Microsoft CEO Satya Nadella: AI Boom, Energy Battle & His Complicated Alliance with Sam Altman

Satya, I wanted to speak with you about, surprise, surprise, artificial intelligence. Are you sometimes bored by the topic to always be asked about it? You know, you can be bored, Matthias, in this thing. You just are excited and every day you thank yourself that we get to sort of do what we do in terms of building these technologies. But ultimately, it's the use of the technologies that need to be celebrated. I mean, what was the info that you got, or the experience that you made, that surprised you most about artificial intelligence in the very recent past? Yeah, I think, you know, it's interesting you ask the question the way you asked it, because in some sense, it is a little bit of, when I even look at my journey in this new generative AI, how we even got to open AI, the partnership, how we sponsored them when it was obviously not conventional wisdom, taking the big bets and all of that. It's really, Matthias, having a prepared mind, right? So Bill started Microsoft Research in 1995. You know, his idea was that, look, Xerox PARC was the place that inspired a lot of the companies, whether it was Microsoft or Apple, and we wouldn't have existed, the PC industry wouldn't have existed. So he wanted to create a fundamental research organization. And in fact, the first group I think we had was a speech group. And, you know, we've been at it. And in fact, we've been obsessed about natural language, right? After all, Microsoft is a productivity company. We think about text, we think about natural language, whether it's in speech or in text form. And so, interestingly enough, we have taken many shots on goal on that. And we continuously kept at it. And then when Sam and team, you know, came and said, hey, they're going to apply these transformers to text. In fact, interestingly enough, in 2016, they were much more RL focused. Although lately reinforcement learning has come back, their original focus was more, you know, not as much text. And so in 2019 is when they said, let's really go after, you know, scaling laws of transformer applied to transformers applied to natural language. And that was magic, right? I mean, if you sort of say what has played out and we have added to it a few other techniques over the years, and now I think incrementally, even new things that are coming together, but no one had thought of scaling laws applied to basically text to. And then with instruction following one of the magical moment, there are two products, GitHub Copilot and ChatGPT, right? GitHub Copilot actually happened first with a codex like model where we could see next, next, next, right? So tab, tab, tab, and you could have code completions. And when I saw that, that is when I really felt like, wow, this is crazy, right? This is kind of completing your thoughts, right? It's not just spell correction type things but it is more completing your thought and doing it at speed. And so that was the first breakthrough. Then given the instruction following chat, right? So you could take a model and just teach it a little on how to sort of think about conversations. And it was able to hold a conversation. That was the GPT 3.5 ChatGPT moment. And then the rest obviously has followed since then, but I think that's it, right? Which is if I had to take away, the lesson is that compute or log of compute is intelligence. We obviously need to get a lot more efficient because log is not a fun thing, right? It's not the best way to spend compute to get intelligence. But that said, there are these scaling laws that are working. Were you surprised that large language models are particularly convincing if it is about emotions or faking emotions, if it is about creativity, even if it's about humor, or did you expect that? No, that I think, you know, at some level, it's an imitation game at the end of the day. And so therefore, I think at this point, I think these are the big science questions that still need to be answered, right? Which is, if you look at it, it's not as if we have discovered the cognitive core, right? You know, if I said, we have to have you figure it out, because in an interesting way, the mental model, you should have, and Andrej Karpathy recently talked about it, right? Which is, which I think is the right metaphor. If you really want to create crack intelligence, you need to throw away all the data. Essentially, you can't regurgitate, right? You have to throw away all of your training data, and just have learned the algorithm for learning, right? I mean, think about it, right? A baby is born, a baby experiences the world and learns by experience. So it's not, so we kind of have this quasi machine today that sort of can regurgitate sometimes, but at the same time in that regurgitation, it also does have pattern recognition, which is very powerful. So it does have some, I'll call it emerging cognitive core that's mixed up with this knowledge it has. So in other words, if you can separate out the knowledge of the model from its cognitive core, like that, I think is the next big scientific breakthrough. And by the way, if you do that, it'll be a lot more energy efficient. After all this brain, you know, what is it? 10 Watts, what have you? So therefore, I think we need even a little more science to help us understand what does it mean to have a cognitive core learning algorithm? Satya, you have said it many times in public and also in your shareholder letter, you spoke again about the transformation of Microsoft. It is basically a reinvention of the company. And it is at the same time, it needs to be also a reinvention of yourself. So what would you say is the toughest part in that, both for Microsoft and for yourself? Yeah, no, I think for all of us, I mean, it's sort of such an important thing, right? Which is, so to me, there are two things which I would say. The original premise that Bill had when he started Microsoft was, which I think has really been a very helpful framing for all the success we've had in the last 50 years was, he said, look, there's a software revolution about to happen. And guess what? He said it when there was no software industry, right? In 1975, there was no such thing. There was systems company, there was IBM, there was DEC or whatever. There was no idea that the PC didn't exist. The microcomputer revolution had not happened. And more importantly, it was not even clear that there would be an industry called the software industry. So at that time, he said, you know what? Software is the most malleable resource. Let's become a software factory. It was not about, oh, we'll build an operating system, or we'll build a languages tool, or we'll build an office productivity, because they all came subsequently, right? Microsoft's path, quite frankly, has been a meandering path, right? Most people don't remember. We built Office for the Mac first, and Windows came later. Our first tool was a basic interpreter. We built Flight Simulator before Windows. So we were mostly a software factory, and we said, hey, we'll build whatever the world needs. And so that was the attitude that has really helped, or the metaphor that helped us through 50 years. Because quite frankly, even though technology changed rapidly, after all, we started, you know, on even pre-PC, and then scaled all the way to the cloud. But the fundamental game was always about, oh, you're a great software factory. But now, we have to kind of help you at Axel Springer build your own software factory. So it's sort of, in an interesting way, it's become a different game where whatever we do, we need to make every company and, you know, create their own AI slash software factory. And so that's the paradigm shift, which, by the way, is a pretty big paradigm shift. And that's what we're in the midst of. And so that's where you've got to go back to your core mission, Matthias, which is, what do we stand for? What is our brand around trust and empowerment? And reinforce it with a culture of learning new things, right? So I always go back to a sense of purpose and mission and culture reinterpreted for what is a world of technological shift and business model shift. But in that process, did you yourself also need to change a little bit your skill set? What did you develop that you didn't have before? Or would you say with exactly the same skills you were and are able to manage that transformation? It's a great one, right? One of the metaphors we use a lot coming out of this child psychology and Carol Dweck's work of growth mindset is when you say skill set, I think the most important skill set for long term relevance is, you know, how do you learn it all and not know it all, right? That, I think, is the core. And especially when you have had success, it becomes even harder because the reason why it becomes even harder is you have to unlearn the things that made you successful to learn something new, right? I mean, if you think about it, right, even today, as I sit around this entire weekend, I spent all the time trying to get myself to understand how new companies are building products which are different than, say, how we have built our products. After all, our scale is a massive advantage, but it's also a massive disadvantage. Let's face it, right? Which is it's a massive disadvantage because in order for me to round trip between product, science and infrastructure, I have three divisional heads who manage those three things versus in a startup, they will be all sitting in one little table, all coordinating across the three. And so to relearn what it means to create the magic of, you know, production in the new world, that, I think, is where I would say you have to. So you have to do that. And in order to do that, you have to have a mindset that's curious, that's capable of hitting the books, doesn't kind of have the ego get in the way. And that's a hard thing. So if it is about basically shaping the whole ecosystem and helping companies to be successful, which companies will survive or will benefit from the AI transformation? What's the skill set for leaders in other companies and for the companies as such? That's a great point. So here's sort of, I'm glad you asked it, because I think one of the things that needs to be talked about a lot more is what is the future of a corporation in a world of AI, right? I mean, if you think about it, in the previous era, you had digital systems, office productivity systems, you know, ERP systems, CRM systems, whatever, the newsroom systems, you could sort of build or buy and deploy these digital systems. And what did they do? They basically digitized artifacts around people, places and things. And you were able to make sense of it and get some productivity gains, because once you have a digital artifact, you can automate, you can have new workflows, and it will just speed up the knowledge turns. Andy Grove used to have a nice term, which he described basically, any company is only as good as its ability to create new knowledge and diffuse that new knowledge throughout the organization, right? That's sort of what traditionally corporations have done throughout history, and digital technology gave it a boost. Now, AI, I think, gives it another level of boost. I'm sure we'll talk about sovereignty. I think the real question today is not about even sovereignty at the country level or the continent level in the traditional sense. It is the sovereignty of a company in the age of AI, right? Because if the entire thing is about one model that knows it all, then the question that needs to be answered by every company is, what do I know that is unique? And more importantly, how do I even play the game? How do I, in fact, ensure that I don't so fall in love or I'm not so in awe of one model company versus the other model company? In fact, I think that anyone leading a company in the real world, you take out the tech sector, right? The height. Everyone in the real world has to be most concerned about, do I have my own AI factory that is able to create my own foundation model? It's not about celebrating somebody else's foundation model, because the future of every company will be some unique foundation model, so to speak, that they have that represents their knowledge, not even at a given point in time, as a thing that continues to improve, right? You at Axial Springer will want to build your own foundation model that is unique to your company's knowledge and its ability to continue to compound that. That, to me, I think is what, whether it's in the software industry to build products or whether it's in the content business or whether it is in manufacturing business, it doesn't matter. All of us will have to create our own AI software factories or AI factories that allow us to create foundation models that are unique to us. Satya, do you think that one day there will be companies entirely invented and run by AI or is that startup perhaps already being launched? What do you think? I think that it's sort of too far-fetched for me, Matthias, to sort of imagine. I mean, at some level, you can always say if somebody had said, oh my God, we are going to have these dark factories. When I go to my data center after, by the way, it's been constructed, installed, and it's running, it takes a lot of people to put it together and it's just running autonomously. So I think there will be levels of automation that will be stunning five years from now, 10 years from now, when we look back at it, but there will be people. I'm much more of a believer that, in fact, it's interesting. I'll give you this metaphor. One of the new products that sort of we're building on is, let's say, I call this new metaphor or paradigm of how I work with AI as macro-delegation and micro-steering. So I delegate to a bunch of agents. Let's say I get up in the morning and I say to agents, do this work for me. And then I just say, no problem. And they go off and do their work. Guess what? They come back and report to me. I mean, at the end of the day, they'll say, oh, I got this done. I got stuck here. I need your guidance here. I need your help here or what have you. Guess what? What do I need? I need a new inbox, a new inbox to deal with, not emails, but now with all the agent micro-steering. So as somebody said to me, the future of software is going to be all about a new inbox, a new messaging tool, and a new cursor on a blank canvas. And I think that that's sort of what we will do. And humans will need an interface to AI, even with increasing levels of automation. There will be productivity. There will be levels of abstraction. But I do think human agency is going to be very much part of it. Everybody's talking about productivity, efficiency gains. If it is about artificial intelligence, very few people talk about quality and new tasks and so on. But let's stick to the productivity and efficiency. I saw a Gartner survey that was really interesting. They said that 74% of the CFOs in big corporations believe in the significant efficiency gains and cost advantages through AI, but only 5% have already seen any efficiency gain or cost effect because of AI. How do we explain that gap? So do we miss the capabilities to deal with AI? By the way, this Gartner survey was on another level, also very pessimistic. They said by 27, 40% of AI tools and projects that have been used by corporations will have failed. How do you, where does this new kind of skepticism come from? And how do you see the gap between CFOs optimism and real cost advantages? Yeah, no, I think that's a great question. And in fact, this is probably one of the fundamental things that will cause, I'll call it firm level differences, right? I mean, when I just speak for myself at Microsoft, I mean, this is making massive productivity changes, whether it's in our customer service, whether it's in our supply chain operations, you've talked about finance or in finance forecasting, there isn't, I mean, every day, in fact, it's no longer even the top down, it's the bottom up revolution that in some sense is happening, which is taking out drudgery, improving the quality of the outcomes, improving the flow of work, right? Somebody like, just to give you one little example, we are building, I mean, we have become a very capital intensive industry, as you may have noticed. And so we're building all these data centers. We have 500 plus fiber operators around the world. And fiber at the end of the day is a physical asset, things happen to it, right? They get cut or what have you, and someone has to repair it. And it's a very manual interaction with even the fiber operators. So the person who is like basically managing all of that internally built a full agentic system to deal with all of that, right? Because that's the only way, because if you think about it, we grew as much DC capacity in the last 18 months as in the first 15 years of Azure. I mean, it's pretty crazy in the sense of the scale. So the only way we can manage that type of scale is by using some of these tools. But to your point, though, Mathias, which is, I think, very, very important to your Gartner report or what have you, I think that there are four or five things that companies have to learn. One is a mindset. The mindset is not about sort of applying AI just to what you have today, but it's the change management that has to be imagined with AI. So the first thing is you kind of have to, you know, in the 90s, we used to talk about business process re-engineering. Guess what? It's time to think about business process re-engineering mindset. The second thing you do need is a new tool set, right? So you kind of have, you know, that's the type of stuff that we can provide or what have you. But you've got to equip yourself with a new set of tools that then can be applied to the new process. Then you need the skills to apply that, right? And that to me is the combination. And by the way, then you need data sets. The data sets can't be captive to your existing systems. You have to normalize the data across multiple systems. So if you really think about composing that mindset, tool set, skill set and data set, and then go and apply, oh, I have a new evaluation, right? That's kind of how you do these AI systems. So there's a lot of learning, right? So if you go and say, oh, I'm going to approach it like any other IT project from even last year, it's going to fail by definition. And so that's where I think a lot of the hard work that's got to be done at the economy level is the re-plumbing of the IT skill set. And that's kind of hard because that's change management. It requires enlightened leadership. It requires people to get re-skilled and then scope the projects appropriately, pick the projects appropriately. That's interesting. Nevertheless, even if we get that done, it seems that at the moment there is a bit more of, yeah, let's say, future skepticism with regard to this idea that AI will be the kind of unstoppable growth booster. And more and more people are also asking, could it be a bubble, or at least could there be a where a bubble is bursting? And some people are basically attributing this vibe shift to a conversation that you had recently together with Sam Altman, an interview that you did together. And then Sam was asked whether it's kind of reasonable to have a company that generates 13 billion revenues and at the same time commits for 1.4 trillion, which in a way we have seen all these big deals in the last weeks. It is a bit scary. So what is your take? How will this development be? Will we see a kind of almost linear or exponential development that goes only in one direction or will there be a correction that could be also a kind of healthy shake out of unsuccessful ideas in companies and could kind of bring the euphoria to more realism? What is your take about that vibe shift and the concern that the AI bubble could burst? The way I think, Matthias, is at the end of the day, the test is simple. The test is it does need to have broad impact across the economy in shifting the productivity curve and leading to GDP growth that's broad spread. Because at the end of the day, nothing can outrun the economic growth, right? Because any supply side thing will have to ultimately meet a demand side thing. And so in time, things have to work. Otherwise, to your point, it will be a bubble. But that said, the reason why we have confidence is not because of some future that we imagine, is because this is where I think it's always the case in tech, which is the future is already here, except it's not evenly distributed. And so that is sort of the asymmetry. If I had not seen GitHub Copilot work in 2020 or 2019, even now, when I look back at it, I don't think we would have done the investments that we would have done. Or I see the speed with which Copilot is kind of improving in its reasoning capabilities, right? I mean, even this weekend, some of the things, right? Literally, there was this unbelievable, like even in response to an email, and it's not like, oh, I hit a button and it just said or came up with some English language response. It really reasoned on the question. It felt like a thought partner was working with me to ask precision questions of the person who had sent me an email. I mean, that's magical, right? I mean, that's the type of stuff that can speed up everything. So it's that confidence that these tools are making a real difference. This technology is making a real difference. And by the way, even our revenue, like if you look at it, when we did our last quarterly results, the fact that we showed 40% growth in Azure, that is real numbers today and at a large base. So if we didn't have that, then, you know, it would have not given us even the confidence to do the outlay. So I do think at the end of the day, you rightfully point out that it has to make broad economic difference. The skill set needs to be much more ubiquitous. It can't be few companies in one sector, in one continent, having all the returns. It has to be a much broader phenomenon. Otherwise, it'll be a, quote unquote, a road to nowhere. But I feel pretty confident that we are on a path to this. But at the same time, it won't be linear, like nothing is, right? So therefore, that's why I think we are very, very disciplined. In fact, one of the key things we're doing with our capital as well is building a fungible fleet all around the world with all of the sovereignty controls, such that you can use it for training, you can use it for inference, you can use it for everything in between, across multiple generations. So you kind of have to have great discipline in how you invest, make monotonic progress, and then most importantly, make sure it is really being diffused throughout the world and throughout every company. Satya, we spoke about Sam and yourself, I'd be interested to hear from you. How is the relationship between OpenAI and Microsoft really? What has happened over the last years and what is the current status of it? Yeah, first of all, I would say, you know, one of the things that gives, you know, it's fantastic is from where we started, we had, you know, we basically... But there is competitive tension at the same time, huh? It's a partnership. Yeah, we started as, you know, it was a research lab, which was a non-profit, and we were able to sponsor their work and then benefit, you know, help them scale, and obviously we benefited. And so we feel very good. Obviously, there's competitive tension at the edges, but at the core of it, right? They have a very successful company at this point, in which we are big investors. We have IP rights, which we then can use to innovate on top, which is what we're doing. We have our own ambition to build our own models on top of it. We can do that. And so I feel very, very good. And by the way, in all of this, I'm glad that we created one of the world's largest non-profits as well. Now, the OpenAI Foundation next to the Gates Foundation are two non-profits that are associated with Microsoft. Is it actually true that Bill Gates warned you when you were about to do the first 1 billion investment not to invest in OpenAI? And if so, did he apologize meanwhile? Bill, like all of us, it was a crazy bet, I mean, in all honesty, to go in and say, hey, we're going to spend a billion dollars. At that time, it sounds like, oh yeah, obviously, every CEO of a large tech company should be taking these bets. It's not that easy, right? I mean, to go and say, because it's not, it was absolutely an unknown. And so he, like everyone else, was skeptical that the scaling laws on sort of something like text would work. And quite frankly, even in the first instances when it started working, he was pretty skeptical until I would say GPT 4.0. There was this very famous demo of GPT 4.0 that we did at his house where he saw it. And he had his own tests, all his, you know, I think they were all the MCAT medical tests and so on that he wanted to see. And that's when I think he changed because Bill was a much more, I would say Bill's core was always more symbolic logic. He was not a neural network person, right? He just didn't believe that neural nets could just learn. And he just wanted a symbolic logic to learn the relationships between people, places, and things. And so this was the first time when I saw him cross over. He'll still talk, by the way, if you talk to him ever, he'll still say this thing. You know, it goes back a little bit to my point about that cognitive core. So I'm sure Bill would still say, yeah, it's great, except it needs a cognitive core, which is symbolic in some sense. And so I think he's right, quite honestly, because you can't just say, hey, there's a machine that kind of does something that we are very thrilled about, but except it's unpredictable, it has all this jagged intelligence, because it's very problematic to deploy AI in mission-critical systems, if it makes a mistake sometimes, and you don't know when. Yeah. One question back to the relationship. I mean, from your perspective, it looks a little bit like, I assume, like you have two running horses, yeah? Some old man on the one end and on the other with this kind of little detail that one horse has to help the other. I have a hard time to understand how that could in the long run really work well. I think that the better way to maybe the metaphor is to think about it as we have a set of products, they have a set of products, and that's where we compete. But we share, I think, and the best way to think about it is the common engines, right? So it's going to take, you know, two car manufacturers that will share a chassis design, right? And there is scale economics because of the chassis design. We get to share, you know, for example, even at the system level, we are cooperating a lot, right? Even what we build out as the plans for these AI factories. And so therefore, it's kind of like, hey, we'll have two brands, we'll have two products. We have, by the way, in our products, we'll use both Mustafa's models and MAI. We have a fantastic team. We're going to grow that. We may even use other models from other frontier models too, because we're not beholden. We love open AI. We obviously have royalty-free access to open AI. So we want to maximize the use of that and building on top of the technology. But at the end of the day, we also have a lot of good ways to share, I'll call it the core engine parts and the supply chain of it. Are you occasionally having dinner with both of them or is that not a good idea? Yeah, we do. I mean, I think interestingly enough, you know, we've grown up and I've grown up in all of this, whether it's the partnership with SAP, which, you know, I was very much part of from Haase's time and Bill and Haase got together and said, hey, let's build SAP on top of SQL Server. So it was kind of like SAP built a very successful business. We built a very successful business, the Intel-Microsoft relationship. So I'm a student of and a cultivator of positive-sum partnerships. So I think of the open AI thing for all of its issues, which we have overlap, but think about it this way, right? You know, open AI competes with everybody today, right? It competes with Google, it competes with Meta, with Sora, it competes with, they're going to build a device, so they're going to compete with Apple. So at some level, I'm glad that open AI, to your point, as an investor, has got great ambition, is showing progress, but at the end of the day, Microsoft very much will be able to sort of go do what we need to do. But is that really a fundamental change in an AI economy that partnerships play a different role that almost everybody is partnering with everybody and there is a completely kind of new definition of competition in that context? I think that what is more important perhaps is what I said before, which is, I think when we partner, at the end of the day, you kind of want to know, because why do people partner? People partner to get leverage, right? Because in some sense, it's always the question of vertical integration, right? Everyone can start by saying, let me produce my own electricity to the output, right? But at the end of the day, specialization has a role and you could say, wow, as long as I can bring together partnerships and specialization to compose so that I can do something that's unique and I get leverage, right? That's the core thesis and it's positive sum. And so I feel like that's what needs to be found. So sometimes I think in whether it's in the tech, a lot is written about the tech industry and its partnerships, but foundationally, tech industry knows how to differentiate in a partnership. What I worry the most, quite frankly, is the next drug company, the next media company, the next auto company, right? How do they partner in smart ways with tech companies so that they remain sovereign in their capability? My biggest fear for sort of anyone who is not in the tech industry, because in the tech industry, we kind of have grown up with that DNA of how to partner and specialize and differentiate. I fear that it's not just even about data privacy, right? It's about being able to actually orchestrate multiple of these partners to build ultimately what I described earlier, your own AI capability, your own AI factory, because that I think is going to be the key, right? Just like how we don't have an electricity industry. Everybody uses electricity. AI will be, everybody uses AI to create value-added output. You mentioned it a couple of times today. Energy, electricity is a key factor and seems to become the scarcity. In Virginia, there is this plan of a data center that has a capacity of 47 gigawatts, which equals roughly 47 nuclear power plants. So how is your outlook? How fast, how convincingly is that problem going to be solved? And is it going to be solved with existing energy technologies or do we need new innovation on that front fast, like fusion and other stuff? I think that that's right. I mean, we need really, I mean, it's all of the above, right? I think we have to do the mundane stuff, right? Which is how do we permit better, faster? How do we build? It's not even just about how do we modernize the grid? Can we have faster behind the meter solutions, but that can come online and modernize the grid? There is how do we get even permits to build faster while at the same time making sure that it's compliant to all the regulations, including the environment. And, you know, at the end of the day, I think that this industry to which I belong needs to earn the social permission to consume energy because we are doing good in the world, right? Because at the end of that, it goes back a little bit to your previous question, right? It just can't be, hey, I'm going to consume four or 5%. The good news here is we start from a very low base, right? I mean, if you look at it at the end of the day, it's 3%, 4%, maybe 5%, maybe 10% by the end of the decade. Most of the growth is happening in this space. So that's why I think it's putting a lot of pressure, but still it needs to result in that economic growth that's broad spread in the economy so that we have the social permission to consume energy. And then to your point, we need the innovation in the sources of energy. We need innovation in how we utilize it with the best, I'll call it power efficiency. And so there's a lot of work that's going on in this space. In order to solve it, do AI companies need to become energy companies in a way? I don't think so. I think that this is another place where I think there is again going to be the coming together of a new value chain in which I think there is going to be great innovation and great specialization that will all compound, right? I mean, there may be one or two places where we may do something, you know, that's quote-unquote behind the meter that makes it sound like we're in the energy business. But I'm a big sort of believer that you just can't vertically integrate into every field which you don't belong in, quite honestly, and you don't have the expertise in. But to your point, but one way to think about it is our offtake of energy, you know, of any hyperscaler, may be a great stimulant for the next big energy entrepreneur or an energy company today. But what is then the energy technology that you count most on with regard to future developments? I think right now, quite honestly, in terms of we have some nuclear, you know, purchasing agreements, we'll see how that comes along. Obviously, we are currently in, you know, with all every form of renewable energy. In fact, we may be the biggest buyers of renewable offtake and PPOs everywhere, PPEs rather. And so therefore, I feel that we are going to not be dogmatic about any one. But we do need and the other thing I would say is when I look at my data center bills, it's interesting to see even the amount we spend in the grid modernization, because they are ultimately grid power and stable base power is going to be the most important thing. And so we need to do all of those. If we are talking about energy supply, we have to talk about Europe and particularly Germany, where there is a problem. We dropped out of nuclear energy and I think that is setting limits for your expansion plans to Europe, in particular to Germany, correct? Yeah, in fact, it's a good one. I'm not deep in exactly what our issues obviously are in D.C. You know, we are building in Germany, we will need energy in Germany, and it's something that we will rely, obviously, on German policy and German ability to create energy, because at the end of the day, that is going to be I think any country will have to sort measure itself on how can they be great producers or tokens per dollar per watt. The good news is we are happy to invest capital, right? So in some sense, you know, us and others will come build the data center so that the German economy can benefit from this. But the problem is you need the tokens per dollar per watt that needs to be competitive because pricing will be local ultimately. And so even German companies, if they don't get the best pricing, they'll just go to Netherlands or wherever the prices are there. Should Germany go back to nuclear energy? That's, you know, that's I'm assuming as much of a political question as it is a technological question. And so I would let the German people decide as opposed to sort of me. But I would say I think having energy capability, right, when I look at the United States, right, what is the current administration? And I think rightfully so, they have focused a lot on making sure that there is real energy coming online in order to make sure that, in fact, it is not like there are two things that they're concerned about. They don't want to increase prices on the citizens of the United States because there needs to be a new build for AI. That's one side of it. And then the second side of it is how do we build more energy for, you know, being a leading provider in the AI age? And I think that's what every country needs to do. The next years will be decisive. So if Germany does not return to nuclear energy and new technologies are not at hand, they will solve the problem immediately. The likelihood that we are going to miss out on AI and value creation in that context is high, isn't it? Yeah, I mean, I think there are a couple of things I think about, right? See, if you look at Germany in particular, it's a unique thing, which is Germany is competitive, not just because of what happens in the German market. It happens because of what is the intellectual property of Germany that is all around the world, right? I've always felt that Germany is successful. I mean, Germany is a big market. There's a domestic market. I think it should be a great producer of both AI factories as well as energy and what have you. But every time I go to a jeweler or a dentist, I'm surrounded by the German middle start. That's kind of my thing when I think about the brilliance of Germany comes through. And so that means Germany is much more of an export-led economy than I think is talked about. And so therefore, what is most important, quite frankly, for German competitiveness in the age of AI would be, can they parlay their know-how, which is world-class, mix it up with AI to create more of the world-class know-how and knowledge and breakthroughs and innovation that then are exported everywhere? I look at, for example, even our data center bills, Matthias. I mean, what Siemens is doing and how they're really going to be a critical part. They've built this digital twin with them. And I mean, it's fantastic to see the innovation that they are bringing to the market in order to, for example, be a solution that speeds up everything, like the chilling solutions we need. All of these things, I think, do lend themselves very well to, I would say, Germany's economic growth. If we speak about Europe, do you think Europe should aspire for sovereignty in AI or is that a false goal? I think that every country, whether it's at the European Union level or at the country level like in Germany, I think sovereignty is an important consideration, like it is for the United States, right? So every country would like to ensure that there is continuity of their supply, there is resilience in their supply, and there's agency in which how they operate. And that's one of the reasons why we have made all these commitments. So the first is we are investing in Germany, in the European Union with our capital, putting it at risk. So these are not AI factories or cloud factories that sit in the United States. They are in the continent and in the country. That's first. Second is we have now sovereign services, whether in the public cloud, for all of the data to be protected, encrypted, the data plane, the control plane to be sovereign. We even in Germany have a sovereign cloud. We have private cloud options. So basically, I feel whether it's on the technology side, whether it's on the governance side or even on the legal side, we've put in a framework which is all our commitments for sovereignty. But the one thing I feel, everyone's talking about this, which is very important, but I feel the new chapter of sovereignty is the one that I referenced earlier, right? It's got nothing to do with any things like this. It's about what is a German automaker or a German industrial company? How are they going to have their own AI factory and foundation model that is unique to them? That is, to me, the true definition of sovereignty, not like, oh, I have a control plane and a data plane that I can sort of push VMs to. That is not the, oh, I've encrypted data. That's all important. But if you have done all that, except you don't have your own at a firm level, it's not even a country statement, the ability to be sovereign in an AI world, then I think, you know, that's, I think, the new frontier of sovereignty. Let's be realistic in the AI arms race, Europe doesn't play a role. It happens between America and China, and you can debate who is going to win it. But in any case, Europe is not part of it. If that's the case, shouldn't then Europe decide very clearly for one ally to simply say, okay, then we are in the team US instead of trying to muddle through? I think that again is a geopolitical sort of call, which I think makes sense. And I think, yes, I think in some sense that by definition, the natural relationship across the Atlantic between Germany and the United States and so on, I think makes so much logical sense. But set that aside, I would slightly disagree with sort of your comment about Europe is sort of just got to pick. I think if you think back, I'm a big believer in the Jeffrey Ding, you know, school, which is all about diffusion of technology, right? Everybody today is more about the hype of one AI technology. Quite frankly, the country that is going to really win is going to be the one that can scale up broadly on AI, use AI broadly in their economy, in their health sector, in their manufacturing sector, in their education sector, and grow their economy, right? More so, in fact, it may turn out. The irony of all of this would be, it's not just the production side of it. It is the usage side of it and the diffusion side of it that's going to probably be the winners. And I think Jeffrey's book, you know, I forget the title of the book, but it's a great one to read because in some sense, his entire thing is, hey, history has proven that general purpose technologies are not about leading sectors and who owns the leading sector. It's about the diffusion. And so if you buy into that argument, I would say Germany or Europe could be the big winner as long as they do the hard work of actually getting the technology in, re-skilling, using that technology. And in fact, the next time the Gartner report comes and if the Gartner report says, oh my God, the companies that seem to have the most success rate in using AI are German companies, that's the day when you can go long. Would you take a bet who is going to develop artificial super intelligence first, China or America? And would America's chances to win that be higher with Europe or doesn't make it a difference? I do believe that I always felt like the United States has been super successful because of its allies and its relationships all around the world. And most importantly, the trust the world has in American technology and American institutions. And so the United States is what, 4% of the world's population, 25% of the GDP and 50% of the market. And that didn't happen because of just the U.S. market. It happened because the U.S. companies and U.S. innovation reached all the parts of the world and there was real trust in it. So to me, I do believe the relationships, the market access, the trust between European Union and United States is a pretty important thing. But I mean, trust alone is not going to do it. And the fact that China has a population of 1.4 billion, they have access to all data and behavioral data of their population. The only limitation for them is basically what kind of dilutes their control of their own population. Everything else will be done. So you could easily make a case that it is more likely that China is going to develop that first and with that have a dominance. Well, I look at this and say, at the end of the day, you know, here we are in 2025 and, you know, every sort of benchmark, I mean, first of all, we have to got to give credit to China for all the sort of things that they're doing in the open source world in particular. And I think we should continue to assume, after all, the human capital of China is just tremendous and they will continue to innovate. But here we are where United States still continues to lead, right, whether it's on the AI systems or whether it's the frontier models or the AI products around the world. Right. And so to some degree, time will tell. But I would still be, I would bet on the ingenuity of the American sort of tech sector. But I would also say that it's not just the ingenuity of the American tech sector, but it also has to be about the, I would say, American tech stack being the most trusted tech stack in the world. Because if you go back to my fundamental point, that look, at the end of the day, this is not even about a capability. It is about ultimately who do I trust to help me have sovereignty? That's it. I feel like every CEO and every country leader has to ask one question and answer it crisply. Who do I trust to sort of have as a partner who in the long run is going to make me successful? And when I think about Microsoft, that's kind of opposition. I think when I think about United States, that should be opposition. Satya, you said that the more important technology gets and the more basically we delegate to bots and machines, the more important becomes empathy. Is that one of the reasons or the main reason why you are now basically bringing people back to the office at Microsoft? No, I mean, look, first of all, it's very interesting in this age of AI and automation and autonomous agents and so on, it's even more important to recognize how humans collaborate. In fact, there isn't a substitute to workplace as the best collaboration tool. It's so fascinating. So I think at the end of the day, I do feel we have all learned a lot of what it means to collaborate over distance and remote and what have you. But there is a real sense that something gets lost when you don't bring people together. When I think about what I describe as the new production function, how do you get the product people, the science people and the infrastructure people to really collaborate in new ways to build that? Because this is one of the fascinating things. AI has not learned the context engineering that is needed, whereas humans do. I can pick up social cues. I can pick up emotional cues, bring those things together in how I accumulate knowledge and I collaborate and innovate. And so I think, yes, the physical space is important, but it's not like I'm not so dogmatic that we were all remote once. We were never that. We're not all in person. We always have a lot more flexibility. Empathy, to your point, requires us to understand the context a little better, but at the same time, not be too dogmatic on the two ends. But it really seems that people want to interact in person more than ever. Conferences are booming. Social events are booming. Life events are booming because it is about that interaction. So do you see that as a kind of general trend of civilization that EQ becomes more important than IQ because IQ is more and I've always felt like, you know, I think, yeah, I think, you know, basically IQ, EQ, and I think Mustafa likes to even talk about social intelligence or SQ. I think all these things I think are important in human society. And in fact, if anything, if you're going to have these AIs that can then interface with us, they also need to develop a better sense of how to interface with our needs across these three. And so, yes, I would subscribe to it. I mean, IQ has a place, but it's not the only thing that is needed in the world. And I've always felt at least as leaders, you know, if you just have IQ without EQ, it's just a waste of IQ. Satya, we have to come to an end, but one more question. Is there anything that nobody knows about you that you could share with us now? Nobody knows about me. God, that is a hard thing. What would it be? Can be a strength or it can be a weakness or it can be just an interesting fact. The one weakness I have is for these Indian sweets that, oh my God, like I cannot resist. If anybody gets me anything and leaves it, you know, it'll be gone. Satya, if you had to do an interview with yourself, which would be the question that you would ask yourself that nobody so far ever asked you? Why is it that I start so many books but never finish? Next time we speak about the unfinished books, Satya, thank you very much for taking the time. Love it. Thank you so much. It was really fun. Cheers.