Microsoft Corporation (MSFT) Earnings Call Transcript & Summary

Welcome, everybody. Back-to-back software sessions here to kick off the conference. So I'm Brad Sills, senior analyst on the software team and cover Microsoft and others and very much looking forward to the conversation. Delighted to be welcoming Microsoft to the conference. We're very fortunate to have Matt McBride here with us, who's Corporate VP of the commercial business and CFO of Commercial. So Matt, thank you for joining.

Yes, happy to be here.

Great to have here.

So we'll just get started. Why don't we just start off with a little bit on your background and your role at Microsoft, your area of responsibilities to start it off, please?

Yes, fantastic. Yes. My name is Matt McBride. I'm the Commercial CFO. You should think about that as kind of I have a couple of different additional hats I wear in addition to the CFO. I have all the commercial revenue, over $200 billion, growing super well. And then I have a horizontal role for basically all the capital spend for Microsoft with the exception of like Office space. And then all the cost of goods sold with the exception of Surface and Gaming. And so commercial consumer, anything having to do with the cloud. And they have the same responsibility for R&D. So you can think about my job as being a typical commercial CFO plus a responsibility I have to advise our senior leadership team, Satya and Amy and others, on how to allocate capital for the most efficient return across the company. And it's been a wild ride a little over 19 years, and it's been super fun to be a part of.

That's great. And why don't we start with Azure, an area of responsibility there? Really impressive growth this past quarter at 35%, exceeding your outlook and your guidance and strong guidance for next quarter. Why don't we just start high level? What are some of the key growth drivers within Azure that you'd like to point out that are driving this type of results? Just remarkable.

Yes. It's been -- we had, like I said, a good quarter. We're very happy with it. I'll start first with things like geography, markets, SMB versus enterprise, things of that nature. From that perspective, the performance has been very broad-based. We're not seeing any particular geography outperform another. We're not seeing a particular industry segment or part of the market that's really outperforming another. So it's been good broad-based sales execution across all those different channels. From a product perspective, I'd start with -- maybe I'll cover 3 things, and I'll start with migrations. Migrations have been going strong. It's -- people want to say it's very late innings in the cloud game. And in fact, it's not. There's still lots of customers who are in different stages of their digital journey. They're still migrating workloads and they're replatforming on applications, and they're coming to Microsoft to help them do that. And so we've seen tremendous progress there. Additionally, as people think about their data estate, we've been doing quite well in data. In particular, as customers think about how to get the most out of their data, especially when they think about adopting AI services that are increasingly differentiated using that proprietary data, they're choosing Cosmos DB, a very large, scalable, high transaction volume solution for that type of need. And then lastly, our analytics products are doing quite well. And so again, great progress there, and we've seen good progress on AI as well.

Excellent. Excellent. Why don't we dig into AI specifically, if we could, please? Can you talk about some of the traction you're seeing with Azure AI services? And just generally, when you look across the Azure stack, where are you seeing the AI adoption cycle manifest in the product?

Yes, yes, sure. First of all, I'll zoom out a little bit and just say like Microsoft's total set of solutions for AI from Azure plus all of our Copilots is doing quite well broadly. And I think the pitch to customers is that broad pitch in addition to what you want to talk about with just Azure. And we've shared that 85% of the Fortune 500 have adopted some form and are using some form of Microsoft AI services. And it's tremendous progress. I think customers are increasingly looking at how do they go about doing different tasks. And in some cases, it might be improving the employee experience. It might be helping them deliver new customer value, customer engagement, business process engineering. And then for them, they might be at different places in the AI adoption and innovation curve, and then we want to accelerate that curve for the customer. And we see those 4 things happening pretty regularly when customers are adopting AI. I'll touch briefly on Foundry for a second. Foundry has over 70,000 customers already. And these customers are all different shapes and sizes. And I'll give you a couple of examples here of people who are using these services. Just down the road, there's Stanford Health Center. And they've gone into Foundry and to use -- to solve a solution and solve a problem. And one of the problems they have is they have to come up with ways to get aligned care for people with cancer. And this process they have is called the tumor board. And it's very manual. It has lots of different role types from data scientists to clinicians to stitch together what they need to be able to go understand and align the care and make a plan to move forward. So what they did is they went into AI Foundry. We have a service in our catalog called Health Service Orchestrator -- Health Agent Orchestrator. And they use that to stitch together all those different sources, all the different role types to basically go out and automate that process. And that matters a lot, especially in the space when you talk about health. That time savings for all those people is now spent on getting to a faster plan. So the patient is getting care faster. And then the clinicians are saving time that they can then spend on saving more lives. And so that's one, again, very serious life-saving example. I'm not trying to lay it on too thick, but like it's a real example. Maybe a lighter weight example is Ernst & Young. They're providing tax services around the world. And you think about that as a problem. That's lots of documentation, lots of localized regulation and a practice that's, again, global in nature. And so what they've done is gone into Copilot Studio, and they've created agents and capability to go reason over 21 million documents. I mean it's huge scale over many, many years, again, localized so that now when their tax professionals need information, they can go query, again, in natural ways, natural language ways. And the same is true for the Stanford example. We're talking about Teams, we're talking about Word, talking about Excel, and they go get the information they need and deliver it in a much faster way. That's tremendous efficiency for that company. And you were buried by data before, and now you're getting tremendous value from the data now. And so maybe a couple of examples there for you from Foundry to Copilot Studio about how customers are really engaging with the AI stack in ways that, again, real value to real people and then real efficiency and productivity, which is great for the employees. They toil less to get the job done. And it saves them money, too, in the process. It's a good cost story. So it's a win-win-win. And Microsoft is pretty differentiated when you think about our experience in this space, the stack we have, the relationships we have with leading model makers as well that customers are betting on regularly.

Great. Great. Good. That's super interesting perspective. Maybe we could pivot to the comments that Satya has made around AI and core. They're very interrelated. You almost can't separate the 2. One brings in the other. Would love to get your perspective on how that is. Is AI kind of the tip of the spear for some of these bigger multi-cloud projects you're seeing? And just would love to get your perspective on kind of the interplay between AI and the balance that you're talking about between GPU-based AI and then the traditional CPU-based data center business that you're in.

Yes. It's a really great question. I mean AI is not AI without the cloud, and I'm not just talking about plugging in GPUs here, right? AI is just a part of an application stack, and the application stack sits on the cloud. And that's for any function of what you're doing. You might be in model delivery and you might have an application and you have end users that are querying that app and getting prompts and getting value back or you might be in the process of doing fine-tuning or any other type of Frontier model creation or just using another AI service out of the box. And to do all that, you need the cloud. And you need compute, you need storage, you need networking. Scott Guthrie recently at our Build event, I think gave a great example about how we're working with OpenAI and what they buy from us in addition to GPUs. When you think about what they're doing, the scale is extremely large. I mean, obviously, the fastest-growing consumer app in history. And so that proposes -- that poses some real challenges when you think about cloud delivery that they have to work through. And what they're using today is way different than what they were using 2 years ago, but they need a scalable data warehouse that will do many petabytes of data with trillions of transactions. And the service they use to do that is Cosmos DB. That's what they're choosing to do. To do the CPU work that they have to do to do any type of scenario from app serving to agentic scenarios, they need VMs. They have to run VMs to do that. And in fact, they need over 10 million cores to do it. And that's a lot. And so like they chose Azure Kubernetes Service to do that. And it's just -- it's a tremendous, again, evolutionary thing. It was way different a couple of years ago. It's different today, and it's different tomorrow. And the great thing about betting on Microsoft for customers, not just OpenAI. I'm not trying to just name-drop there. This is true for all the customers, but I think the scale matters. An example is that as their needs change, Microsoft's breadth of capabilities and the services we offer, the geo locations we're in for our data center footprint around the globe mean that they can adapt and grow with Microsoft, and we can continue to deliver a different value as they are at different parts of their journey with AI. And that's one of the reasons why, again, customers are choosing us. And they see us, again, taking the value we learn from Frontier work and then being able to package that up and deliver that value to other customers and then also increasingly use it ourselves.

Right. Wonderful. Speaking of suites, I wanted to ask about Fabric. It's a topic that we hear a lot from the channel and seeing some real traction. I think you cited on the call, 21,000 paid customers, up 80% year-over-year. So real traction there. Can you just take a step back for us, kind of explain what is Fabric and why are you seeing this kind of success in adoption?

Yes, yes. Thanks for the question. I think that the 21,000, 80% number is a fantastic number in and of itself. But I think it's even more impressive when you think about this is the fastest analytics service Microsoft has ever developed. And it's only been GA for 18 months. And so when you think about the growth that it has over that fast of a horizon, there's something special there that we're really proud of. And what CIOs care about, they have a tremendous job. Like in the moment that we're in, to get value, data is your greatest asset. And your ability to get value of that asset is your ability to mine it, manage it and do all the things you need to do. And you can go get a whole host of point solutions. You can stitch them together, and CIOs can spend time. Instead of being the CIO, they can be chief data integration officers and do data labeling and management and stitching and moving data around or they can come to Microsoft, and they can buy a fully spec-ed out SaaS service that has proprietary Microsoft capability plus any open source solution that you want. It's got partnerships with Snowflake. It's got partnerships with Databricks. It has all of that. And you can get that on your terms in a very, very seamless and fast way. And it's fully integrated, again, from engineering all the way to visualization with our great Power BI products. And so it's [ hunting ]. The folks who don't want to spend lots of time toiling in the trenches appreciate the time to value that comes from Fabric. And so we've been really pleased with what we're seeing.

Wonderful. Great. Why don't we shift to Microsoft 365, the commercial business? Wanted to get your thoughts on just the balance of growth between subscribers and ASP. What are -- I guess, just qualitatively, we could start, what are some of those key drivers in both of those sides of the growth equation?

Yes, sure, absolutely. So we shared 15% constant currency net revenue growth. We shared a 7% seat number, pretty simple math. You can impute a ARPU growth of 8%. And let me unpack that a little bit. So we have -- in that business, we have over 430 million subscribers is the number we've shared in the past. And so obviously, with that type of penetration, seat growth is going to moderate in the future. We're still doing well in SMB. We're still doing well -- sorry, small and medium business. We're also doing well with our frontline workers. And so we will continue to see seat growth. But more and more, as you even see in the numbers I just shared, it's an ARPU game. We're delivering new value in E5. More and more customers are adopting it and stepping up to it. And especially when they think about the full breadth of what Microsoft offers in terms of security, there's a good story there. And so we see lots of good step-ups from -- up to the most premium suites that we offer. Copilot is another driver there. We expect to see lots of growth over time from Copilot. And it's early days, but we're doing well there. And we're going to be continually delivering new value to those suites over time. And like when you think about -- we just recently rolled out Copilot Chat to all of our users to keep -- let them get the value of AI and all the suites from E3 and above is another example of Microsoft delivering new value to the suites over time. And customers will continue to renew. They'll continue to true up and step up, and we think that will drive ARPU over the long run.

Excellent. Excellent. While we're on the topic of Copilot, would love to get your perspective on kind of where we're at in that adoption cycle. Where have you seen success? And going forward, how are you thinking about the incremental attach rate or user for Copilot?

Yes. Yes. Look, we are seeing good traction. We have hundreds of thousands of customers up 3x year-over-year. We are doing larger and larger deal sizes with Copilot every day. And so we see good signals. It's moderate growth. It's pacing well. It's the fastest suite we've ever launched, which is great. And we're increasingly learning how to deliver new value, how to deliver it in ways that customers want and can use easily. I'll use an example of like some new value we've delivered in the suite. We just launched Researcher and Analyst, and they're fantastic products. And they're products that are deeply integrated and differentiated from Microsoft because they're sitting on top of our graph, and they're sitting on top of the customers' data that's stored within SharePoint and other locations in Microsoft services. And like I -- I use Researcher all the time. I recently had a meeting that -- it was a strategy meeting. It was over a project/product that had gone through many strategic revisions over the years. And we are coming in for yet another strategy review. And I'm going to spend lots of time combing through all my e-mails, all my documents to remind myself of the fact set and what we were thinking at the time we made a different strategy decision a year ago or things like that. And then I'd spend a bunch of time doing research externally and say, what are my -- what's my competition doing? And so instead I just outsourced it to Researcher. Asked it to pull it together, summarize it, give me some really nice tables and do a comprehensive analysis much deeper than what you get from a deep thinking kind of standard Copilot Chat. Came back minutes later, it was fantastic. I mean it was so simple. I saved so much time doing that. And then I sent it out to all my friends who were going into the same meeting. And then we entered the meeting with shared context, and we were able to get to decision-making faster. And Satya does this all the time. We have meetings every Friday with an SLT and some of them earlier in the week, too. And we'll regularly come into meetings and Satya will go to start jumping into like a PowerPoint or a Word doc or whatever. And Satya will just blast out on Teams like check out this Researcher I did over the weekend on this topic. And so we all pause and then we read the Researcher and then we come together. And again, we get to the point of having decision-making of reasoning over what we're going to do as a company with that shared understanding and that shared context. And it's just so good. It saves so much time in big meetings that matter, like the ones I'm talking about. And so I'm a big Researcher fan.

That's great. That's great. I guess now that you're in -- it's been over a year, 1.5 years into the Copilot cycle, what have been some of the learnings? You mentioned some learnings that -- what's gone well in kind of the go-to-market and how you positioned it? And what are some of the areas where you think, well, we can -- we're figuring this out, and we're working on driving the adoption rate higher?

Yes. Like years ago, when Satya became the CEO, he really pushed this kind of like growth mindset, learn at all culture and changed within Microsoft the way that we think about learning and growing. And that shows up in everything from the way we work together as employees to the way we engage with customers and the way we build product. And so I think that we were capitalizing on an asset that was built years ago in a time when it matters most, when we're willing to go and put features out, ship those features, let customers use them and then just deeply wallow in the data and what's happening, how delighted our customers, what are they doing with it. We've changed the way that the prompts show up, the way the boxes show up, the way the shading looks on -- again, we've already shipped it all, but we're in there tinkering. We're trying to figure out exactly how we can take that human diffusion problem with new technology and accelerate it as fast as we possibly can. So we've made thousands and thousands and thousands of changes to the product in real time as we see customers engage with it. And so that's the beauty of what we're doing is we get to take that learning mindset and take it and turn it into a capability that becomes an asset as we go deliver new value over time. And so we're definitely not done. We'll continue to make changes, but [ we make ] informed changes. We spend so much time every week as a senior leadership team wallowing in the metrics, what is the user data telling us? What are the sessions telling us? And how should we think about tweaking products to give? Again, being very customer-centric and getting the technology in their hands to deliver value is our prime focus. And so we expect -- we have learned a lot. We expect to learn more, and we'll keep iterating our way to it.

Wonderful. Wonderful. Great. Why don't we shift to security? We hear quarter after quarter when we do our calls with some of your partners, that security is one of the #1 use cases that are driving workloads to Azure and across the business. So security is such a broad thing. Microsoft has such a robust stack in security. So maybe just if you could just start with what are -- if you could unpack a little bit, what is the suite and what's driving the success?

Yes. So we've shared that we're like 1.4 million customers and like 900,000 of those are using 4 or more services, and that's up significantly year-on-year. And as we look at like what those customers are doing and you look at that number of services going up, is what we see customers doing is saying, look, I've got a broad security set of threats that I got to manage. And it's in applications. It's all over the place. And they can go buy lots of point solutions or they can consolidate to a provider that has broad breadth of signal, broad capability, and can deliver that in a single pane of glass with AI-infused to allow them to do the work they need to do without taking on additional risk and without any seams between the different providers that they would have if they had point solutions. And these seams matter. They really do. We can tell you firsthand. And so Microsoft is, I think, well positioned from a consolidation breadth of services and then ultimately, the trust because when you have share, when you have large signal gathering capability across your clouds, whether that's end user applications or platforms, you have a capability to get signal and to manage threats. And we think we're uniquely positioned, and customers agree, and that's why they're choosing us.

Wonderful. Why don't we shift to CapEx? You're on track to spend over $80 billion this year in CapEx, and you've guided to spending more next year. What's informing your decisions and your thinking on investments in CapEx?

Yes. Look, I mean, we spend CapEx based off of demand is the simplest way to say it. And demand is a thing that changes all the time. I mean I wish 3 years ago, I would have known how much I needed 2 years ago because I would have been in a much better position and had a lot more revenue, but I didn't. And so it's a dynamic space right now. We're responding as quickly as we can to the demand signals we have. But when we think about putting the capital down, we have, look, strong signal in Azure, strong signal with our Copilots and our application stack on what customers want. Their bookings we beat the last 5 quarters in a row. When you look at our RPO balance, it's $315 billion. I mean that's all very strong signal that customers are betting on us. They're betting on us on a multiyear basis. We have strong multiyear contracts that say, hey, we're going to bet on you. And I just think it's really important that you know that we take that really seriously. It's a reputation thing. Customers are saying, we believe in your ability to innovate. We believe in your ability to do business right, to do it in a secure way and to deliver it with high quality. We're betting our company on your ability to have quality uptime and performance. And we take that super seriously. And so we're very grateful that customers do that. But it's -- again, CapEx is really based upon that signal. And we are mindful that things change all the time, and there'll be adjustments up and down, slight tweaks here and there. And we're being -- managing the risk of that capital as well. But it's really demand based, and that's how we think about it.

Sure. Great. As CFO of an organization that is as broad as commercial, how are you thinking about margin across the broader organization? What are some of the puts and takes on how you're thinking about margin going forward?

Yes. We're managing margin very tightly. It's something we have a legacy doing over time within the commercial cloud gross margin that we report out on a regular basis. If you go put a historical plot, you'll see it being managed super well. And -- so we're taking all the learnings and what we did to manage that over many, many years and applying them to AI just like we did before. And to give you a sense of what that looks like day-to-day, each product we have that we go to market with has a target margin that it's eventually going to get to. We build that out deeply with the engineering teams and with the product teams, and we build multiyear, multi-semester plans to go drive down the way that those applications and services utilize the underlying hardware and software of the platform. And so that's one layer of the stack. Then you go and click deeper and you say, okay, great, we have platform teams that are driving utilization of how you bin pack or how you drive up utilization of a server. And we have -- those teams have targets and efficiency plans as well. And then you go into something like how do you adopt the latest silicon, how do you do innovative things. Like, for instance, we have a great -- we talked about Overlake. We have a network card that is allowing us to offload CPU and drive up utilization significantly. We're building our own silicon that we've shared previously. And so at every layer of this stack, we're after efficiency. We're after unit cost. Satya talked about all the time. He talked about it in his script again in his last earnings call about being world-class at driving efficiency from our capital deployments. And we, again, [ manage ] every layer of the stack directly with teams. And then the last thing I would say that we do that I'm quite proud of is we create financial incentives to drive user behavior. And so like if it's better for us, for a team to build to a certain region or a certain type of a VM because we have the ability to do better bin packing or better utilization, we drive financial incentives inside the company to get them to do that. And so that multifaceted approach we've built over time is being applied to AI. And Satya talked about the cost per token going down by more than half. He talked about our power utilization on GPUs being better by 30%. And -- again, this is just the output of all the same thing we've been doing for years in the cloud, applied to the new most expensive scarce resource we have in GPUs.

Makes sense. And while we're on that topic, since you mentioned AI, would love to get your perspective on those workloads and whether you see the path to those getting to the same gross margin level as core cloud in Azure [indiscernible] the puts and takes...

Yes, it's super early days. I mean we had quite a journey on the cloud side originally. We're going to be going on a journey here. But I would just say, I believe we can deliver significant margins there. And there's good tailwinds, I think, that are moving in our favor. And I'm not just talking about GPUs. Satya talked a lot about in his Build talk, if you haven't seen it, I encourage everybody to go look at it. Talked a lot about how much of the benefit, the tailwind in terms of utilization and efficiency is coming from models, is coming from post training, is coming from the way that those application stacks are utilizing the GPU. And in many cases, saying, hey, what can we offload, what can we parallelize into a CPU or another part of the stack. And there's tremendous scaling laws on that side. Beyond just scaling laws, you might look at in terms of model capability, size of training runs, total token sizes and parameter sizes and things of that nature and hardware. There's a lot of capabilities [indiscernible]. So I think there's good secular trends behind us, along with our experience, are reasons to believe that we'll continue to make progress on that. But it will take time, and we'll work through it. And we're not -- so we're not predicting anything. We're not sharing anything in detail on that. But I think that there's good foundational elements of why you should believe that we'll continue to chip away at this over time, like we've done in past cloud endeavors and cloud investments.

Wonderful. One more before I ask the final question, but I wanted to go back to the AI adoption question that we were talking about earlier within Azure. You cited a health care and a financial use case. Are there any industries or segments of the business where you're seeing that uptake higher for some of the AI services within Azure? How would you distinguish between just the business overall as to how the adoption cycle is tracking?

Yes, sure. I mean I think -- again, I'm not trying to just zoom out every time, but like it really is broad-based. I mean I think customers -- our customers understand the value AI can deliver. So they're all -- many, many, most are trying to figure out how exactly they're going to do it. And -- so they're just at different parts of the journey. Some are very sophisticated and they're way down the path of having taken an open source model and completely customized it and done all the post-training work they want to do, and they're doing very sophisticated things with delivery. And some customers are just taking a straight API, and they're trying to figure out how to use that in a way that makes sense for their product. And we're -- again, we have that great breadth. We're with every customer, no matter where they're at, with them on the journey, helping them figure it out. As we have figured it out ourselves, we can then take that learning and share it with them. And so it's a great journey. And it's, again, very broad-based. And it's naturally very diverse because every customer is in a different place in terms of sophistication and modernization.

Makes sense. That makes sense, Matt. Well, maybe just to wrap it up, is there anything that we haven't talked about that you think is important to touch on here?

Yes. No, it's a great question. I would say, for me, I'll say one that maybe you didn't expect. I think the pace, the pace of it all right now is very noteworthy. It's easy to miss. I mean we've shared at Build conferences and other various things Microsoft has done publicly. We've shown dot plots of number of shipped products and number of shipped features, and it just is this massive curve up and to the right. And it just continues. I mean, Microsoft is more aligned moving at a faster pace than ever before. And that takes real leadership, and that takes real determination. And in many cases, it takes real stamina because I can tell you, personally, I feel like I'm just constantly sprinting every day, every week, night and day to go keep pace with what our customers want to do, keep pace with the technology and deliver it. And then we talked about the learning earlier and tweaking and adjusting. And it feels very vigorous inside of Microsoft. It's challenging, but it's so exciting. And like for me, having been at the company for 19 years, the energy is fantastic. And I think we're doing a great job executing at pace. And that's not easy to do when you're Microsoft scale. And I think that's something that we're really proud of and that I feel personally, I get the benefit from. I feel very blessed to be part of it.

Absolutely. We can see in the results. And this has been great, Matt. Thank you so much for joining us. Really enjoyed the conversation.

Yes. I appreciate it. Thank you.

Thanks.