-time, non-recurring license revenue in return for much bigger strategic customer commitments from multi-year cloud revenue, from which we expect to further accelerate our revenue growth rate.
This is exactly really what we've been targeting. And it bolsters my confidence that our overall revenue, earnings, and cash flow performance, as well as our growth rates, will only get stronger and accelerate. In short, this Q4 marks the full emergence of our high-growth cloud businesses.
Now I started talking about this tipping point four years ago, and you've seen it continue to play out in our results since then. As a reminder, we accelerated our US dollar revenue growth rate from negative 1% in fiscal year '20 to plus 8% this past year if you exclude Cerner.
In addition, EPS has grown at a 10% compounded annual growth rate over that same period and both operating cash flow and free cash flow, which of course we report on a trailing 12-month basis, were each declining 10% four years ago. This year, they grew 9% and 39%, respectively.
Now, customer conversations are now absolutely only focused on our cloud services as the results clearly show. So let me give you just a couple of -- a few examples. First, as you saw, OpenAI selected Oracle to run deep learning and AI workloads on Oracle Cloud Infrastructure. Like many others, OpenAI chose OCI because it is the world's fastest and most cost-effective AI infrastructure. In total, we signed over 30 AI contracts for over $12 billion this quarter and nearly $17 billion this year.
Second, we continue to expand our work, helping companies use our Cloud Applications portfolio to reinvent their businesses. As an example, a very large enterprise tech company signed a contract in Q4 for over $600 million where we will be helping them transform their operations with Fusion to enable them to become more agile, faster-growing, and more profitable.
May I say, in the process, we will replace out many of our competitors' product, these cross-pillar cloud deals or suite deals focused on business process reengineering that incorporate multiple-cloud applications that no one else can offer.
And I would point out, by the way, that today is day 11 of our new fiscal year and we are once again announcing our results not only for the quarter, but the year, and giving guidance, making us faster than any other public company by a long shot. We are able to do this because of Fusion applications, and that is why companies are choosing Fusion and our wonderful teams are showing them the way.
And third, I'm pleased to announce we signed another multi-cloud partnership this time with Google. OCI and Google Cloud Network Interconnect is available immediately in 10 regions, and we will be live with Oracle Database at Google Cloud in September where customers can get direct access to Oracle Database services running on OCI deployed in Google Cloud data centers.
So what's driving this? Well, it is all about our comprehensive, highly differentiated, and secure cloud offerings. Customers have progressed from their initial curiosity about Oracle Cloud into full-blown rollouts. We have the most secure, complete, and cost effective set of enterprise applications and infrastructure cloud technologies of any vendor.
Not only are our cloud technologies vertically integrated to work together, but we offer flexible deployment models like public cloud, multi-cloud, sovereign cloud, dedicated cloud, or any other way our customers ask us to deliver.
And we also offer Oracle Alloy where Oracle partners become cloud providers, offering customized cloud services alongside the Oracle Cloud.
Now, I'm now going to dive into the details of Q4 and finish my prepared remarks with how this strength and momentum will impact fiscal year '25 and beyond. Okay. So let's start.
In Q4, the dollar strengthened from the time of my Q4 guidance. So we saw a 1% currency headwind to total revenue and a $0.01 currency headwind to EPS. As usual, I'll be discussing our financials using constant currency growth rate because this is how we manage the business.
Total cloud revenue, that is SaaS plus IaaS, excluding Cerner, was $4.7 billion, up 23%; including Cerner, total cloud revenue was up 20% at $5.3 billion; and SaaS revenue of $3.3 billion, up 10%; and IaaS revenue of $2 billion, up 42% on top of last year's 77% growth.
Total cloud services and license support for the quarter was $10.2 billion, up 10%, driven again by our strategic cloud applications, Autonomous database, and OCI. Applications subscription revenues, which includes product support, were $4.6 billion and up 6%.
Our strategic back office SaaS applications now have annualized revenue of $7.7 billion and were up 16%. Infrastructure subscription revenues, which includes license support, were $5.6 billion, up 13%. Infrastructure cloud services revenue was up 42%.
Excluding legacy hosting, OCI Gen 2 Infrastructure cloud services grew 44% with an annualized revenue of [$7.4 billion]. OCI consumption revenue was up 63%. Were it not for continuing supply constraints, consumption growth would have been even higher. Database subscriptions, which includes Database license support, were up 6%, highlighted by cloud database services, which were up 26% and now have an annualized revenue of $2 billion.
Very importantly as on-premise databases migrate to the cloud, either to OCI directly or using database at Azure or database at Google Cloud, we expect these cloud database services will be that third leg of revenue growth alongside OCI and strategic step. Consistent with our strategic direction and reflecting customer preference for cloud services, software license revenues were down 14% to $1.8 billion. So all in, total revenues for the quarter were $14.3 billion. That's up 4% if you include Cerner, up 5% excluding Cerner.
Shifting to margins. The gross margin for cloud services and license support was 77%. This is a result of the mix between support and cloud, in which cloud is growing much faster than support. The gross margin percentages for software support and SaaS are consistent with last year.
While IaaS gross margins improved substantially, gross margins will go higher as more of our cloud regions fill up. We monitor our expenses carefully to ensure gross margin percentages expand as we scale. To that point, though, the gross profit dollars of cloud services and license support grew 8% in Q4.
Non-GAAP operating income was $6.7 billion, up 9% from last year. The operating margin was 47%, up from 44% last year as we continue to drive more efficiencies in our business. Looking forward, as we continue to benefit from economies of scale in the cloud, we will not only continue to grow operating income, but we will also expand and the operating margin percentages.
The non-GAAP tax rate came out over a 1% higher than my guidance at 20.1%, and non-GAAP EPS was USD1.63, and GAAP EPS was USD1.11. As a reminder, the non-GAAP tax rate last year was 9.2% and this had an adverse effect on this quarter's EPS growth.
Non-GAAP pre-tax income grew 14% in constant currency. So you can figure out that had we have the same tax rate last year as this year, net income would have grown 14% and EPS would have been up 12% in CD, 11% in USD.
For the full fiscal year, total company revenue was $53 billion, up 6%. Total cloud services and license support revenue, which is entirely subscription base and accounts for nearly three-quarters of total revenue was $39.4 billion, up 11%.
Total Application subscription revenue grew 9% and Infrastructure subscription revenue grew 13%. Total cloud services, excluding Cerner were up 26% to $17.1 billion -- $17.2 billion. SaaS revenue, excluding Cerner, was up 13% to $10 billion -- $10.4 billion for the year.
IaaS and Cloud Infrastructure revenue was up 60% to $6.8 billion for the year, with consumption revenue up 66% from last year. Non-GAAP EPS for the full year was USD5.56, up 9% in USD. And the full-year operating margin percentage was 44%, up from 42% last year.
At quarter-end, we had nearly $10.7 billion in cash and marketable securities. The short-term deferred revenue balance was $9.3 billion, up 4%. Over the last four quarters, operating cash flow was $18.7 billion, up 9% and free cash flow was $11.8 billion, up 39%.
Capital expenditures were $6.9 billion. As I mentioned, our remaining performance obligations or RPO is now $98 billion, up 44% in constant currency. And the portion excluding Cerner, if you're curious, was up 60%.
We signed several large deals in this quarter and we have many more -- many, many more in the pipeline. Approximately 39% of total RPO is expected to be recognized as revenue over the next 12 months. And this reflects the growing trend of customers wanting larger contracts as they see firsthand how Oracle Cloud services plan are benefitting their businesses. Now, while we spent $3.5 billion on CapEx this quarter, the $2.8 billion shown in the cash flow statement is lower simply as a result of timing of payments.
We are working as quickly as we can to get cloud capacity built out given the enormity of our backlog and pipeline. At this moment, we have 76 customer-facing cloud regions live with 47 public cloud regions around the world, and another 19 being built. We have 11 database at Azure sites live and more locations with Microsoft coming online soon.
We will have 12 Oracle Database at Google Cloud sites live this year. We also have 13 dedicated regions live and 15 more planned. We have several national security regions and EU sovereign regions live with increasing demand for more of each.
And finally, we already have two Alloy cloud regions live with 11 more planned. Of course, we also have many, many, many cloud customer installations as I mentioned earlier, the sizing and flexibility and -- the size and flexibility and deployment optionality of our cloud regions continues to be incredible -- incredibly advantageous for us in the marketplace.
This quarter, we purchased 1.25 million shares for a total of $150 million. In addition, we paid out dividends of $4.4 billion over the last 12 months, and the Board of Directors today declared a dividend of $0.40 per share.
Before I discuss my guidance for Q1 and fiscal 2025, I do just want you to have a couple of notes. The first is that in Q4, we decided to exit the advertising business, which had declined to about $300 million in revenue in fiscal year '24. Also, I will no longer be breaking out the Cerner business in my results. And even though it will begin to grow modestly throughout the year in both revenue and operating margin, it's not necessary to break it out anymore and -- because it is now operating in a growth mode.
Now to guidance. Throughout fiscal year 2025, I expect continued strong cloud demand to push Oracle sales and RPO even higher and result in double-digit revenue growth this fiscal year. I also expect that each successive quarter should grow faster than the previous quarter as OCI capacity increases to meet demand.
We believe our momentum, our current momentum will continue as our pipeline is growing even faster than bookings and our win rates are going higher as well. I expect fiscal year '25 Cloud Infrastructure services to grow faster than the 50% we reported this year. CapEx In fiscal year '25 will probably be double what it is in fiscal year '24 -- what it was in fiscal year '24.
Okay. Beyond this fiscal year, I remain firmly committed to our fiscal year '26 financial goals for revenue, operating margins, and EPS growth. However, given our strong bookings results, I believe some of these goals might prove to be too conservative given our momentum. We are going to provide you a more fulsome update on all of this at the Financial Analyst Meeting at Oracle Cloud World in Las Vegas in September.
Okay. Let me now turn to my guidance for Q1, which I'll review on a non-GAAP basis. Now if currency exchange rates remain the same as they are now, currency should have a negative 1% effect on my revenue and either $0.01 or $0.02 negative on EPS in Q1. However, as you all know, actual currency impact may be more or less; I just can't get that now.
Total revenue for Q1 are expected to grow from 6% to 8% in constant currency and using the currency situation as it is now, they're expected to grow from 5% to 7% in USD. Total cloud revenue is expected to grow from 21% to 23% in constant currency and 20% to 22% in USD. Non-GAAP EPS is expected to grow between 11% to 15% and be between $1.33 and $1.37 in constant currency.
Non-GAAP EPS is expected to grow between 10% to 14% and be between USD1.31 and USD1.35, but this time in USD. My EPS guidance for Q1 assumes a base tax rate of 20%. And as always, one-time tax events could cause the actual tax rates to vary from my guidance.
Okay. I know that was long. But with that, let me turn it to Larry for his comments.

Lawrence Ellison

Thank you, Safra. I'm going to start by repeating something Safra said. In Q4, Oracle's company-wide RPO increased 44% to $98 billion. In AI alone, we signed contracts with 30 different customers for $12.5 billion in new AI business. These astonishing RPO numbers 44% and $98 billion were driven by massive increases in sales of Oracle Cloud Infrastructure, OCI.
So who are the companies choosing to use Oracle Cloud Services and Oracle data centers? Well, here are a few names: NVIDIA, Microsoft, Google, xAI, OpenAI, Cohere, and dozens more. In other words, the world's largest cloud companies and the world's most successful and accomplished AI companies choose to use Oracle Cloud services and data centers.
So why are they working with Oracle? Because Oracle's Gen 2 Cloud Infrastructure is different. OCI's (inaudible) main network moves data much faster. And when you charge by the minute, faster also means less expensive. OCI trains large language models several times faster and at a fraction of the cost of other clouds.
OCI's critical cloud software, the operating system and the database are fully autonomous. At OCI, human beings do not run the operating system or the database. Autonomous software robots do. No one else has this level of autonomy in the cloud. Eliminating human labor eliminates human error. Almost all cloud security breaches begin with human error, eliminating the possibility of human error is the only way to make certain your cloud data is not stolen.
That's it. The most important technology companies in the world are using OCI because it's faster, less expensive, and more secure. Easy to say, not easy to do. Back to you, Ken.

Ken Bond

Thank you, Larry. Krista, if you could please poll the audience for questions. We'll begin the Q&A portion of the call.

Question and Answer Session

Operator

Raimo Lenschow, Barclays.

Raimo Lenschow

Perfect. Thank you. Congrats from me. These are very impressive numbers. Safra, can you try to help us bridge the strong RPO number and how we need to think about feeding that into revenue? Is that just the capacity function? Or is there anything on the customer side that you need to deliver or the technology side you need to deliver? Just help us to bridge the gap on those. Thank you.

Safra Catz

It's all about capacity. It is -- as we bring the capacity online, wherever it's going online around the world is when those workloads are coming over. A lot of the engineering work is done in advance so that those customers know how they can operate.
They bring smaller workloads, but the bigger workloads, they're just waiting for us to go online and make it available to them. It is really that level. We are scheduling them on our availability. And, as I mentioned, our pipeline to take more deals is all about us just getting the capacity up and live and moving forward.

Raimo Lenschow

So it's just a mechanical problem in a way.

Safra Catz

Yeah. Well, it's not a problem. It's just the schedule. As things come online, as the data centers go live, or as we deliver the computers, they're just getting -- it's just very straightforward. There's no magic here. These customers have done a lot of the analysis and the engineering in advance and have tested us or competed us against our competitors and have chosen us very -- already understanding how we work, and they're just waiting for us to give them more capacity.

Raimo Lenschow

Good. Very good. Thank you.

Operator

Brad Zelnick, Deutsche Bank.

Brad Zelnick

Great. Thank you very much and congrats from me as well. Larry, it's great to see the amazing momentum in OCI, especially given it's a competitive market and the leading names in AI are coming to you, wanting to partner with Oracle. Can you talk about the innovation road map for OCI and your AI services in particular? And why we should expect Oracle to keep on winning not just today, but over the next several years to come in this market?

Lawrence Ellison

Okay. Well, I think in OCI, we've talked for a while about our ability to build very small data centers, one you could put in a shipper, a submarine, or a full cloud, a full Oracle Cloud, we will soon have in six standard half racks to go into a conventional data center.
So virtually any one of our customers could choose to have the full Oracle Cloud in their data center with every service, every service in the cloud. And they could scale that up quite extraordinarily large. So we talk about the fact that we can start very small and that's a huge difference between us and our competitors. So we can actually put it again customer by customer, small countries, we can do.
What we haven't talked so much about is we're also building the largest data centers in the world. We talked about -- I think we talked briefly about one last call, where we can park -- it's a 70-megawatt data center where we can park eight 747s nose to tail in the data center. It's a huge AI training data center.
Well, we're also building a 200-megawatt data center. In fact, this past quarter, we sold about half of that data center for the -- for a period of time. So we're now bringing 200-megawatt data centers online. So we are literally building the smallest, most portable, most affordable cloud data centers all the way up to 200-megawatt data centers ideal for training very large language models and keeping them up to date.
This AI race is going to go on for a long time. It's not a matter of getting ahead, just simply getting ahead in AI, but you also have to keep your model current. And that's going to take larger and larger data centers. And some of the data centers we have -- that we're planning are actually even bigger. There -- some are getting very close to our there, say, 1 gigawatt, which is a pretty good-sized city or one enormous AI cloud training data center.
No one else can span this range. And in every case, we have unbelievably fast networks that are part of this. The data centers we're building include the power plants and the transmission of the power directly into the data center and liquid cooling. And because these modern data centers are moving from air cooled to liquid cooled, and you have to engineer them from scratch.
And that's what we've been doing for some time. And that's what we'll continue to do. And currently, we are leading the pack and being able to deliver that quality and that scale of data center.

Brad Zelnick

Amazing. Thank you so much, Larry.

Operator

Siti Panigrahi, Mizuho.

Siti Panigrahi

Thank you. Larry and Safra, it's impressive to see how fast you ramped OCI as you're now available in 11 data centers. And then now with this Google partners, we'll have Oracle Database at Google Cloud. So I have two questions: one is, as you embark on offering this multi-cloud flexibility to customer, when can we see a similar partnership with AWS? And second is, how should we think about these partners is helping your customers migrate their on-prem Oracle workloads to cloud?

Safra Catz

I don't know, Larry, you wanted (multiple speakers)

Lawrence Ellison

No. I could start. Well, we believe in giving customers choice. And customers want choice. Customers are using multiple clouds, not only infrastructure cloud, but they might have Salesforce applications or Workday applications. They use multiple cloud in their business right now.
So it's very important, we think, that these -- that all the clouds become interconnected. So we're thrilled to have the connection with Microsoft and be building OCI data centers inside of it -- right inside of Azure. So the computers are next to each other to minimize network costs and network latency, which is all good things.
We're doing the same thing with Google. We would love to do the same thing with AWS. We think we should be interconnected to everybody. And that's what we're attempting to do in our multi-cloud strategy. I think that's what customers want. So I'm optimistic that's the way the world will settle out.
We'll get rid of these fees for moving data from cloud to cloud and all the clouds will be interconnected, and customers can pick their favorite service from their favorite cloud and mix and match whatever they want to use and do it easily and seamlessly.

Safra Catz

Yeah.

Siti Panigrahi

Thank you.

Operator

Alex Zukin, Wolfe Research.

Alex Zukin

Hey, guys. Thanks for taking the question. I wanted to dive a bit deeper on just precisely how many deployment models you guys are offering for OCI because it feels as though that is getting particularly differentiated as we start to think of sovereign cloud, GovCloud, more private cloud, given the conservative posture for AI and data privacy.
So how do we think about how much of an advantage that is providing in sales cycles? And maybe in that massive $30 billion-plus in the second half RPO, but also just comment on the magnitude of that opportunity going forward.

Lawrence Ellison

I'm going to take a swing at this one. We can -- every medium-sized on-premise customer that Oracle has could have a private -- full Oracle Cloud where they have no neighbors. They are the only user of that Oracle Cloud, and we can install that in their existing data centers. Nobody else can do that. You have to move to the public cloud.
Now we have public cloud, we have a lot of public cloud regions. We love the public cloud. But if you're very conservative and you want to absolutely maximize security and that's important to you, we can put in a cloud, a full Oracle Cloud, and we run it. We pay for the -- again, it's an Oracle region. We put in Oracle Cloud region and let me just make up a name.
Samsung, we could build a cloud region for Samsung. In fact, two cloud regions is for Samsung. We could do two cloud regions making up names, General Motors, Ford, anybody, any company. Those are pretty big companies, but much smaller companies as well.
So we're the only ones that give you an option to have the full capability of a public cloud run by Oracle, all of our services, every single one of our services. You don't pay for the hardware, you just pay for what you use, put that model directly on your premises, and you can use it, and no one else is in that cloud.
We can do that. No one else can do it. We can put them on ships and on submarines, no one else can do it because we can start very, very small. All Oracle Clouds are identical, except for scale. All Oracle Clouds have all Oracle services. All Oracle Clouds are fully automated because they're identical. They're fully automated.
So one of the reasons we took a little bit longer to get our cloud out was because we built something quite different than what our competitors have. And that allows us to go from very small to very large, using the same automation software. I think some of our competitors, they're large data centers, some are quite different than other data centers.
They might have different -- some services might be available on some data centers and not in others. They're not -- they did a very different approach to what we did. We had the advantage of seeing what all the other guys did and we took a different road. It took us a bit longer, but we think we're better off in terms of security.
We're better off in terms of scalability. By the way, that means the ability to go down in size and up in size. It allows us to get to every corner of the globe and provide a level of privacy for your data that other cloud providers cannot provide.

Safra Catz

Yeah. And because, as Larry said, because whatever the deployment model is, you don't have to compromise. Some of our competitors may offer some level of sovereignty or some level of disconnected, but they don't actually have all the services. For us, and the reason we've been so successful is whether it's disconnected or sovereign or whatever it is, the customer always gets everything all services, not just some services.
And they get to deploy it any way they want, and they get the security or the regulatory requirements. Sovereignty may be very critical. And for most governments, they don't want their data in the public cloud out and about. They want to have it sovereign to their country. And so no compromises, no compromises on the services and no compromises on security.

Alex Zukin

It also sounds like you guys have a better price in most cases.

Safra Catz

Oh, much.

Alex Zukin

So thanks, again, and congrats on a good quarter.

Safra Catz

Much. Because we're so much faster, when you use our cloud, it is new. It's modern, but it also has technical advantages and so it runs your workload so much more quickly. And when you pay by the minute, the second, the hour, if your workload ends in 1/10 time, you pay a 1/10th the price. That's very hard to compete with.

Lawrence Ellison

One last comment -- maybe one last comment. The other thing is our cloud was designed not for hundreds of regions, but for thousands or possibly even tens of thousands of data centers and regions. That's the -- that's why we had to put in a high degree of automation. There is no way we could run these data centers manually.
There are too many of them, and we're building them to do fast. We couldn't hire people fast enough and train people fast enough. And the risk of them making a mistake, an error is the risk -- well, they start exposing our customers' data. So they are highly automated. It's a little bit like I apply to myself, and comparing it to the satellites that Elon Musk puts in the sky.
StarLink has -- they're more -- he has more satellites than everyone else in the world combined because, again, it's a very different -- it's a satellite system, Starlink, that's designed for a very large number of satellites that are highly automated. And same model lots and lots of them, 100% or nearly 100% automation to run these clouds.

Operator

Kirk Materne, Evercore ISI.

Kirk Materne

Yes. Thanks very much and I'll echo the congrats on the cloud momentum. Larry, Safra, I was wondering if you could just expand a bit on the OpenAI announcement this afternoon. Just what that entails in terms of how you'll be working with them or Microsoft? Are there certain workloads they'll be working on with you directly? Can you just give us whatever additional color you can on that deal? Obviously, very excited. Thanks.

Safra Catz

Well. Go ahead. No, you can. Go ahead.

Lawrence Ellison

Okay. All right. Well, we're building a very, very large data center, very big -- about half of a huge data center. We're building for them, lots of NVIDIA chips, the new NVIDIA chips, the new NVIDIA interconnect, liquid cooled, and they're primarily for training. I mean node inference things. It's we're doing masses and masses of training.
And I don't know, that's what we're doing and the training goes beyond languages because now these systems are -- even the call -- even though they're called large language models, they really -- part -- the proper name is probably neural networks, they're neural networks, and they're trained not just with language, but masses of images as well.
For example, Oracle is very involved with taking biopsy slides and using microscopes to read biopsy slides, recording those images, and then using AI to diagnose cancer from these biopsies. It's one of the projects we're working on the medical side of our business. And these large language models, strangely enough, are also looking at biopsies.
They're not just reading things, language. They're also looking at images and interpreting images. So that is actually a bigger and more complicated problem than understanding language. That's what's so exciting about -- again, second time I mentioned Elon and an Elon company. Tesla is very close to getting full service driving authorized in China.
I'm not speaking at it. It's cool. I think the Chinese government is moving along the full self-driving -- full self-driving in China. In order to train a car to do full self-driving, you train it on vast amounts of images because the car has to look at these images and then decide what it's going to do next.
That's what it does. It doesn't speak, it responds to what it sees. That's a very different problem than answering a question posed in any language. So everyone is going to be training their models on imaging. That's a huge amount of additional data. It's a huge amount of additional training and we're right in the middle of it.

Kirk Materne

Thank you.

Operator

John DiFucci, Guggenheim Securities.

John DiFucci

Thank you for taking my question. My question, I think, is for Safra. Safra, the IaaS revenue growth has been really impressive, and it has been for a while here, but perhaps even more so the last couple of quarters, especially is the backlog given its scale. And this may be somewhat of an obvious question for you, but it's based on my conversation with investors.
There's two high-profile topics that I want to make sure we understand what the contribution has been today versus next year. And that's Oracle database at Azure and AI in general. We've heard a lot of conversation about the former when we speak to partners and customers in the field.
And you've spoken a lot about the latter today. So beyond conversation volume, can you talk a little bit more about what the contribution of these two topics has been to that impressive IaaS revenue growth in this quarter versus what we should expect that contribution to be in fiscal '25?

Safra Catz

Okay. I would tell you that both of them, both -- whether it's database at Azure or even the AI workloads as they come on board, they are all incremental to anything you saw so far in our revenues, okay? The database at Azure, those centers are just going live now. So even though we are selling quite a bit of ARR there, these are small and growing very, very fast.
So the revenue in Q4 of, let's say, Azure was very small. Q1 will be 10 times as much. Q2 will be potentially 30 times as much. So it is extremely incremental to our current run rate. By the way, that is also true to -- we've already -- we have revenue -- AI revenue so far. Yes, we do, and we've been announcing those.
These contracts that we are signing -- that we've signed at the end of Q3 and that are signed at the end of Q4 are so much larger in size that they will be incremental to everything you saw this past year, literally, incremental added by quite a bit. So it is going to be -- this is a very exciting time, obviously. And everything is incremental to what you've seen so far because it dwarfs it in many ways.

John DiFucci

That is really clear, and it really speaks to, I think, what you started talking about a long time ago, especially a lot more publicly, geez, I don't know, the fall of '22, but thanks. That's really clear.

Safra Catz

Thank you.

Ken Bond

Thank you, John. A telephonic replay of this conference call will be available for 24 hours on the Investor Relations website. Thank you for joining us today. And with that, I'll turn the call back to Krista for closing.

Operator

Ladies and gentlemen, this does conclude today's conference call. Thank you for your participation and you may now disconnect.