Company Management
Amy Hood - EVP & CFO |
Brett Iversen - General Manager, IR |
Satya Nadella - Chairman & CEO |
Analysts
Aleksandr Zukin - Wolfe Research |
Brent Thill - Jefferies |
Karl Keirstead - UBS |
Kasthuri Rangan - Goldman Sachs Group |
Keith Weiss - Morgan Stanley |
Kirk Materne - Evercore ISI |
Mark Moerdler - Sanford C. Bernstein & Co. |
Mark Murphy - JPMorgan Chase & Co. |
Keith Weiss - Morgan Stanley - Analyst
A very nice end to a great fiscal year. Satya, you started your comments talking about how every customer conversation has the customer asking you about utilizing generative AI technology and how fast they could utilize that generative AI technology. What's the answer? What do you tell them in terms of the pace with which that could get into the marketplace and your customers can start using it?
And then for Amy, how should investors think about just the fundamental gross margins behind these generative AI technologies? We understand there's going to be a lot of CapEx to ramp up underneath these. But what should we expect in terms of what the ultimate gross margin looks like underneath all these new generative AI solutions?
Satya Nadella - CEO
Thank you, Keith, for the question. The fundamental guidance and conversation that we have with customers is twofold. One is the easiest path to value out of generative AI is to adopt certain solutions, for example, GitHub Copilot. In some sense, it's sort of a no-brainer to add productivity leverage for all of the software developers in any organization. Whether you're a bank, you're a retailer or you're a software company, it applies to everyone. So that's probably one of the things that we have seen very good -- even productivity data and great adoption.
And then obviously, the excitement that there is already around the M365 Copilot. So first thing we sort of talk about is how we ourselves are deploying all these Copilots across, whether it's Sales Copilot or M365 Copilot or GitHub Copilot, how do you get maximum value out of these horizontal tool chain? And then on top of that, we have taken what we did underneath these products and built it out as a first-class tech stack, right, which we talked at our developer conference, called the Copilot Stack, and then with Azure AI tooling, made it possible for a someone like Moody's to build their own Copilot for their people.
So to us, we want to be able to help customers build their generative AI applications on top of Azure AI, and with speed, if you will. And so those are the 2 things that we ask them to identify where they can get the maximum productivity leverage. And then we even swung with our own resources to help them get those things done. And the last comment I'd make is the cloud and data in the cloud enables all this because I think the diffusion cycle here, in some sense, we have a new set of cloud meters that are getting adopted faster because of everything else that came before it in the cloud. So those would be the observations.
Amy Hood - CFO
And to your question, Keith, on gross margins and how I think about those going forward, the first thing I would say is that I expect gross margins here to transition over time, just like they did in the prior cloud transitions. I would also say I expect workloads and the gross margins of the workloads to be different, just like they are in the cloud today.
I would also add one thing that's different than last time, we talked a bit about this before, is that we start out in a different place with more of a shared platform, which allows us to scale those gross margins a bit faster than last time. And we do expect, as you asked and Satya talked about, the pace of this adoption curve, we do expect to be faster. So you're seeing the CapEx spend accelerate in Q4 and then again in Q1, and we've talked about what it should look like the rest of the year.
Now that being said, we're talking about all that and going through that transition while delivering in FY '24 over FY '23, effectively 1 point higher operating margins. Because if it's flat year-over-year, as we guided, with the headwind from the useful life change, when you correct for that, it's about 1 point higher. So I think the real focus here is being able to be aggressive in meeting the demand curve and focusing on the transition and growth in gross margins and delivering the operating leverage.
Brent Thill - Jefferies - Analyst
Satya, on the optimization headwinds that you continue to see, when do you think we hit peak optimization? Are we getting close to hitting that peak and getting some relief in the back half of the year and maybe AI helping provide us a tailwind? Any color from what you're seeing from your perspective would be helpful.
Satya Nadella - CEO
Sure, Brent. Thank you for the question. Yes, a couple of observations. One is, I think, overall in the cloud, you do see new project starts and then those project starts get optimized and then you sort of time-series all of that, and that's sort of what you see in the normal course. What happened here was during the pandemic, obviously, there were lots of new project starts and optimization in some sense was postponed, and that's where you're seeing, I'll call it, catch-up optimization. And that's something that, to your point, we will lap. Going into the next couple of quarters, I think, it will come down.
And we are seeing new project starts, both traditional type of project starts, even cloud migrations, data applications and, of course, obviously, the AI applications. But we'll get back to I'll call the normal pace of new project starts and optimizations going forward, that we will cycle through, I think, in the next couple of quarters, what is the last catch-up optimization.
Amy Hood - CFO
I would just add, Brent, I think to Satya's point and maybe to build a bit of a line for you, I think it felt very similar to last quarter where we made the same comments, which is we're seeing sort of the normal optimization plus we're seeing new workload starts across these workloads Satya talked about. And I think that's what we're saying going forward and really what the change is just that lapping of, I think, a bit of a catch-up from a year ago. And you're right, we'll continue to do that through H2.
Mark Moerdler - Sanford C. Bernstein & Co. - Analyst
And congrats on the quarter. Amy, CapEx moved up significantly Q-over-Q and year-over-year and it's increasing moving forward. Can you give us some color? Is it physical data centers? Is it predominantly servers? Is it predominantly AI-driven? How should we think about the useful life of it? And then quickly for Satya, can you give us some -- status on the general availability of the full Copilot development stack? And how long it's taking clients and partners to build Copilots?
Amy Hood - CFO
Why don't I start on the CapEx question, Satya, then I'll turn it over to you. Mark, really, first of all, both in Q4 and then talking about Q1, the acceleration is really quite broad. It's both on -- both the data centers and a physical basis plus CPUs and GPUs and networking equipment, think of it in a broad sense as opposed to a narrow sense. So it's overall increases of acceleration of overall capacity.
And I think if you look back over really FY '23, you wouldn't have seen some of the pace on normal, what I would say, capacity adds, even for the normal Azure workload. So you're seeing both accelerations, the normal Azure workloads plus some of the AI workloads, is partially the reason. So it's why I do comment quite often that it's both overall Commercial Cloud demand and building out capacity for AI. It's both.
Satya Nadella - CEO
Yes. And I think just for perspective, I think it's sort of always good to think about it, right, where we have, what, $111 billion commercial cloud business growing at, what, 22% year-over-year. And then you had a CapEx growth, which is around the same number, 23%, 24%. So in some sense, it's sort of replacement capital plus some new capital that is going to drive new growth. So that's, I think, the scale. And we feel good about that structure of overall growth rates and how it translates into future TAM opportunity for us.
And then to your other question on how all this translates into project starts effectively, the Copilot stack is available today on Azure. So we have everything from Azure AI tool chain where you can use obviously, Azure OpenAI or even you can use open models from Llama and Hugging Face models. You have all the Fabric and all of our operational data stores for what is one of the most useful patterns around generative AI as what is called Retrieval Augmented Generation, which is you take the data that you have in the data stores, use it in a prompt to generate completions, summaries, what have you. And so that's something that we've seen a lot of and the Copilots are fundamentally orchestrations of that. And so we have these services available.
The thing that's fascinating is when you use something like Power Virtual Agent, you have a low-code, no-code tool to build effectively these AI products or full-fledged Copilots like we've built. And all the underlying primitives for that are available on Azure. The tool chain is available on Azure and the speed with which customers are able to deploy them, ISVs are able to build them, is pretty impressive.
Kasthuri Rangan - Goldman Sachs - Analyst
Congratulations on the quarter. If I could, I just wanted to get your thoughts to shift the discussion away from COGS and the CapEx to more of the top line outlook. It looks like Azure growth rate is definitely starting to stabilize and generative AI contribution to Azure is measurably improving quarter-over-quarter, and optimization in a broader sense is also starting to settle down.
Where does this leave with the company's outlook for Azure growth rate in the future quarters? Are we at a point where we've bottomed out, and we could start to see some acceleration due to the trends we discussed? And also if we take the superset of Microsoft Cloud, when you throw in the new pricing for Copilot, it certainly looks like the TAMs are opening up in a pretty significant way. So when you take the broader lens, that 21%, 22% growth rate that Satya and Amy referred to, what could be the outlook? Could we be too optimistic in entertaining hopes of some kind of acceleration in the years ahead? Or how do you think about the outlook on the top line?
Satya Nadella - CEO
Sure. Kash, thanks for the question. So maybe I'll start and then, Amy, you can add because I think -- we do think about what's the long-term TAM here, right? I mean this is -- you've heard me talk about this as a percentage of GDP, what's going to be tech spend? If you believe that, let's say, the 5% of GDP is going to go to 10% of GDP, maybe that gets accelerated because of the AI wave. Then the question is how much of that goes to the various parts of our Commercial Cloud and then how competitive are we in each layer, right?
So if you sort of break it down, you sort of talked about how Microsoft 365, we think of this Copilot as a third pillar, right? We had the creation tools. We then had all the communication and collaboration services, and we think the AI Copilot is a third pillar. So we are excited about it. Amy talked about how we want to get it out first and part of this preview. And then in the second half of the next fiscal year, we'll start getting some of the real revenue signal from it. So we're looking forward to it.
But we think of it long term as a third pillar, like we thought about something like, say, Teams or SharePoint back in the day, or what have you. Then Azure, the way I think about it is we still are, whatever, you're inning 2 or inning 3 of even the cloud migration, especially if you view it, right, whether by industry moves to the cloud, segment move to the cloud as well as country adoption of the cloud, right? So there's still early innings of the cloud migration itself. So there's a lot there still.
And then on top of that, there's this complete new world of AI driving a set of new workloads. And so we think of that, again, being pretty expansive from a TAM opportunity and we'll play it out. But at the same time, we are a $111 billion commercial cloud that has grown in 20s, and so therefore, we do hit law of large numbers. But that said, we do think that this is a business that can have sustained high growth, which is something that we are excited about.
Amy Hood - CFO
And I think the only thing, Kash, I would add is, I think in some ways, what we're really pointing to is there's a process here. We see the demand signal is quite strong. It remains strong. I'm thrilled with all the product announcements we've made. I'm thrilled with them moving to paid preview and then moving to GA. They absolutely are expansive in terms of addressable market. They reach new budget pools is almost the way I talk about it a lot in terms of how CIOs or CFOs that I talk to think about that investment. So a growing opportunity.
And as you know, we're focused on executing against that. And then revenue is an outcome. But it certainly does require -- the demand signal requires the capital expense and then creates the opportunity. And that's why I think, in some ways, we're spending a little more time talking about some of that investment is because it is the demand signal.
Karl Keirstead - UBS - Analyst
Okay, great. Amy, if I could double-click a little bit on the exciting news around M365 Copilot as everybody on the line looks to layer that opportunity into our models, I just wanted to get your views. Are there any guardrails you'd offer us to sort of keep us in line? Is there a degree of gross margin pressure in the Office segment? In other words, is it a fairly cost-intensive new product that we should keep in mind? And also, could it pull along Azure in the sense that you need Azure AD and perhaps some of the other cybersecurity products? So a little color there might help everybody with their modeling exercise tonight and in the coming weeks.
Amy Hood - CFO
Thanks, Karl. I think maybe I'll first start with the process we have when we release new products. And I absolutely understand we are excited, too, by the demand signal, the customer reaction, really the requests we're getting to be in the paid preview. It's all encouraging. As you know, we've -- last week, we announced pricing, then we'll continue to work through the paid preview process get good feedback. Then we'll announce the general availability date, then we'll get to the GA date. Then we'll, of course, be able to sell it and then recognize revenue.
And that is why I continue to say that I am just as excited as everyone else about this, and it should be more H2 weighted. And we've, I think, given you some sizing opportunities. And I think I would use all that. But I do think this is really about pacing. And of course, we've still got to get our Security Copilot and some of the Dynamics workloads priced and released. And we'll continue to work toward that.
And of course, I think one of the things that people often, I think, overlook is, and Satya mentioned it briefly when you go back to the pull on Azure, I think in many ways, lots of these AI products pull along Azure because it's not just the AI solution services that you need to build an app. And so it's less about Microsoft 365 pulling it along or any one Copilot. It's that when you're building these, it requires data and it requires the AI services. So you'll see them pull both core Azure and AI Azure along with them. And I think that's an important nuance as well.
Satya Nadella - CEO
Yes. If I could just add to what Amy said, the platform effect here is really all about the extensibility of the Copilots. You see that today when people build applications in Teams that are built on Power Apps and those Power Apps happen to use something like SQL DB on Azure. That's like a classic line of business extension. So you'll see the same thing. When I have a Copilot plug-in, that plug-in uses Azure AI, Azure meters, Azure data sources, Azure semantic search. So you'll see, obviously, a pull through not only on the identity or security layer, but in the core PaaS services of Azure plus the Copilot extensibility in M365.
Mark Murphy - JPMorgan - Analyst
Satya, there's so much evidence now that GitHub Copilot is boosting developer productivity by 40% to 50% or more, and it's resulting in higher quality code. Do you envision a similar level of productivity boost for the Microsoft 365 Copilots or the Security Copilot or the Sales Copilot? In other words, can every room in the house be remodeled to a similar extent such that, that value proposition is pretty elevated across the entire stack?
Satya Nadella - CEO
Yes. Judson Althoff would love you for having used his metaphor of remodeling every room of the house with AI because you're absolutely right. I mean that's the opportunity we see. I think what you're also referencing is now there's good empirical evidence and data around the GitHub Copilot and the productivity stats around it. And we're actively working on that for M365 Copilot, also for things like the role-based ones like Sales Copilot or Service Copilot. We see these business processes having very high productivity gains. And so yes, over the course of the year, we will have all of that evidence.
And I think at the end of the day, as Amy referenced, every CFO and CIO is also going to take a look at this. I do think for the first time -- or rather, I do think people are going to look at how can they complement their OpEx spend with essentially these Copilots in order to drive more efficiency and, quite frankly, even reduce the burden and drudgery of work on their OpEx and their people and so on. So therefore, I think you're going to see all of that translated into productivity stats, and we're looking forward to getting that data out.
Aleksandr Zukin - Wolfe Research - Analyst
I guess maybe just a multi-parter. You mentioned a couple of times that with the AI workload adoption that you're seeing on Azure, it's starting to look maybe a little bit different from an incremental share gain perspective versus previous generations. Can you maybe expand upon that? How should that drive for Azure consumption, particularly as we get through the year? And do you see a scenario where either the combination of lapping the optimization headwind, plus the AI contribution, plus this incremental tailwind that you're seeing around the workloads actually does drive a reacceleration in Azure, particularly in the second half when you're going to start to see some of those things kick in?
Satya Nadella - CEO
Yes. I mean the thing that we are both seeing and excited about is both the new workloads. I mean if you think about Azure, we have grown Azure over the years coming from behind. And here we are as a strong #2 in the lead when it comes to these new workloads. So for example, we are seeing new logos, customers who may have used out of the cloud for most of what they do, or for the first time, sort of starting to use Azure for some of their new AI workloads.
We also have even customers who've used multiple clouds who use that for a class of sort of workloads also start new projects when it's transferred in data and AI, which they were using other clouds. So what I think you will see us is more share gains, more logo gains, reducing our CAC even. And so those are the things of points of leverage. But at the same time, we are not a small business anymore in any of these things. We're significantly -- we have significant scale. And so, yes, we celebrate. That's why we're even giving you the visibility at 1 point of it showing up this quarter, a couple of points showing up next quarter. And those are material numbers.
And so that's kind of what I think will track. And I think Amy mentioned it because we want -- there are 2 parts to even AI, right? There's the models themselves with our partnership with OpenAI. That's sort of one type of spend on compute. And the other is much more revenue-driven, right, which is we will track the inference cost to the revenue and demand. And you're already seeing both of those play out.
Kirk Materne - Evercore ISI - Analyst
Satya, I was wondering if you could expand a little bit on your comments on data platforms. I think we've heard a lot over the last quarter or so about if you don't have a data strategy, it's tough to have an AI strategy. Can you just talk about where customers are right now in that journey to have a more, I guess, thoughtful data strategy? And what does that mean in terms of their ability to adopt AI services? Meaning do they have to sort of tackle the data issue first before they can really take advantage of all the AI services? Or how should we think about that sort of juxtaposition?
Satya Nadella - CEO
Yes, sure. Thank you for the question. Yes, absolutely. I think having your data, in particular, in the cloud is sort of key to how you can take advantage of essentially these new AI reasoning engines to complement, I'll call it, your databases because these AI engines are not databases, but they can reason over your data and to help you then get more insights, more completions, more predictions, more summaries, and what have you.
So those are the things when we say Copilot design pattern, that's sort of what that design pattern is all about. The thing that perhaps even in the last quarter, and I had that in my remarks, that's most exciting is how with Microsoft Fabric, especially for the analytics workloads, we brought together compute, storage, governance with a very disruptive business model.
I mean to give you a flavor for it, right, so you have your data in an Azure data lake. You can bring SQL Compute to it. You can bring Spark to it. You can bring Azure AI or Azure OpenAI to it, right? So the fact is you have storage separated from all these compute meters, and they're all interchangeable, right? So you don't have to buy each of these separately. That's the disruptive business model. So I feel that we are well -- Microsoft is very well positioned with the way our data architecture lays out our business model around data and how people will plan to use data with AI services. So that's kind of what I mean by getting your data estate in order. And it's just not getting data estate in order but you have to have it structured such that you can have the flexibility that allows you to exercise the data and compute in combinations that makes sense for this new age.
Brett Iversen - General Manager, Investor Relations
Thanks, Kirk. That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with all of you soon.
(Tips:This transcript is converted by recording, we will do our best, but cannot fully guarantee the accuracy of the conversion, it is for reference only.)
公司管理
Amy Hood-执行副总裁兼首席财务官 |
Brett Iversen-投资者关系总经理 |
萨蒂亚·纳德拉——董事长兼首席执行官 |
分析师
Aleksandr Zukin-沃尔夫研究 |
Brent Thill-杰富瑞集团 |
Karl Keirstead-瑞银 |
Kasthuri Rangan-高盛集团 |
Keith Weiss-摩根士丹利 |
Kirk Materne-Evercore ISI |
Mark Moerdler-Sanford C. Bernstein & Co. |
Mark Murphy-摩根大通公司 |
基思·魏斯 -摩根士丹利-分析师
为一个伟大的财政年度画上了非常好的句号。Satya,你在评论开始时谈到了每次客户对话中客户是如何询问你使用生成式 AI 技术的,以及他们能以多快的速度利用这种生成式 AI 技术。答案是什么?就它进入市场的速度以及你的客户可以开始使用它的速度而言,你能告诉他们什么?
然后对艾米来说,投资者应该如何看待这些生成式人工智能技术背后的基本毛利率?我们知道在这些之下会有大量的资本支出增加。但是,就所有这些新的生成式人工智能解决方案之下的最终毛利率而言,我们应该期待什么?
萨蒂亚·纳德拉 - 首席执行官
谢谢你,基思,提问。我们与客户之间的基本指导和对话是双重的。从生成式 AI 中获得价值的最简单途径之一是采用某些解决方案,例如 GitHub Copilot。从某种意义上说,为任何组织中的所有软件开发人员增加生产力杠杆都是不费吹灰之力的。无论你是银行、零售商还是软件公司,它都适用于所有人。因此,这可能是我们所看到的非常不错的事情之一,甚至是生产率数据和很好的采用率。
然后很明显,M365 Copilot周围已经令人兴奋了。所以我们首先要谈的是我们自己是如何部署所有这些副驾驶的,无论是 Sales Copilot、M365 Copilot 还是 GitHub Copilot,你如何从这些横向工具链中获得最大价值?最重要的是,我们利用了我们在这些产品之下所做的事情,将其构建为一流的技术堆栈,对,我们在开发者大会上谈到了这个堆栈,叫做 Copilot Stack,然后使用 Azure AI 工具,让像穆迪这样的人能够为自己的员工构建自己的Copilot。
因此,对我们来说,我们希望能够帮助客户在 Azure AI 之上快速构建他们的生成式 AI 应用程序,如果你愿意的话。因此,这是我们要求他们确定在哪里可以获得最大生产率杠杆的两件事。然后我们甚至动用自己的资源来帮助他们把这些事情做好。我要说的最后一句话是云端和云端的数据可以实现所有这一切,因为我认为这里的扩散周期,从某种意义上说,我们有一套新的云计被更快地采用,因为云中还有其他所有东西。因此,这些就是观察结果。
艾米·胡德 - 首席财务官
对于你的问题,基思,关于毛利率以及我如何看待未来的毛利率,我要说的第一件事是,我预计这里的毛利率将随着时间的推移而发生变化,就像之前的云过渡一样。我还要说,我预计工作负载和工作负载的毛利率会有所不同,就像今天在云端一样。
我还要补充一件与上次不同的事情,我们之前谈过这个问题,那就是我们从另一个地方开始,有了更多的共享平台,这使我们能够比上次更快地扩大毛利率。而且,正如你所问和萨蒂亚所说的那样,我们确实预计,这种采用曲线的步伐会更快。因此,你会看到资本支出在第四季度加速,然后在第一季度再次加速,我们已经讨论了今年剩余时间应该是什么样子。
话虽如此,我们正在谈论所有这些,并在24财年比23财年交付的同时经历了这种过渡,营业利润率实际上提高了1个百分点。因为如果按照我们的指导,同比持平,受使用寿命变化的不利影响,当你纠正这个问题时,它会高出大约1个百分点。因此,我认为这里真正的重点是能够积极满足需求曲线,专注于毛利率的过渡和增长以及提供运营杠杆作用。
布伦特·蒂尔 -杰富瑞-分析师
Satya,关于你继续看到的优化不利因素,你认为我们什么时候达到优化峰值?我们是否快要达到这个峰值并在下半年得到一些缓解,也许人工智能可以为我们提供顺风?从你所看到的角度来看,任何颜色都会有所帮助。
萨蒂亚·纳德拉- 首席执行官
当然,布伦特。谢谢你的提问。是的,有几点观察。一个是,我认为,总的来说,在云端,你确实会看到新的项目开始然后这些项目开始得到优化,然后你会对所有这些进行时间序列,这有点像你在普通课程中看到的。这里发生的事情发生在疫情期间,很明显,有许多新的项目启动,从某种意义上说,优化被推迟了,这就是你所看到的,我称之为追赶优化。就你而言,这是我们要圈的东西。我认为,进入接下来的几个季度,它会下降的。
而且我们看到新的项目开始了,传统类型的项目都开始了,甚至是云迁移、数据应用程序,当然还有人工智能应用程序。但是我们会回过头来谈谈我称之为新项目启动和未来优化的正常节奏,我认为,在接下来的几个季度中,我们将循环进行最后一次追赶优化。
艾米·胡德 - 首席财务官
我只想补充一点,布伦特,我想对于 Satya 的观点,也许可以为你划一句话,我认为这与上个季度我们发表了同样的评论非常相似,那就是我们看到了有点正常的优化,而且我们看到在 Satya 所说的这些工作负载中开始了新的工作负载。而且我认为这就是我们未来要说的话,我认为,实际的变化只是与一年前相比有点追赶。你说得对,我们将在 H2 继续这样做。
Mark Moerdler -Sanford C. Bernstein & Co.-分析师
恭喜这个季度。艾米,资本支出环比和同比大幅上涨,而且未来还在增加。你能给我们点颜色吗?是物理数据中心吗?主要是服务器吗?它主要由人工智能驱动吗?我们应该如何看待它的使用寿命?然后对于 Satya 来说,你能不能给我们一些 —— 关于完整 Copilot 开发堆栈的全面可用性情况?客户和合作伙伴建立 Copilots 需要多长时间?
艾米胡德 -首席财务官
我为什么不从资本支出问题开始,Satya,然后我会把它交给你。马克,实际上,首先,无论是在第四季度还是在谈到第一季度,加速确实相当广泛。它两者兼而有之 —— 无论是数据中心还是物理基础再加上 CPU、GPU 和网络设备,从广义上而不是狭义上来看待它。因此,这是整体容量加速的总体增加。
而且我认为,如果你回顾一下23财年,你也不会看到正常的节奏,我想说的是,容量增加,即使是普通的Azure工作负载。因此,你会看到两种加速,即普通的 Azure 工作负载以及一些 AI 工作负载,是部分原因。因此,这就是为什么我经常评论说,这既是商业云的整体需求,也是人工智能能力的建设。两者兼而有之。
萨蒂亚·纳德拉 -首席执行官
是的。而且我认为,从角度来看,我认为考虑一下总是件好事,对吧,我们有什么,1110亿美元的商业云业务同比增长22%。然后你的资本支出增长了,差不多是同样的数字,分别为23%、24%。因此,从某种意义上说,推动新增长的是替代资本加上一些新资本。所以我想,这就是比例。我们对这种整体增长率结构以及它如何转化为我们未来的TAM机会感到满意。
然后关于所有这些如何有效启动项目的另一个问题,Copilot 堆栈现已在 Azure 上线。因此,我们拥有所有东西,包括你可以使用的 Azure AI 工具链、Azure OpenAI,甚至你可以使用 Llama 和 Hugging Face 模型中的开放模型。你拥有所有的 Fabric 和我们所有的运营数据存储,这是围绕生成式 AI 最有用的模式之一,比如所谓的检索增强生成,也就是说,你获取数据存储中的数据,在提示中使用它来生成完成内容、摘要,以及你拥有什么。所以这是我们经常看到的东西,Copilots 从根本上讲就是精心策划的。因此,我们有这些服务可用。
令人着迷的是,当你使用像 Power Virtual Agent 这样的东西时,你就有了低代码、无代码的工具来有效地构建这些 AI 产品或像我们构建的那样成熟的 Copilots。所有底层原语都在 Azure 上可用。该工具链可在Azure上使用,客户部署工具链的速度以及ISV能够构建它们的速度令人印象深刻。
Kasthuri Rangan -高盛-分析师
恭喜本季度。如果可以的话,我只是想听听你的想法,将讨论从 COGS 和 CapEx 转移到更多的收入前景上。看来Azure的增长率肯定开始稳定下来,人工智能对Azure的创造性贡献逐季度明显改善,从更广泛的意义上讲,优化也开始稳定下来。
这与该公司对未来几个季度Azure增长率的展望将走向何方?我们是否已经触底反弹,由于我们讨论的趋势,我们可能会开始看到一些加速?而且,如果我们以微软云的超集为例,当你为Copilot设定新的定价时,看来TAM正在以相当大的方式开放。因此,当你从更广泛的角度来看,萨蒂亚和艾米提到的21%、22%的增长率时,前景会怎样?我们能否过于乐观地抱有未来几年某种加速的希望?或者你如何看待收入前景?
萨蒂亚·纳德拉 -首席执行官
当然。Kash,谢谢你的提问。所以也许我会开始然后,艾米,你可以补充一点,因为我想 —— 我们确实在考虑这里的长期 TAM 是多少,对吧?我的意思是 —— 你听过我谈论这个占国内生产总值的百分比,科技支出将是多少?比方说,如果你相信,GDP的5%将达到GDP的10%,那么由于人工智能浪潮,这种情况可能会加速。那么问题是其中有多少流向了我们商业云的各个部分,然后我们在每个层面的竞争力如何,对吧?
因此,如果你把它分解一下,你有点谈微软 365,我们如何将这款 Copilot 视为第三支柱,对吧?我们有创作工具。然后,我们拥有了所有的通信和协作服务,我们认为 AI Copilot 是第三个支柱。因此,我们对此感到兴奋。艾米首先谈到了我们想如何发布它,也是本次预览的一部分。然后在下一财年的下半年,我们将开始从中获得一些实际收入信号。所以我们很期待。
但是从长远来看,我们将其视为第三个支柱,就像我们过去想过的(比如)Teams 或 SharePoint 之类的东西,或者你有什么。那么 Azure,我的看法是,不管怎样,你仍然处于云迁移的第 2 局或第 3 局,特别是如果你看的话,对吧,无论是行业向云迁移、细分市场迁移到云端还是国家对云的采用,对吧?因此,云迁移本身仍处于初期阶段。所以那里还有很多。
除此之外,还有一个由人工智能驱动一系列新工作负载的全新世界。因此,我们再想一遍,这与 TAM 的机会相比相当广泛,我们会试一试。但与此同时,我们是一个价值1110亿美元的商用云,在20年代发展壮大,因此,我们确实达到了大量定律。但话虽如此,我们确实认为这是一项可以持续保持高增长的业务,这让我们感到兴奋。
艾米胡德 -首席财务官
我想唯一要补充的是,卡什,我想在某些方面,我们真正指向的是这里有一个过程。我们看到需求信号相当强烈。它仍然很强大。我对我们发布的所有产品公告感到非常兴奋。我很高兴他们转到付费预览版然后转到 GA。就潜在市场而言,它们绝对是广阔的。就我经常谈论的首席信息官或首席财务官如何看待这项投资而言,他们达到新的预算池几乎就是我经常谈论的方式。因此,机会越来越大。
如你所知,我们专注于解决这个问题。然后收入就是结果。但这确实需要——需求信号需要资本支出,然后创造机会。这就是为什么我认为,在某些方面,我们多花一点时间谈论其中一些投资,是因为它是需求信号。
Karl Keirstead -瑞银-分析师
好吧,太棒了。艾米,如果我能稍微双击一下围绕 M365 Copilot 的激动人心的消息,因为在线上的每个人都希望将这个机会分层到我们的模型中,我只想听听你的看法。你能给我们提供护栏来让我们保持排队吗?办公板块存在一定程度的毛利率压力吗?换句话说,这是一款成本相当密集的新产品,我们应该记住吗?而且,从某种意义上说,你需要 Azure AD,也许还有其他一些网络安全产品,它能否与 Azure 相提并论?因此,在今晚和接下来的几周里,稍微涂一点颜色可能会对每个人的建模练习有所帮助。
艾米·胡德 -首席财务官
谢谢,卡尔。我想也许我会先从我们发布新产品时的流程开始。而且我完全理解,需求信号、客户的反应,以及我们即将进入付费预览版的请求,也让我们感到兴奋。这一切都令人鼓舞。如你所知,我们已经 —— 上周,我们宣布了定价,然后我们将继续完成付费预览流程,以获得良好的反馈。然后我们将宣布正式上市日期,然后我们将进入正式发布日期。当然,这样我们就能出售它然后确认收入。
这就是为什么我继续说我和其他所有人一样对此感到兴奋,而且应该更加 H2 的权重。我认为,我们已经给了你一些规模调整的机会。而且我想我会用到所有这些。但我确实认为这实际上与节奏有关。当然,我们仍然需要对我们的 Security Copilot 和一些 Dynamics 工作负载进行定价和发布。我们将继续为此努力。
当然,我认为人们经常忽略的一件事是,当你回过头来谈谈 Azure 时,Satya 简短地提到了这一点,我认为从很多方面来说,这些人工智能产品中有很多都使用 Azure,因为构建应用程序所需的不仅仅是人工智能解决方案服务。因此,与其说是微软 365 的支持,不如说是任何 Copilot。就是说,当你构建这些时,它需要数据,也需要人工智能服务。因此,你会看到他们同时使用核心 Azure 和 AI Azure。我认为这也是一个重要的细微差别。
萨蒂亚·纳德拉 -首席执行官
是的。如果我能补充一下艾米所说的话,这里的平台效应实际上完全与 Copilots 的可扩展性有关。今天,当人们在 Teams 中构建基于 Power Apps 的应用程序时,你就会看到这一点,而那些 Power Apps 恰好在 Azure 上使用 SQL DB 之类的东西。这就像经典的业务线扩展。所以你会看到同样的东西。当我有 Copilot 插件时,那个插件会使用 Azure AI、Azure 仪表、Azure 数据源、Azure 语义搜索。因此,很明显,你不仅会看到身份或安全层的突破,还会在 Azure 的核心 PaaS 服务以及 M365 中的 Copilot 可扩展性中看到。
马克·墨菲 -摩根大通-分析师
Satya,现在有太多的证据表明 GitHub Copilot 正在将开发者的生产力提高了 40% 到 50% 或更多,从而产生了更高质量的代码。你是否设想 Microsoft 365 Copilots、Security Copilot 或 Sales Copilot 的工作效率会有类似的提升?换句话说,能否将房屋中的每个房间都改造到类似的程度,以至于整个堆栈中的价值主张都得到了相当高的提升?
萨蒂亚·纳德拉 -首席执行官
是的。贾德森·阿尔索夫(Judson Althoff)会很高兴你用他的比喻用人工智能改造房子的每个房间,因为你说得绝对正确。我的意思是这就是我们所看到的机会。我想你还指的是现在有很好的经验证据和数据围绕 GitHub Copilot 以及围绕它的生产率统计数据。我们正在为 M365 Copilot 积极研究这个问题,也为诸如 Sales Copilot 或 Service Copilot 等基于角色的角色开发这个问题。我们看到这些业务流程具有非常高的生产率提升。所以是的,在这一年中,我们将拥有所有这些证据。
而且我认为,归根结底,正如艾米所提到的那样,每位首席财务官和首席信息官也会看看这个问题。我确实第一次认为 —— 或者更确切地说,我确实认为人们会考虑如何用这些 Copilots 来补充运营支出支出,以提高效率,坦率地说,甚至减轻运营支出和员工等的工作负担和繁琐工作。因此,我想你会看到所有这些都转化为生产率统计数据,我们期待着这些数据公布。
亚历山大·祖金 -沃尔夫研究-分析师
我想可能只是一个多方参与者。你曾几次提到,随着你在 Azure 上看到的人工智能工作负载采用情况,与前几代产品相比,它看起来可能与增量份额增长的角度略有不同。你能再详细说明一下吗?这应该如何推动 Azure 的消费,尤其是在我们度过这一年之际?而且,你是否看到这样的场景:无论是优化逆风,再加上人工智能的贡献,再加上你在工作负载周围看到的这种增量顺风,确实推动了 Azure 的重新加速,尤其是在下半场你将开始看到其中一些事情开始出现的时候?
萨蒂亚·纳德拉 -首席执行官
是的。我的意思是,我们既看到又兴奋的是两个新的工作负载。我的意思是,如果你想一想 Azure,这些年来,我们已经从后面发展了 Azure。在这些新工作负载方面,我们在 #2 中处于领先地位。因此,举例来说,我们看到了新的徽标,客户可能已经在云端使用了大部分工作,或者是第一次开始使用 Azure 来处理一些新的 AI 工作负载。
我们甚至还有使用过多个云的客户,他们将其用于一类工作负载,当它们在数据和人工智能中传输时,他们也会开始新的项目,而他们使用的是其他云。因此,我想你会看到更多的股票收益,更多的徽标收益,甚至减少我们的CAC。因此,这些都是杠杆点的东西。但与此同时,在任何这些事情上,我们都不再是一家小企业了。我们意义重大 —— 我们的规模很大。所以,是的,我们庆祝。这就是为什么我们甚至在本季度为你提供1点的知名度,下个季度有几个百分点的曝光度。这些是物质数字。
所以我认为这将是可以追踪的。而且我想艾米之所以提到它,是因为我们想要 —— 即使是 AI 也有 2 个部分,对吧?还有我们与 OpenAI 合作的模型本身。这有点像一种计算支出。另一个更受收入驱动,对,那就是我们将追踪推理成本与收入和需求的关系。而且你已经看到这两者都在发挥作用了。
Kirk Materne -Evercore ISI-分析师
Satya,我想知道你能否稍微扩展一下你对数据平台的评论。我想在上个季度左右的时间里,我们听说了很多关于如果你没有数据策略,就很难制定人工智能策略。我猜你能否谈谈客户在制定更周到的数据策略的旅程中现在所处的位置?就他们采用人工智能服务的能力而言,这意味着什么?这意味着在真正利用所有的人工智能服务之前,他们必须先解决数据问题吗?或者我们应该如何看待这种并置?
萨蒂亚·纳德拉 -首席执行官
是的,当然。谢谢你的提问。是的,绝对可以。我认为,特别是将你的数据放在云端是你如何利用这些新的人工智能推理引擎来补充(我称之为数据库)的关键,因为这些人工智能引擎不是数据库,但它们可以推理你的数据,帮助你获得更多的见解、更多的完成、更多的预测、更多的摘要,以及你拥有的东西。
因此,当我们说 Copilot 设计模式时,这就是设计模式的全部内容。即使在上个季度,也可能最令人兴奋的是,通过微软Fabric,尤其是对于分析工作负载,我们将计算、存储、治理与极具颠覆性的商业模式融为一体。
我的意思是给你一个味道,对吧,这样你的数据就在 Azure 数据湖里了。你可以把 SQL Compute 引入其中。你可以把 Spark 带到它身上。你可以把 Azure AI 或 Azure OpenAI 带进去,对吧?所以事实是,你的存储空间与所有这些计算计量器是分开的,而且它们都是可以互换的,对吧?因此,您不必单独购买每个。这就是颠覆性的商业模式。所以我觉得我们很好 —— 微软在我们的数据架构围绕数据布局商业模式的方式以及人们计划如何将数据用于人工智能服务方面处于有利地位。所以我所说的整理好你的数据资产就是这个意思。而且它只是无法按顺序排列数据资产,但你必须对其进行结构化,这样你才能灵活地使用数据和计算组合,这对于这个新时代来说是有意义的。
布雷特·艾弗森 - 投资者关系总经理
谢谢,柯克。今天财报电话会议的问答部分到此结束。感谢您今天加入我们,我们期待尽快与大家交谈。
(提示:此笔录是通过录音转换的,我们会尽力而为,但不能完全保证转换的准确性,仅供参考。)