Microsoft Corporation (MSFT) Earnings Call Transcript & Summary

Okay. Good afternoon, everyone, and welcome. I'm Mark Murphy, software analyst with JPMorgan. And it is a great pleasure to be here again with Dave O'Hara, who is EVP and CFO of the Commercial division of Microsoft. So Dave, first off, welcome, and thank you so much for taking the time.

Thank you, Mark. Happy to -- anytime anybody offers me a trip to Boston, I'm here.

Yes. I know you recently ran the Boston Marathon.

Did. One of the crazier things I've done, but yes.

Great to have you back here.

Maybe you could just briefly introduce yourself and describe some of the teams and some of the products that you oversee within your role and maybe size that business for us.

Sure. So I'm the CFO for the Commercial business. And sometimes people ask me, "What's the Commercial business in Microsoft?" And I say, "Everything is not consumer," which is sort of the easiest way to define it, but it's all the names everyone is familiar with. It's Office, Windows and Azure and SQL and BizApps, all from a commercial perspective. So give or take, it's sort of $150 billion a year.

Okay. Around $150 billion. So let's begin with the topic of generative AI and the copilots. We have, since last year, been very outspoken bulls on this topic. And we've been bulls on Microsoft's leadership within this category of generative AI. We actually referred to your OpenAI investment as potentially some of the best money ever spent. And so we've noticed that we have Amy and Satya saying that they're committed to leading the AI platform wave. Having been with the company for 22 years, what I'm interested in is, how would you describe the level of internal excitement and the magnitude of this opportunity for Microsoft to try to go out and lead the AI era?

Sure. Sure. That's certainly the hot topic with customers, with partners and investors. And so I think the urgency and the passion around the topic are very high. A lot of people come and ask us, "What do you think about AI?" And they're not even sure what they're asking yet. They just want to know what's going on and where is the world going and what do we as a company -- these are customers talking -- what do we, as a company, need to do to get in front of it. So I would say high sense of excitement with customers and partners, high sense of urgency inside Microsoft. It's been a bit of an expanding circle. When we first made the investment in 2019, it was a relatively small group of people involved, and now it just permeates everything that Microsoft is doing and everything that we're working on. So it's a big focus for us, big focus for customers. And I think it's really early but really, really good signs of interest and demand.

So we have been impressed on this topic, Dave, with the product known as the GitHub Copilot. And we started noticing about 18 months ago, we would speak with developers and they would say, "Well, I wrote code all day long, but 40% to 50% of the code was generated by this GitHub Copilot." And the math on that is pretty staggering when you think about the bottleneck out there in terms of there just aren't enough developers in the world. So soon, we're going to have copilot for Security. We're going to have it for Dynamics. We're going to have it for Teams and Word and Excel and PowerPoint. We have been told that the previews for these products are going to blow people's minds by some of the people that have been in the private previews. Where do you think we're going to see the greatest AI value creation among all of these products?

Yes. It's -- I mean, blow their minds is always an interpretive thing, but I do think they're very impressive. And to me, having been there, watching them be developed, the amazing thing to me was the speed with which they were developed. So when we started out with OpenAI, the logical place to start was with Bing, for all the reasons that make sense. And so once the Office team got to building some of the copilot products, just how quickly they came together and how quickly they surfaced in the product and how much direct value add, all was very impressive. And I think certainly from my expectations, exceeded the expectations that we had. So I think the ability to not only tap into OpenAI and all the benefit it brings but to make it so usable and drive productivity increases, it's been impressive with the speed at which that moves. So I think everything is just accelerating, and that's why our goal, as a company, is to not only stay -- is not only be a leader, but to stay a leader. And we -- to do that, we need to just keep moving really, really quickly. And I think you would see that urgency in everything that we do and everything that we build. But I think in terms of where do we see the value most, I think it has a lot of value in -- for something like Bing that's a search engine, but it really has a lot of value for something like Office or Windows or any of the Modern Work products where you can see big productivity increases.

So we definitely see the speed. We're impressed by it. Everyone out here is also wondering how is Microsoft planning to monetize it. I know you had mentioned you'd talk about Microsoft 365 Copilot monetization in the coming months. But just generally speaking, how do you think about pricing and packaging AI kind of up and down the stack?

Sure. At a general level, there's really a few ways to think about it. One is we have many of our products that are sold on a per-seat basis. And so -- and that works in some instances where you have certain functionality that people use all day, every day. Companies would prefer to pay for that as a per seat. There are other products maybe that are more consumption-driven, data being an example, where maybe a per-user construct makes more sense, but maybe a consumption model makes sense. And there are others where you might just have point AI services where people just go out and use it, pay as they go. And they use it for a while, disappear, come back a month later, they use it again. And so I think all of those are opportunities for us as a company. So per user, we'll see that show up in some places. Consumption, we'll see it show up in others. And others will just be services on demand.

Okay. So it's a pretty broad range. The other question we tend to get very frequently from investors is what is this upfront investment profile going to look like? First, you have to train these large language models, right? And then you're going to bear the cost of actually operating them as you embed them into all these products. How should we think about in the short term and then in the longer term the margin impact for Microsoft as you make those investments?

Yes, it's a good question. I think it's important for folks to just remember, like to be successful in building a large language model and to operate a large language model, you need a lot of data center capacity. And so you need a footprint to kind of -- I don't know that I'd say you have to be a hyperscaler, but it helps to be a hyperscaler to already have all that footprint, and we have that. And then after that, it's a matter of making sure that all the machine capacity is available to do all the inferencing and training. And so you -- there's obviously incremental cost to that, and it's not cheap. But -- and then we have to come up with the right monetization model. But I think the important thing is that like in the cloud business, you had to build out that data center footprint first. It took a long time. It was super expensive. What we're doing now is we're adding capacity to those data centers that allows us to serve the LLMs. And so I think that is -- I think it means that the margin opportunity probably comes a little quicker than it would have otherwise in a cloud model.

Comes in quicker.

Probably a little quicker just because you already have the data center footprint there and you're just taking advantage of what exists.

Okay. That's very encouraging to hear. So Dave, a moment ago, you mentioned this great speed at which Microsoft is moving in all these directions. How do you think about weighing the -- you have this benefit of being first to market by moving quickly with ChatGPT and the copilots. You have this opportunity to gain a leadership position. You also have a risk of moving very quickly and putting these things out there in the wild. There could be risk around bias. There could be risk around hallucinations. You have to try to protect a customer's own proprietary data as well, right? How do you think about weighing all of this?

Yes. I mean it's obviously a very important question, Mark. We've been thought leaders since the very beginning on how -- on responsible AI and how we think about managing AI. I mean Sam Altman has been -- has gone on record several times saying, hey, we need to be regulated. You should come in and regulate us. And so I think there's -- I think everybody is going in with an open mind that we need to make sure that we have some guardrails on how we think about growing the business. With regard to hallucinations, et cetera, we're all very aware. We want to eliminate that to the degree possible at all. And so we're very focused on that. And then from a responsible AI perspective, making sure we're managing people's data, making sure we're abiding by our privacy regulations, making sure that we're doing things compliantly, all of that is super important. And we've been -- it's been a focus of ours for years. It's sort of one of those things where this seems like it popped up overnight. But again, we started investing back in 2019. I think we came out with our responsible AI principles a few years ago. And so we've been at it for a while, and we've been trying to drive some thought leadership on how to do this in a responsible way and we'll keep doing that.

So Satya had recently discussed how he believes AI is going to reshape pretty much every single software category. We -- in our work, we came across a comment saying that this would be the next big battleground in IT. Can you maybe help us, as you look out into the future and you think about all the existing AI-enabled products that you've announced, where do you think you're going to see the largest growth opportunity? And then even beyond that, how are you going to evaluate all the future opportunities for building AI into your products?

Well, I think the very first -- to sort of answer that in reverse order, the very first area where we'll go is where our customers can see value. And so we work -- I spend a lot of my time, as part of the Commercial business, talking directly with customers and partners, and you get a sense for what really matters. And so we'll go build in areas that really matter. I also think from a company perspective, one of the very first things we do is we look at the total addressable market and say how big can this market get. And so if we have permission to play, like if it makes sense for us to be there and it's a really big market and we think we can deliver that customer value, that's where we're going to go. I think in terms of how do we think about the most immediate ability to monetize, I'll go back again to, I think, the work that we're doing on our productivity tools in Office and in Modern Work more broadly. I think that's going to have some immediate value. And I also think, just to sort of add on to that, there's a lot of talk about how AI is going to eliminate jobs. I think that what it will do is automate a lot of functions, but it does free up time for those workers to go do other things. Even within our own finance team, when we automate functions, it frees up time for our finance team to go work with customers directly and help them understand what they can do. And I honestly would rather have my finance team out working with customers than doing manual reporting or something on the back end. And so I think there will be productivity gains. It frees up people to go work on other things. To your point about GitHub, there's a shortage of developers. I think it expands our capacity to build better products. So there's a lot of goodness that comes with it.

Yes. Yes. And especially in Office, you have the potential to get this out there in front of -- we could dream the dream of putting this in front of 1 billion people pretty rapidly, right?

Yes. I'm a bit hard to impress. But the day that the Office team came in and did the demos, it's like, hey, that's good stuff. And it happened fast.

It really did. All right. Let's shift gears and talk a bit about getting to Azure and talk about this trend we've seen around optimizations. And I'll frame it this way. We -- in our last Microsoft partner survey, we were asking them for how long do they expect to see an elevated level of cloud optimization activity continue. They came back and said, on average, something like 4 to 5 quarters. And they're basically saying that customers are trying to do more with less, right, in this environment. How does that time frame strike you? Is it a bit pessimistic? Because if you already have a chunk of customers that have kind of done all the optimization that they can do, right, then some of them would be nearing the end of that.

Yes. I think it's natural to say after 4 quarters, you sort of -- the optimization starts to lap itself. But I do think that from an optimization perspective, first of all, we see optimization as a good thing because we -- I've never had a customer come in and say, "I need to cut my budget in half." They're like, "Hey, I would like to do more, but I'm kind of running up against my budget. Is there a way you can optimize some of this so we can spend more over there?" And so generally, if we do it right, the optimization is a growth opportunity even if it causes a little bit of short-term compression. So we see optimization as a good thing, and we do think they're going to reprioritize and keep spending their budgets on cloud. It's just going to be in new areas of the cloud. But I do think -- to say after a year, you start to lap, I think on a per customer basis, that's true. Some customers have been optimizing faster than others. And so there will always be some of it going on.

Okay. So compression but also kind of -- it sets the table for growth opportunity later?

Compression and expansion at the same time.

So on the expansion part of that, Amy had this comment in the most recent earnings call that AI services would contribute one point of Azure's growth in the current quarter, in this June quarter. And so of course, we ran the math on that. It equates to something like a $500 million run rate. But as people pointed out to us, the number of Azure OpenAI customers increased 10x sequentially last quarter. So can we do some combining here and extrapolate that a bit? Is it logical to conclude that AI services are going to become a pretty meaningful boost to Azure growth if we're looking out, let's say, 12 to 24 months? So in other words, could this be creeping up where it's actually contributing 2, 3, 4, 5 points of growth and it is truly moving the needle for Azure?