Microsoft Chairman & CEO: AI, chip shortage, empathy, and poetry

Satya,

you are the CEO of the most valuable company in the world, what's on your mind these days?

And so here we are on what is essentially a complete new platform shift that we are in the midst of, right? I kind of say it's my 32nd year, but it's year two of my fourth platform shift. And so what's on mind is, okay, what is this platform shift really all about? And as a company, can we be all in and innovate, right? I mean, at the end of the day, when I say there's no franchise value, it also means that you get to play for it all up again, even the battles we won and the battles we lost are all up for grab again.

And so therefore, there's a freshness to it. So what's on my mind is that ability to ground myself back yet on another platform shift, and it's exciting.

But at the same time, hey, who wants to be in a business where you don't get to reinvent yourself again and again?

Do we have a breakthrough in healthcare driven by AI? Do we have a breakthrough in material science and energy transition because of AI? The list goes on, right? So to me, that is fundamentally the way I think about it, which is, I think that one of the things that might be most important for us is to consider how a general purpose technology, somebody was telling me this, right, which is in the height of the Industrial Revolution. in the United Kingdom, they spent 10% of their GDP building the railroads.

And obviously, the railroads is not about the railroads, it was about the entire economy of the United Kingdom. And so something like that, I think, is what, that's the unit of analysis, at least for me, as to how tech and its future will impact the broader society and economy.

I mean, you know, one of the things that I've always looked at over my years at Microsoft is look for high ambition technology, innovation companies and partnerships, like whether it's Intel and Windows came together and that was successful, SAP and SQL Server, so it came together and was successful. So I'm always looking for partners that we can innovate with. And that's what I found in Sam and team. And at that time, it was not like. it was a real shot in the dark, right?

It is not like, oh, wow, this is a sure, sure thing. Everybody now talks about it as if, you know, this is the issue with tech, right? Which is long before it's conventional wisdom, you have to be all in and hope it works. And this is one of those things where we backed it long before it was conventional wisdom. And here we are.

But I don't take, you know, like there's going to be severe amount of competition. You know, Google is a very competent company. And obviously, they have both the talent and the compute, and they have, you know, they're the vertically integrated player in this, right? They have everything from data to silicon, to models, to our products and distribution, and there's others as well. And so, yeah, we will have a significant amount of competition.

And I think, if anything, Microsoft's partnership with OpenAI is bringing more competition to. otherwise what would have been a default. Google should be the default winner. And if we partner well and we innovate well, we can bring some competition to them.

So, that's kind of the first thing that we want to do, which is the best work in being able to build the infrastructure out for both training and inference. And then the next layer up, we want to have the entire data tier, right? So, you can imagine as these models and model capabilities become more capable, I think the data tier will be completely redone, right? We've talked about the retrieval augmented generation already. You have all these things, whether it's embeddings, vector search, how do you chunk data such that retrieval augmented generation can work well.

So, that's an entire layer. Or when context lens become bigger, that's a different sort of data layer, like what's the throughput between data and your inference fleet? How do you sort of think about that? So, therefore, we will innovate on the data layer. And then, of course, on top of it, is where we will innovate on our co-pilots.

One of the first products we built was GitHub co-pilot. In fact, my entire confidence in this generation of AI started when I started seeing from GPT-3 to 3.5 and that implementation in GitHub. And so, we now have not only GitHub co-pilot, we have co-pilot for all knowledge work in Microsoft 365.. We have co-pilots for these functions, whether it's service or sales or finance. So, we're going to innovate in our app layer on our own.

And so, that's, I think, fundamentally how I look at it. It's a full stack approach. And each layer, by the way, we will innovate, we will have partners, we will have others innovating, there will be competition. even. It's not like one of the things of being a platform company.

is you got to be comfortable with many third parties competing with you on different layers? because that, to me, is core. Otherwise, you kind of try to do everything in a monolithic way. And at least what we have learned over the years is the best thing to do is to keep each layer competitive on its own.

I think one of the interesting things is each one of these built on the previous, right? So, I don't think the web would have happened if there was not a ubiquitous PC. After all, the first time I saw Mosaic, you know, was as a browser on top of Windows, and then Netscape came about, and then IE, and what have you. So, therefore, I think you sort of see each one of these births, the next, and then it goes beyond what births do, right? That's, I think, the real thing.

Right now, we're seeing that, which is the cloud, as we know of it, and mobile and PCs on the edge have really birthed the AI age. The question is what happens next, right? Which is, does it go beyond that? I think there's going to be AI that is not just about cognitive work, AI that is also going to accelerate science, so I think that that's an exciting space. AI that is going to be embodied in the real world, so what may happen in robotics is an exciting space, and so there is a bunch of things that are going to happen over time, but clearly, yes.

This is year two of a complete new paradigm that obviously reinforces what happened previously, or builds on what happened previously, but also showing early signs of what happens next.

Is there going to be another model architecture breakthrough or what have you? I think that one has to see. It's unclear, but we definitely are not going to bet against scaling laws, nor are we going to say that this is the last model architecture breakthrough. In fact, a great example of it is even what we've seen. even with our small language model innovation like PHY, which is, we are able to create what is significant capabilities in a small language model, which doesn't require, obviously, the same amount of compute, which is just like attention is everything.

Attention is all you need. Textbooks is all you need. That sort of intuitively speaks to how I think learning can happen. Therefore, I would say, yes, if more capacity there is in the world, the chances are that we will be able to make progress, but at the same time, I wouldn't discount a real breakthrough in model architecture that doesn't perhaps require the same type of compute. That's why I think, at least I want to keep myself open-minded about it.

There are many things one could do, even on top of foundation models, like an OpenAI model or a Mistral, or what have you. There are ways you can add unique knowledge, unique value, on top of even what exists, before you go off and say, let's build all the compute, all of it, and do the same pre-training run. There's nothing stopping any country from doing any of this, but the question is, what is the value add? Therefore, I think that you start from the value add and then back into whatever is needed for it.

Whenever I am, whether I'm in Norway or I'm in Jakarta or New Delhi or wherever, I am always grounded on. fundamentally, the fact is, are we able to make – look the local politicians and political leadership and society at large in the eyes and say that we contributed to their public sector becoming more efficient, or large multinational companies in the region becoming more globally competitive because of some tech input, education outcome, health outcome, small businesses and their productivity. At the end of the day, your social contract with the country comes from your ability to contribute to their local progress, and that – so I don't think we can ever be beyond. Geopolitics will exist, with or without us, and our goal has to be how do you participate and have permission to operate.

I mentioned science, right? One of the things I look at it and say is in order to make progress on science, you need great. in silico simulation. Quantum is the ultimate breakthrough, right? So, when we have a complete new system architecture that breaks free of the Von Neumann limitation, you are then finally going to be able to simulate something like the dynamics of a cell or a molecule, right?

So that, I think, when you can do that, then everything else, even in terms of biology or what have you, becomes more feasible. The interesting thing is, AI is kind of like an emulator of that simulator. In other words, you can kind of simplify the search space, and we see this already, Nikolai. One of the things we did recently was we have a model for material science, which we used to generate a new novel compound, which we then went and manufactured. We worked with the Pacific National Lab locally here to effectively go reduce the lithium content by 70% in a new battery material and produced it, right?

Not just conceptualized it, but simulated it, produced it. And so to me, something where quantum plus AI, I think, can be the ultimate accelerator of science. And we are making progress. Even on quantum, we are taking a very full stack approach. We have our software stack with our Qsharp, which is our quantum programming stack.

So we are excited about the progress we're making on quantum and how it complements AI.

We have cloud gaming. We have console gaming, PC gaming. So we are a full stack game publisher, as well as a game systems provider. And so our goal there is, one, we're in gaming for our love of gaming, right? So I always sort of say we should never be in businesses as a means to some other end.

It has to be an end. Otherwise, it's not a business. So to me, gaming is something where we want to bring joy of gaming. That's the one pure consumer entertainment category. I love the fact that gaming on a secular basis is probably going to be, if not already, the biggest entertainment category out there.

So that's one. And of course, it has real implications on the rest of it, right? If you think about, even, remember, it's interesting. I've not talked to Jensen about it, but one of the greatest sort of successes of GPUs was fostered by innovation in gaming, right? DX, which was the Microsoft graphics stack, is what made GPUs an accelerator, right?

After all, the GPUs were created for PC gaming. One of the things that we want to use AI is to be able to find these bugs in these even closed worlds before they're out there. And so, therefore, we have some very great use cases there. But beyond that, I think gaming as data in the context of some of the innovation in our models, I think is going to be important.

So one of the things I look back, even in my career at Microsoft, when we first became the largest market cap company, I forget, I think in the early 2000s, I think we crossed G, you know, people on our campus were walking around, including me, thinking, oh, we must be God's gifts to humankind because we are so brilliant. and what have you, except what we needed to be grounded on that day was to say, wow, we now have a real responsibility to reground ourselves, to innovate again, so that we are relevant in the future, right? And so that's why I was lucky enough to have read Carol Dweck's book on mindset, which is around child psychology, called Growth Mindset. And I love that book. I read it more on the context of sort of our children's education, but I must say I got educated because I felt like this is what makes individuals, children in school, it's very clear, right?

It's better to be a learn-it-all versus a know-it-all, because even if the know-it-all has great innate capability, the learn-it-all, you know, even if they start from behind, they will surpass the know-it-all, right? That's sort of, you know, true for children in school. It's true for CEOs in my seat. It's true for companies. And so we took that approach, Nikolai, we said, let's be a learn-it-all versus a know-it-all.

And the day you say you've achieved that cultural transformation means you're a know-it-all. So therefore, it is a good way to sort of say every day you make a bunch of mistakes, you at least have the courage to acknowledge them and continue on it. And so it's not a destination you ever reach.

So that's why I, you know, I credit more of this work by Carol and team and so on, because it's not like, you know, it's not like, I don't think anybody at Microsoft views growth mindset as a Microsoft dogma, or definitely not Satya Nadella dogma. It is something that speaks to them as humans, right? Which is. it's good for them as friends, colleagues, partners, parents, neighbors. It integrates work and life.

They can bring their own personality and passion to it and benefit from it. Right. I always say, like, you're not like this. Microsoft. culture of growth.

Mindset is not for Microsoft. It's for you. And if you, you should only practice it if you feel like it speaks to your own. thriving at Microsoft and in life. That's, I think, what I attribute it to.

Right. It wouldn't have taken off if it was just another thing. There is a top down slogan. It is. I always have believed that, which is ultimately people work and find meaning in work only if they can find some true, deeper meaning for themselves.

And so that's, I think, what I've been always trying to invoke.

Why is it so important for you?

Innovation comes from us being able to drive the solutions to unmet, unarticulated needs of customers out there. Right. So the key being unmet and unarticulated. So that means you have to have a better sense. when you're even looking at some log data or, you know, some customer interview data or whatever.

You know, it's not just the words that they're saying, but you've got to be able to walk in their shoes. And the good news there is this is innate in us all human beings. We have the ability to empathize with the other person. In fact, design thinking is that right. So when people go and say, let's do learn about design thinking, design thinking is applied empathy.

And so to me, that's what I pull the thread on, which is let's not think of empathy as something that, you know, is just a soft skill that you reserve for your friends and family. But I think it's at the root of all innovation. It's about being able to meet the unmet, unarticulated needs that comes from your innate ability to have curiosity, to learn about others, walk in their shoes, innovate on their behalf. And that, I think, is, I think, what we have to do. And that's why I think empathy is an important, important skill for all of us.

And but, you know, when he was born, it was a lot about what happened to me. I was sort of, you know, essentially, quote, unquote, you know, all about why did this happen to us? Why did it happen to me or how? And then I realized, after watching, in some sense, my wife, who was there as a caregiver, as a parent, you know, taking him up and down Seattle to every therapy possible. Quite, frankly, you know, it took me years, not, you know, days or months or weeks.

And and then I realized that nothing had happened to me, but something had happened to my son and I needed to be there for my son. And that is the experience I talk a lot about. But there's experiences like that every day, right? Some colleague of mine comes with some, you know, parent of theirs who needs care. Right.

That sort of, you know, I learned from it. I am or at least let me put it this way. I am more attuned to learning from other people's experiences today than I was in the beginning of my career. And I think that happens to all of us. Right.

Which is life's experiences. If they accrue that ability to build a deeper empathy for other people. And that also helps you be a better manager, a better co-worker, a better innovator.

Right. You have to be accountable for making real progress. One of the things I think about, why do businesses exist? Businesses exist. At least.

I like this Colin Mayer definition that you have to create profitable solution to challenges of people and planet. Because that's at least a good way to allocate the global resources that are available. Right. The profit motive is a good motive, because it means you're competing and allocating resources in the most efficient ways and face competition. And so, therefore, you have to have great execution, have to have great accountability.

And so I think of empathy as a necessary condition to create great solutions that are profitable and that are competitive, solutions that are winning in the marketplace, as opposed to somehow this being a trade-off.

. I don't know I have causality here well understood, because, quite frankly, it's so much easier for others to opine on this than or other. It's just really for others to judge and assign causality there. But the way at least I come at this is I don't start with, what am I good at? I am very keenly shaped by.

what am I not good at? In other words, I'm always looking, what can I learn from someone else? If there's one attribute I have, I don't start each day by thinking, oh, all the stuff I know and I'm good at. I'm like, wow, what am I weak at? Whom do I talk to?

Whom do I meet? How do I really shape the colleagues around me who have better skills than me on many fronts? That's what I'm wired. Maybe that helps, but I don't know whether that's the causality. But I don't start each day with saying, wow, I'm so good at this, so therefore I'm going to go do this.

No, I come at the exact opposite, which is.

And I don't come at it with paranoia, right? I mean, I don't like paranoia. I like this. That's why I go back to my own words for this. That's why the growth mindset, or confronting your own fixed mindset, having confidence.

And wow, what an unbelievable world we live in that every customer can teach me, every partner can teach me, every colleague can teach me. Like, what more can I ask for in life? So it's not paranoia. It's not like, oh, wow, we have to go in every day that if I don't learn something, I'll fail. I'm more about like, what do I learn so that I can innovate?

Maybe that's how I come at it. And that's right. Bill and Steve, in their own unique way, had that mindset. And so I've grown up around it.

And from ancient Greece to modern Silicon Valley. And so that's why I think you have to sort of really get that calibration, that you've got to have some confidence in your own capability.

And that's essentially how that's my framework for life as well.

more is when I sort of felt like, wow, poetry is basically natural language compression.

And it is able to describe, you know, it's a model of the world in the most succinct form. And so there's, and I got into Urdu poetry in a big way in my mid 30s. And so I grew up in Hyderabad, where obviously Urdu poetry was in the air. And now, of course, I love it. But even, you know, late, you know, the American poets, the English romantics, Germans, they're fast.

I mean, like, so I I'm, at least I'm not, I wouldn't say I know much poetry, but I at least I'm fascinated by the ability of the human mind to compress thought, whether it's code or poetry.

And I'm not saying you shouldn't have ambition. You shouldn't strive for your next promotion. You shouldn't put advocate for yourself or have others. You absolutely should do all that. But at the same time, really, my advice would be also to take the job you have at hand and do a wonderful, you know, go at it with all of you, the vigor and all of the energy, and also define it as broadly as possible.

Right. I mean, that is perhaps one of the things when I look back at it, I never defined my job narrowly. And that, I think, was both very satisfying in the moment and it helped, I think, land me the next job. And so that is my perhaps the one advice I would leave people with.