Microsoft Corporation (MSFT) Earnings Call Transcript & Summary

Microsoft has a very special place in my heart. I was a Microsoft developer in the 90s. And so I said to Takeshi, I wouldn't be here without Microsoft. So I am a little biased because it helped me get up to the main streets in Minnesota. Takeshi, welcome, 27 years at Microsoft. I know you said that there are others that have beat you in terms of being at Microsoft longer, but 27 years, just run us through what you've seen over those 3 decades and maybe lead us into your current role and what you're doing now.

Yes, I'm a bit of an oddball at Microsoft. I'm native Japanese. I grew up in Japan, and worked for the Japanese government as an MITI during trade negotiations before joining Microsoft in the engineering group, of -- what was then Windows NT in 1997. And then I [split] over to marketing in 2003, working on sort of Office as a traditional shrink-wrapped software back then and then worked on sort of transforming that business into a cloud business, starting with Office 365. And then it’s a lot of it. It’s about the cloud and I worked on Azure. I became the commercial CFO a few years ago and then the CMO last October -- no, a year before October. So it's been just over 1.5 years maybe. And they spent a lot of time on cloud, AI and growth.

We'd like to talk about cloud and AI. Let's talk about AI. A few years ago, you really started to push before anyone. You guys got out in front. You led the charge. It's now been a couple of years in. Maybe just bring us up to speed where you're at? What are the next steps? How do we think about -- I know you unveiled a lot of Copilots to build, and we're getting new reasoning models, where Jefferies is very excited. We're going to roll out Copilots to 90% of our firm. We're 6,000 employees. We're tiny, relative to other financial services firms, but we're going to 90%. We're at 10% today. So we're a huge believer.

Well, I mean it's one of those things where you're never satisfied. Certainly, I'm facing a room full of people who is never satisfied with what we do. We need to do better. That's great. And our CFO constantly reminds us of that. Our CEO reminds us of that. I get love letters of that ilk very often. But at the same time...

Is that true that there's a room called the -- what's the room that Amy called the biggest room for improvement? Biggest room in the house is the room for improvement, yes.

But at the same time, having worked on multiple transitions, first, actually from sort of a software license sales back in the office shrink-wrapped days to moving to an annuity sales model and then to a cloud subscription model with the Office 365 and then adding a consumption business model in Azure. Like, having watched these and participating in these transitions over the multiple times with Microsoft, the thing I can say about sort of what's happening in AI is just the speed is so different. It's crazy to think that we talked about Copilot, like it's a very well-established thing, but we only made the product generally available in November of '23. So the whole product has been in market generally available just for 1.5 years plus. And so the pace of both engineering throughput and then sort of the resulting go-to-market adjustments have been high. And unlike some of the transitions before, like the cloud, where we had the sort of the opportunity to fast follow a bit because there were precedents set before us that we can learn from. We're now in a place where we are leading and we often make decisions that have no precedents before. So we have to make -- be the first one to make a lot of decisions and set the pattern and lead the industry and sort of our customer engagement scenarios and invent new ways to do that. So that's really exciting. But the other thing I can say is the AI work really builds on sort of the strong foundation we built in the cloud. When we were working initially in the cloud transition from on-prem software, we really didn't have a base of cloud, so we have to build that up from basically scratch. Whereas now we have a base of cloud and AI sort of essentially layers on top of it in a nice way, given the fungible infrastructure we have. And so that gives us a lot of scale fairly immediately.

There's a -- I think the industry is trying to grapple with how you price AI. Some have said it's foundational. Some have said it's a Copilot, some have said it's by the work done. There's different ways to do it. How do you think about how customers should pay for AI at Microsoft?

I don't think it's one size fits all, and there are different constituents. For example, developers certainly would like to consume things in a very consumptive way, whether it's in the form of an API or tokens. And so we have that and that's doing really well on the Azure side with the variety of AI services we offer and the application platform. We think of -- as what we call AI Foundry as effectively an app server for developers to be able to build applications quickly and efficiently and manage it. But on the SaaS side, like Copilot, there are subscriptions that we have and we also are giving customers options to basically say, "Hey, you have a -- sort of a population in your employee base for whom paying for a subscription makes total sense." But then you have all sorts of frontline workers or broader constituents that are in your enterprise that you still like to provide AI capabilities in their daily lives for, but not necessarily ready to buy a sort of a monthly subscription for every one of them. So we sort of opened up new options with business models like Copilot Chat that essentially enables customers to consume even sort of SaaS AI services on a consumptive business model. So we're sort of giving choices across the board, and we're continuing to learn.

Last week at Build, you had a lot of announcements. Maybe what was most important to you that you unveiled last week?

Well, Build is our main developer conference, so the entire context is very, very focused on developers. So we had our CTO, Kevin Scott, talk about sort of the notion of an agentic web and how we're embracing open protocols like MCP and A2A to be able to enable scenarios where agents can work across individuals, teams and organizations to unlock new scenarios. So this notion of new ecosystem and embracing open standard was a key part of that story. But given that it is a developer conference, it was a great opportunity for us to talk about how we are reimagining the software development life cycle. It's actually interesting to remember that the name Copilot came from our first Copilot product, which was GitHub Copilot. GitHub Copilot was our first Copilot product. And it has started with things like code completion. So it's interactive. You're the developer and you're writing half the line and then the remaining line code completion, the AI helps you. But then it's evolved since to be able to do things like multi-file edits and from being a sort of a payer program to now more of a peer programmer where you can assign issues and essentially sort of treat AI as sort of an autonomous thing that can complete tasks on their own. And so the Copilot -- GitHub Copilot has been a great showcase in how the capabilities has evolved over time, and we sort of talked to developers about how that really changes the workflow and make them more productive. And we had announcements like Visual Studio extensions to GitHub Copilot being open sourced, which really generated a lot of excitement amongst developers to be able to sort of use GitHub Copilot and VS Code in more flexible ways. And then I think for us, because we have such an enterprise footprint, we spend a lot of time talking about how we are delivering this ability for developers to build AI applications in a way that meshes with the governance and security and compliance structure that enterprises have. So how does an agent application that gets built on Azure intrinsically support the identity framework that we have in Entra? How does the -- sort of all the logging capabilities that you can use in AI Foundry as you build AI -- as you build your AI applications can compose with our governance, data governance solutions like Purview or integrate with security in terms of our Defender products. And so we sort of paved the -- articulated our point of view on how AI development can be done in a way that's secure and manageable in a way that fits with the enterprise estate.

You get the magic wand and you can wave it over in enterprise to get AI adoption going quicker. I think maybe it's been maybe slower than maybe some of us all thought it would be. What are the -- what do you wave this over? Is it data governance, security in your data estate, right? What do you think the magic wand?

Well, I mean, if it were that simple where you have just one silver bullet, you would be just working on that one silver bullet. It is a net new asset for enterprises to manage, right? It's not like, hey, I used to have this thing. I know a database, I'm replacing a database, upgrading a database, migrating something better way. Those are all relatively known things. In many aspects, AI is new for companies to adopt. So it's not just about the product. It's about sort of the policies that they now have to think through. What are the business workflows within the company, organizational structures, skill set. So all those sort of things that are actually not necessarily the tech, but more about skilling, more about organizations, more about how people within the customers collaborate in different ways. That is the typical enterprise adoption cycle for net new technologies versus linearly sort of upgrading things you already know. So that accelerating, that's of course, the biggest lever, frankly, beyond any particular given tech per se.

On the last earnings call, Amy pointed out that strength in core drove a lot of the upside. Everyone kind of looks around and says like what's happening in the core to accelerate growth now at Azure?

Well, the -- I think for financial communities, we talk about AI and non-AI somewhat separately. That is not how we engage with a customer. The customer conversation isn't like let's have an AI conversation and a core conversation. That's not how we do go-to-market because every migration project that from a financial community standpoint might seem core is already always connected with what are we going to do once we migrate and modernize and we want to use. We're doing this because we want to be able to apply AI on top. So for example, just as an example, we talked about AI Foundry with a great customer momentum with 70,000 customers. But -- and that counts in sort of the AI growth. But Fabric, which is what our data platform solution is growing basically 80% year-over-year. And I think the customer usage of OneLake, which is where you would store your data for Fabric, I think it is growing like 6x year-over-year. And that's all in furtherance of customers getting ready to basically apply AI solutions on top of the data estate they have, but that counts as core. But that's the same customer engagement in the same meeting with the customer, we're talking about AI Foundry and Fabric together. And so we need to get this balance right in terms of driving growth on both sides. And certainly, Q2, when Amy calls out execution issues, we don't take this lightly. And the Shockwave went through the entirety of the company.

I can imagine.

We heard it, and we sort of worked hard.

Yes. It was interesting because I had a call back with Amazon post third quarter, like what are we going to stay off of that Microsoft print, they blew it out. Even Amazon was like, wow, like what -- like I know there are multiple things that went right, but what's at a high level going really well for Azure right now?

Well, right now, I think it's -- the execution forecast has really helped us, particularly amongst our -- what we would call sort of the large enterprise customers, maybe the top 500 customers that were balancing the conversations about net new projects in AI that might actually consume a lot of engagement cycles, but not necessarily immediate consumption versus migration projects that might actually not consume as many sort of customer engagement and meeting minutes, but actually drives lots of consumption. How do we actually balance the execution in the face of the customer to make sure we're having the right conversation on the AI side, but also driving things that can drive consumption sort of we're getting that balance, I think, better calibrated. Also our scale motion, sort of the smaller enterprises and mid-market and below, we made some internal changes that's still sort of working through the system, and that's starting to show improvement, although we have more to do there, too.

Macro has been on everyone's mind. Maybe you don't hear it as much as we do in our community, but it doesn't seem like it's having any issue or any big concern in your execution.

Well, of course, no one is immune from macro. But for us, as far as we talked about in our Q3 results, too, we haven't seen a lot of impact from that. And the good news is, I think the customer spend in this area, digital transformation, cloud migration, security, those are some of the most defensible spends, I think, that customers would want to prioritize. And so I think we're in a good spot. Not to say we're immune, but we feel pretty good right now.

My only criticism is we got to get Copilot to tell better jokes. Why did Microsoft employees bring a ladder to work? Because they heard the company was going to the cloud. That was from Copilot. We're going to work on.

I'll tell the...

Okay. Sorry. There's just a side injection there. I know there are questions from the audience. If you guys have a question, you can raise your hand, you can jump in. Any big open questions for Microsoft?

[Technical Difficulty]

If I understood the question correctly, the question was about how we think about data center capacity for Azure?

[Technical Difficulty]

And the investment on capital? Like I would say we're basically building to demand. We're very sensitive to demand signals. And so we want to make sure that we're basically building to meet the demands that we see. And so that's how we think about it.

[Technical Difficulty] there's obviously anytime you guys pull back or adjust those plans [Technical Difficulty] how nimble can you be with those CapEx plans [Technical Difficulty] when there are changes in demand?

Well, I wouldn't claim to be a super expert on it. But basically, there are long-term assets, right? A lot of the CapEx spend is for assets that are going to be sort of utilized over a long period of time, 10, 15 years. And so some of those are somewhat lumpy and there's not a lot of near-term variability you have. But then the kit that you installed in these data centers, the server racks, the CPUs and the GPUs that we install can be as close to the demand signals as we can. But then the investment from an aggregate volume standpoint encompasses both those kind of server and kits versus the long-term investment is kind of a blended number. And so that's all I can say.

I know you love all your Copilots equal, but is there one category where you're excited maybe usage isn't as high today and you think it could explode? Maybe one thing has come up maybe in security that was a little later in terms of getting that out, is that, could it be security, could be another category? What are you most excited about in that family?

Well, for me, the exciting thing about Copilot is the fact that it actually gives us an opportunity to reinvigorate every one of our franchises because a little bit like the Internet, AI and Copilot type use cases aren't unique to one category. We basically get to reinvent productivity with Microsoft 365 Copilot. We get to reinvent security with Security Copilot. We get to reinvent coding and productivity for developers with GitHub Copilot. So for us, Copilot isn't one thing, it's actually something that actually can reinvigorate all of our product franchises, inclusive of all of our major franchises, Windows, Microsoft 365, all the others. So that's what's exciting to me as a CMO of the company.

Other questions?

[Technical Difficulty]

Well, I would say the AI adoption by customers, of course, varies for each customer. But I would say there's definitely be -- a distinct phase shift where it's not about POCs. We're really seeing large-scale deployments, particularly when it's targeted against known business processes. So it really boils down to not sort of an ROI in the broader sense, but against a given KPI. Let's talk about customer service or let's talk about HR processes, or let's talk about essentially, very operational things like, invoice reconciliation or those kind of things and where customers are saying, "Hey, I can target this. I have a known KPI and I can develop and deploy a solution against it and get -- see returns." And that creates the next project, the next project cycle where it's way more than a POC and much more into production deployments is what we're seeing.

The whole concept of data governance is a big theme for getting ready for AI. What are you doing to help companies get ready on that side? Maybe talk through the products, the strategy, and the adoption of.

Well, actually, our data governance offering is called Purview. And certainly, if I think about even the search traffic volumes that we get on customers searching for Purview, it's actually dramatically increasing. And so the customer interest is not just AI, but how do I govern data associated with the introduction of AI technologies like Copilot is very, very high. And so I would say interest in products like Purview is very high. The other thing I would say is we have been quite responsive to use the customer engagement and the feedback we get to actually add some basic data governance capability into offerings like Microsoft 365 Copilot in the base. So you don't have to buy another product. Like as an example, Copilot essentially only accesses things that you can access. But then oftentimes in an enterprise, there are lots of things that anybody can access, but you didn't know that everybody could access, but Copilot would efficiently find them. So we've actually enabled scenarios like Advanced Access Analytics so that you can say, hey, who can access this file and how do you govern it and actually making that part of the Copilot offering so that IT, doesn't have to buy another product as they sort of go deploy Copilot. So those are changes we made in the last 12 months based on customer feedback. So we're sort of improving the base, what you get with Copilot, but also having a sort of a Purview solution that can really look at your data state across the board, and then that's having a lot of customer interest.

Our CIO of Jefferies went to a Microsoft briefing and he came back and I didn't think he was going to say this, but he said, the one thing I was like blown away by was Power Apps and what we can do in an age of AI to kind of create custom apps for Jefferies that are -- that's different for us than others. And maybe if you can just expand on Power Apps, what it means. Some may -- not understand what it is, but what, yes, explain.

Yes. Well, basically, like we often talk about it as a sort of low-code/no-code solution, and it might mean very little to people in this room. But the way I think about it is, if you think back in the early days of productivity, almost early days of Office, document creation used to be a back-office operation and if you have to respond to an RFP, there was a back-office operation process to be able to respond and have a full faith and full-throated response to an RFP. But then with the Office and Office productivity applications becoming ubiquitous, doing things like RFP responses or doing financial analysis become much more of a frontline operation. Power Apps basically is doing that, exactly the same for app creation, just like you can create a document. Basically, it doesn't have to become an IT project for something to -- for somebody to build an app and essentially fill that last mile of application needs, because otherwise, the IT backlog per application is becoming longer, not shorter. So being able to empower the frontline office people to be able to essentially empower themselves, with applications they can create themselves is what Power Apps is doing, and it's growing very, very rapidly.

So do you envision, you sit in front of the -- your system and you say, I'd like this customer [training] app built and here are the characteristics, you input it and it builds the application for you. Do you envision a world where...

Yes, yes. I mean, like if you actually look at some of our build demos with Power Apps, it's doing exactly something like that. You literally tell Power Apps what kind of apps you want and Power Apps will help you build it and actually create a mobile app and a web app that you can then deploy or share just like a document. I [would make] one, and I'll share it with you in a sort of a secure compliant way.

One question on the back.

[Technical Difficulty]

I can't speak to what we expect on the sort of the device sales impact, but we're certainly taking advantage of the fact that we have a deep history of innovation on not just the cloud, but software that you can deploy to on-premises and edge devices. And so one of the announces we made actually at Build was this notion of Foundry Local. So then you basically can train your models in Azure, let's say, with that AI Foundry, and you can deploy it as a container to different local devices. And then we also announced Windows AI Foundry, which is essentially a way to have a layer in your Windows device that can actually let developers build applications really quickly using local AI. So we are definitely looking to continue to invest in this notion of AI, both in the cloud, but on the edge and having those 2 work really well together.

One more question over here.

[Technical Difficulty]

Well, the competition, I think, at AWS or anybody else, I think, happens in a very project-by-project basis, I think. So whether you can talk about data, you can talk about infra. But for us, I think the -- our strength really lies in our ability to serve the enterprise needs holistically. It's across all the stack from infra to data, all the way to SaaS layers that integrates with everything we offer with Microsoft 365 and Dynamics 365.

Great. Thank you so much for joining. Appreciate your time.