Transcription
Satya, what brings you here to Germany? You know, it's our AI tour, uh, today in Munich, and I'm excited to be here. Is it more difficult these days to come here and, um, be in Germany and answer questions?
Not really. I love it. I mean, it feels like the atmosphere has changed a little bit. There's more skepticism. There's some governments in Europe even thinking about, um, only allowing European tools. There's even a supermarket giant that's developing their own cloud.
Well, I mean, look, you know, first of all, I think there is real reason for why there is competition always. There is local competition. There is multinational competition, and we welcome that. And I think what Germany and what Europe deserves is great innovation, both coming out of this region and from the rest of the world, because, in fact, when I look at Europe and look at Germany, uh, in fact, I, I always describe it, you know, whenever I am in a jeweler in the United States or when I'm at the dentist, I see the German middle and its excellence, uh, and so the world, I think, will remain, uh, a connected world where the innovation from any region reaching the other regions is going to be the true sign of progress.
But you, but you sympathize. You can understand that some German or European companies are a little worried these days.
I mean, I, first of all, I think that any country should sort of think about how should they think about their own innovation agenda, their own competitiveness, their own sovereignty. That's the topic of the day today. And I talked about this even today in my talk, which is how one can take all the options. For example, if you look at it, right, we are definitely an American company. We're a multinational company, but we have significant capital that we have invested in Europe, in Germany, putting our capital at risk, so the German companies can use it and then create more value, right? So it's not like we're coming here with just software. We have both innovation and investment here locally, and we have all the sovereign options, right? You have the ability to manage your own keys, you can encrypt your data, you can have confidential computing, you can even have private cloud, cloud options, and we have even local partners. And so, in some sense, I think as long as one takes the approach that, look, at the end of the day, if you want to manage risk, you manage it using a portfolio, like everything else in life, um, and then to be able to think about it in clear ways and say, at the end of the day, Germany needs to have sovereignty and also produce things out of Germany that get exported around the world, and we can be a great partner for that.
All right, um, maybe briefly on your career. It's amazing. I mean, when you started as CEO, O, 2014, um, the company was 10 times less valuable than today, and you, I think, increased the revenue by a factor of five or six. Amazing numbers.
How, look, look, at the end of the day, um, what matters the most is what happens next, as opposed to what happened in the past. And if I sort of had to recount how is Microsoft, not just in my tenure or whatever 12 years, but what's the real question is, what is it that after 50 years after our formation here, we are in 2026, uh, and going at it, um, and the fundamental thing to me is having that sense of purpose and mission that allows us, uh, to be relevant going forward, and a culture that allows us to build the capability, right? Even take the, the thing that we're talking about in AI, it was back in what, 2019, 2018, when we decided that there was going to be essentially a new regime, uh, of AI innovation, and see it long before its conventional wisdom.
Is it true that you invested in OpenAI in 2019, I think, for the first time, and Bill Gates was not a big fan of that?
We, I think, we've, Bill has talked about it. I've talked about it, uh, you know, externally. And I mean, after all, you know, who would have thought in 2019 that there is a smart idea to go put a billion dollars? Everybody thought we were just burning a billion dollars. And so, you know, it turns out some of these things, when they work, they are celebrated, and when they don't work, nobody knows about it. And so, therefore, look, at the end of the day, I'm, I'm, I'm more like the thing that we were obsessed about, and Bill, in particular, was obsessed about, was natural language, right? I mean, throughout our history, from the beginning days of Microsoft research, was, can we really have AI breakthroughs in natural language? And so when, uh, the OpenAI team started pivoting from what at that time they were doing a lot more RL and, uh, and what have you, interestingly enough, they switched to natural language and scaling laws. Um, and that's when we sort of picked it up, and, uh, and rest is history.
I mean, there's the legend that you met Sam Altman at Sun Valley at the conference. Um, what did you see in him after the first meeting that, in the end, made you make this investment?
You know, I, I've known Sam for a lot longer than that. I met him, you know, maybe in, you know, when he was in his first startup. Uh, and so it's not that I ran into him in Sun Valley for the first time. Okay. So, I've known Sam, uh, for a long time. And, um, and quite frankly, some of the other founders of OpenAI and what have you. And I think, I mean, fundamentally, interestingly enough, I think the paper that was originally written, uh, there was even written by, I think, Dario, when he was at OpenAI, on scaling laws. Uh, and so, you know, look, I think it's a fantastic team they assembled. They had real belief in scaling. Um, they, uh, and then we thought, uh, that we can be great partners. Um, and so, and, uh, we went all in, and, uh, it's been great for them and for us.
Are you sometimes concerned we are still in an AI bubble, though?
You know, look, I think it's, at the end of the day, there's only one thing that matters, right? Which is, it's not about talking endlessly about the technology. It's about seeing that technology in the real world having real impact. Right? I come here to Munich and I sit and sort of listen to a startup out of Munich, uh, that's building one of the world's best customer experience, Parloa, right? Which is fantastic. You see they're building, they're winning share, uh, right, uh, that's great to see. Talking about German innovation, which is not obviously born in Germany, but winning world over, and they are partnering with us, uh, or they go the other extreme and say, um, you see the Munich fire department using AI to get better at their emergency transfers. Right? To me, the bubble is a bubble if you don't have these examples of diffusion. I give you an example for when I thought it's might be a little bit overhyped. Um, the third most, the three most important drum companies, SAP, one of them, Deutsche Telekom, um, Siemens, they're all together less valuable than OpenAI is these days. Is that real?
Well, I mean, I think at the end of the day, there is going to be more of a broadening, uh, of the economy. In fact, if anything, the, in the fullness of time, uh, right, if, what is AI going to do? AI is going to do two things. One is, if you think about it, it's a real physical investment, right? After all, you brought up Siemens, right? I think Siemens is going to be one of the largest suppliers of a lot of the equipment that goes into data centers. So I think Siemens is going to be a major beneficiary, uh, of what is going to be the buildout of AI, not just in Germany, but world over, right? Um, and so the, there is going to be a, and energy is going to be a massive requirement. So I think there's going to be a broadening of the supply side of it. Then on the demand side, right, as it gets used, it's going to create surplus, it's going to create local surplus, whether it's in the health outcome, or whether it's in the public sector efficiency, or what have you. So ultimately, there should be and will be GDP growth, right? That's the real test of any of this.
Pure valuations is staggering.
Valuations are, look, at the end of the day, you know, stock markets work because people are really looking at what the forward earnings potential of anything is. And it's sort of like, you know, in the short run, it is what is it that Warren Buffett thing, right, which is it's a, uh, voting machine, and in the long run, it's a weighing machine. And I'm not going to comment on, uh, stock prices and valuations. That's for investors to make sense of. But what I think of is an all-up economy benefits from new general-purpose technologies lifting all boats when it comes to productivity, and this is what's happening.
Do you think there's like some new breakthrough innovations in AI? Is there something after the transformer models on the frontier already?
Yeah, I mean, this, it's a good, great question, right? When I, I think about, um, even, it's not no longer just one transformer model. I mean, if you look at, you know, where we've, Oh, it was all about pre-training scale. Then we, it was about post-training. Then we came up with reasoning. Uh, then we said, "Oh, there's RL." Uh, and so there's constant innovation, um, that's happening even in what is considered transformers and transformer architectures. Uh, there is obviously going to be, I always say that we are one sort of innovation away from the entire regime changing, even, uh, in terms of what's a new model architecture that could even be more efficient, uh, in its performance. And so I think we have to be very open-minded about it. I brought three, five years out questions. The first one is, in five years out, will there be vast amounts of people without a job?
I think it's going to be one of the biggest questions for us is how does work, work artifact, and workflow change. So let's take software development, right? Because after all,
this is the place where it all started, right? In fact, my own belief in this particular transformer models and its scaling properties was, you know, reinforced when I saw next, it suggests in GitHub Copilot. Um, if I look at what's happening with software development, right, and if, in fact, if you look at even the history of software development, you know, at least, you know, I started learning with assembly, and then compilers came, and then we went to higher-level languages. We went to interpreted languages, and we've had, you know, many, many tool chain changes and levels of abstraction that made us more productive. This one is a, another big leap forward in making it possible for both reducing the floor and raising the ceiling for anyone who is a software developer. Right? What does it mean reducing the floor? Anyone can be a software developer, just like how Excel reduced the floor for anyone to be an analyst. Now, anyone can be a software developer, but it's also raising the ceiling on, oh, what is the new sophistication you need, uh, in order to be productive with these new tools, so that these code bases that are getting generated are not black boxes? That's going to be what all of us will have to reskill ourselves on. So, to me, what I think is more important in the short run, I'm not saying that there is not going to be displacement, we have to be clear-eyed about it. What's the best protection against displacement? It's to one, understand the new medium, the new tool, the new skills required, and transform yourself in job, and that is what's happening in software development. And you can see it where software developers are getting more productive. The IT backlog is huge. I, I, I totally buy it for, for the software developers bubble or that industry. But I mean, there's talk about like whole economies.
But let me say this, right? If somebody in in the early '80s had come to us and said, three billion people, four billion people are going to get up every morning and type, and we would have said, why? Right? You know, we have a typist pool and we have a slide pool. Why would four billion people type? We invented this entire thing called knowledge work. Uh, and so, I, I think at some level, I'm not like being polyannish about this, but I'm saying, let's at least have a bit of optimism in our ability as humans, and human societies, uh, and as political economies, right? One thing that is discounted is like, we have control. It's called political economies and elections, and people are not going to tolerate anything that is not going to be beneficial at society scale. And so I'm a big believer that the, what is the real governor? It's just not going to happen to us. We're going to shape what happens to us.
It's, it's going to be difficult. I mean, if this disruption happens fast, I mean, we're up for huge challenges. I mean, if this unemployment really comes, it's going to be challenging.
For sure. If unemployment suddenly comes, it's going to be challenging. And so that's where I would say, let us all focus on making sure that we are doing all the things, uh, to help us skill ourselves such that we have more economic opportunity with what is going to, uh, the new productivity curve.
So, but you're not afraid of like huge waves of unemployment?
I, I, I am afraid of anything that is going to not help human societies, uh, really make progress, right? So, anytime anyone gives me a dystopian worldview that we are all going to be unemployed, of course, I'm afraid of it. Uh, but I'm also saying, I'm not so out of control of it. Like, we all as a c, you know, as a society get to sort of make sure that the technology we build and how it diffuses, what happens to wages, how we litigate, what expertise we will reward and not reward. There's so many things, uh, that we are have control over that, that we should not abdicate all that.
Um, next five years question. Um, do you think Germany will still be a top three market for Microsoft in the next five years?
I mean, I, I definitely hope so. I mean, to me, you know, what stood out and made Germany, uh, unbelievably successful in the world is the German ingenuity, the German engineering, and its ability to produce things that the world needs and the world wants. And so if anything, uh, I think that industrial base that Germany has now, coupled with even the digital capability, in fact, I always feel again, going back to my middle example, why not think of, you know, now all this digital data, right, and the exhaust of digital data coming back from everywhere, being again, uh, translated into the intelligence layer that goes with the Siemens equipment, right? One of the things that we're working with Siemens are all their digital twins, twins, and the intelligence layer in their digital twins.
You talk to German leaders regularly.
Absolutely.
Like who do you talk to?
I mean, I talk to Ronald at Siemens. I talk to Ola at Mercedes. I talk to, in fact, I'm going to talk to many people today. And so, yeah, I'm, I'm indefinitely in contact with, uh, uh, German CEOs all, all the time.
Do you think in the next five years, um, Microsoft will still rely heavily on revenue from Office and Windows?
You know, when we say rely on Windows, Office, and, uh, and Windows, it's fascinating. What is Office? I mean, if I asked you, right, uh, the question, Office, when we started, people bought Office clients, then they bought servers, and when it came to the cloud, they literally kind of bought really data, uh, and the data storage and the sharing and collaboration. And so there is no such thing as even what is Office. It has gone through both business model transformation, value transformation, right? And so I look at even our quote unquote, uh, in fact, in the agent world, like the first thing people are doing for agents is provision Office for it, because agents need to collaborate, need to be able to collaborate with humans. Uh, and what's the best way to do it in a Teams channel, to have access to email? Like when I want, I have a coworker, my coworker that takes my identity needs, in some sense, a separate Office instance. And so I think that one way to think about is, if we remain static in anything, of course, that business goes away. But when you think about what happens next, and this happened to the cloud, right? We thought we had a fantastic server business, and we thought the cloud will destroy the margins. It, of course, destroyed the cloud margin, I mean, server margins, but it expanded the market so massively. And I look at what's going to happen with agents, and I look at that and say, wow, this is the best news if you really are in the Office business, because every agent will need Office.
You always seemed like such a positive, sunny guy, and what you touch turns into gold. But I'm sure you had to make some huge and, and some very tough decisions in your life, in your business career. What was the toughest one?
Well, I mean, at the end of the day, you know, toughest calls are when you have to make redeployment calls that impact people. I mean, at the end, you know, that's, I think the hardest stuff. But in our case, when it comes to business strategy, you know, let's, you know, call it what it is, right? When we realized that, uh, when it came to the phone market, we were too far behind, uh, and we needed to pick where, uh, we focused our efforts, and we picked the cloud. We always thought that the cloud as a control plane to go across devices was more important. And that's, I mean, even, you know, making sure that Office was available everywhere and what have you. And so, um, you know, those were tough decisions. But at the end of the day, I, I, I look at this and say, look, at the end of the day, for us, making sure we are looking at the future, um, irrespective of, you know, what happens to us, we have to lean into it, uh, because we can't deny what is going to be a secular shift.
Any tech innovation that you'd still like to see in your lifetime?
You know, to me, my, my sort of dream would be the coming together of these, um, two things that we're working on, right? Quantum plus, uh, AI. And, in fact, the juxtaposition of the two, right? One of the things we showed even at my keynote today was this model we built called GigaTime, which basically, you know, does, um, a simulation, uh, of spatial proteomics, right? So it's, it's, it's a test that is used in, you know, to check whether a particular immunotherapy can work for a particular cancer tumor. And today, it's an expensive test. And, um, and so therefore, if we have a model that really is in silico, it's then going to be available in every secondary hospital. Now, think about training that model, right? Training that model, a quantum computer can do the best simulation. After all, we're trying to learn the language of nature, and nature is quantum. And so therefore, a quantum computer will allow us to get training data to train AI models that then in silico can be much better at simulation of the real world, and that will allow us to give it more predictive power.
Last question. How does a Microsoft CEO think about longevity and how do you work out?
Well, I get up, try to get up every morning, irrespective of the jet lag, and get to hit the gym. That's.
Every morning?
Every morning.
For how long?
Half hour.
Half hour, and that's it?
That's it.
And then you do pull-ups and push-ups?
Well, I mostly nowadays have like, since my knees have no longer in shape, I'm mostly doing incline walks.
Okay. Okay. Thank you very much, Satya. Thank you so much.
That's a pleasure.