Sundar Pichai, CEO of Alphabet | The All-In Interview

We're sitting here at the Googleplex with the CEO of Alphabet, Sundar. Thanks for being here.

Great to have you here, David. I look forward to it.

Is Google at risk of being truly disrupted by AI?

We've recently been testing it in Labs. There's a whole new dedicated AI experience called AI Mode coming to Search.

OpenAI has Sam, xAI has Elon, Meta has Zuck, and Microsoft has Satya. Are you willing to share your perspectives on those 4 competitors?

I think maybe only 1 of them has invited me to a dance, not the others.

What's your biggest regret?

Look, there are acquisitions. We debated hard and came close.

Just give me 1 name or get in trouble.

Maybe Netflix.

We just leaned into the user experience, and over time we figured out monetization to follow. It's like 1 of the original principles of Google: follow the user. All else will follow.

What's going on? I'm really excited for this conversation. You and I started working at Google on the same day in 2004.

I didn't quite realize that. Same Noogler class. We had the hats on that same week, on the Friday all-hands.

I'm now a podcaster. You've done a little bit differently.

You're more than a podcaster, but you're very good at podcasting.

Well, I appreciate it. I respect the other stuff you've done as well. So, no, I appreciate it.

In your tenure at Google, you ran Chrome, Chrome OS, Drive, and Google Maps, and it's been 10 years now since you've been the CEO here at Google, now Alphabet. Amazing, and congratulations. Under your tenure as CEO, the stock has gone up by 4.5x, to a $2 trillion market cap today.

You've grown revenue from $20 billion a quarter to nearly $100 billion a quarter. It's been a really incredible run to see someone who started as a PM grow their way into this incredible role. So, congrats. How have you liked the job?

No, look, I love building products. In some ways, Google was really set up—I think the founders set up this deep computer science approach—and you take that and apply it to build things that can impact people on a day-to-day basis. So, it's that kind of product and technical culture which is the essence of the company, and I love doing that.

There's not a single week which goes by where I feel like I don't get to do that. Those are the parts I really enjoy. But obviously, running a company of this scale, where you impact so many people, is a privilege. So, I've enjoyed every part of it.

You're at a pivotal moment in the company's history today. Have you read The Innovator's Dilemma?

I'm obviously very familiar with the concept. I don't think I've read the book, actually. But it's 1 of those things which is so much in the ether—you think you know it.

I say it in jest because that's the talk of the town: the talk on Wall Street, the talk in Silicon Valley. Is Google getting disrupted in this moment?

AI seems to create a fundamentally different paradigm for human-computer interaction. Consumers are asking AI questions through chat interfaces. They're getting complete answers. They're engaging with AI systems in a way that they traditionally didn't with the classical Search interface.

Is Google at risk of being truly disrupted by AI? Is the core Search business—which, with the ad revenue, is about a $200 billion run rate out of $360 billion of your total revenue, and most of your profits—at risk? It seems like Google is in a really challenging quandary where, if you disrupt yourselves too quickly, all of that revenue can go away. It can be really impactful.

So, is Google being disrupted by AI at this moment, or is Google leading?

It's a good framework and a good question to talk about. For almost a decade, 1 of the first things I did was to think of the company as AI-first. It was very clear to us. We had Google Brain underway in 2012, we acquired DeepMind in 2014, and in 2015, when I became the CEO, I said, "Look, the technology is really evolving."

The reason we were excited to approach our work as AI-first is because we really felt that AI is what will drive the biggest progress in Search. I think even in the last couple of years, I've viewed this as an extraordinary opportunity for Search.

If you look at how much information means to people, each person is going to have access to information in a way they've never had before. So, it feels very far from a zero-sum construct to me, and we're seeing it empirically when people are using Search.

Obviously, there are a couple of major things we've done with Search. Transformers drove some of the biggest innovations in Search, with BERT and MUM dramatically improving Search quality. We launched AI Overviews about a year ago. It's now being used by over 1.5 billion users in over 150 countries.

It's expanding the types of queries people can type in, and we see it empirically: the nature of queries is expanding, so there are whole new use cases coming into Search. For queries where we trigger AI Overviews, we see query growth, and the growth continues over time.

Getting the feedback from AI Overviews, we've recently been testing in Labs a whole new dedicated AI experience called AI Mode coming to Search. We'll speak about it more at Google I/O. In AI Mode, you can have a full-on AI experience in Search, including follow-on conversational queries, and we're bringing our cutting-edge models there.

The models are actually working to answer your questions using Search as a real, native tool. There, the queries people are typing in are literally long paragraphs. The average query length is somewhere between 2 and 3 times what we see in Search as it existed 2 years ago.

So, we are seeing people respond. Search, from the outside, looks easy to do. The craft of Search is very hard. Over 2 decades, I think we've had a real north star of understanding what users want in Search.

You've been here; we're a very metrics-driven company. We know what works. Users are our north star, and empirically we see that people are engaging more and using the product more.

So, all that to your question about The Innovator's Dilemma: I think the dilemma only exists if you treat it as a dilemma. In technology, you have these massive periods of innovation, and you lean into it as hard as you can. It's the only way to do it.

When mobile came, everyone was saying, "You're not going to have the real estate. How will ads work?" Mobile was a transition which ended up working great. I can give great examples. TikTok has come in, and YouTube has thrived since the moment TikTok came in. It was a whole new format.

We did Shorts when we launched Shorts. Shorts absolutely didn't monetize anywhere near long-form, but we just leaned into the user experience, and over time we figured out monetization to follow.

To me, you don't think about it as a dilemma. Users—you have to innovate to stay ahead, and you lean in that direction. It's 1 of the original principles of Google: follow the user. All else will follow.

I think the "Google is dead" disruptor narrative, as you point out, has been repeated a number of times. Today, people are pointing specifically—and I appreciate your points about there being new Search experiences coming. The Search experience, it sounds like, is going to evolve.

As people look at standalone apps, they compare Gemini as a standalone app to ChatGPT and to the Meta experience. The stats that came out in recent court testimony, which had some data revealed from March—I don't know where the data came from—said the Gemini AI app had 350 million monthly users, compared to ChatGPT at 600 million and Meta AI at 500 million.

Is that the wrong way to think about it—that the Gemini standalone app isn't the future or the AI bet that Google is making? It sounds like there's going to be much more of a kind of tight integration into how the Search experience evolves and what happens to Gemini in Search.

Maybe the most widely used GenAI product today might be Search with AI Overviews. People are using it intensely. Obviously, we have a standalone Gemini app.

I think we're making progress there. Particularly with the introduction of Gemini 2.5 Pro, we've seen a real uptick in engagement and usage growth in the product. We have a lot more to come.

Just in the last few weeks, we've shipped Deep Research and updated Canvas and Audio Overviews. You can now generate video with Veo 2 straight in the Gemini app. On Android phones, with Gemini Live, you can screen-share, and it can talk about what's on your screen.

There's a lot coming that way, and users are responding. ChatGPT has obviously had phenomenal success, but I think it's still early days, and we're definitely seeing traction and growth.

To me, what matters is: if you innovate, are users responding and using it more? That seems to be the case, so it's in our hands to continue innovating.

I think it's a fiercely competitive moment, but across our products, people are coming and using and consuming information across Search, using the Gemini model increasingly in YouTube, in the Gemini app, and so on. So, I think we have a much broader view.

If I were to think about the unit economics of Google's business...

There's a cost to serve a search query, and there's revenue per search query—ad revenue per search query. How is that number changing, or how will it change in this evolution in search toward more of an AI interface? I've got to assume that to serve an AI-driven query is much more expensive than to serve a search query.

Look, this is something people were really worried about two years ago, but I've always felt that, to the extent that something is about the cost of serving it, Google, with its infrastructure, I would wager on—and on our chances to do that better than pretty much anyone else. We've actually seen that, for a given query, the cost to serve that query has fallen dramatically in an 18-month time frame.

What is probably more of a constraint is latency, I would say. It's less the cost per query; I think it's our ability to serve the experience at the right latency. Search has been near-instant. So how do you think about that frontier? That's been more of a question. The cost per query is not what I think will end up being the constraint. I think we'll be able to handle it; we've done the transition well. That's not a primary driver of how it'll impact things.

Do you have a point of view on ad revenue per AI query?

We already, with AI Overviews, are at the baseline: it's the same as without AI Overviews. We've reached that stage. But from there we can improve, right? I've always felt the reason ads have worked well in search is because commercial information is also information people, when they have that intent, are looking for—the most relevant information.

So I don't see any reason why AI, just from a first-principles standpoint, won't do a better job there as well, right? And so I think we're comfortable that we can work the transition through. Some of it may take time, but all indicators are that we'll be able to do it well over time. It's already at the baseline when we show ads with AI Overviews.

Do you feel that pressure on Wall Street and the board? What's the tension that you feel as a leader in trying to manage this transition in the product and the revenue model for an organization of this scale? I don't know how many leaders have done it successfully in the history of business. Where do you feel the tension? Where do you feel the pressure? And how much leeway are you being given by the founders and the board to do what's needed here?

Two things. The main thing is that it's a moment of acceleration. If anything, the good thing about these moments is that you don't even have time, a lot of times, to think about some of those questions. I think a lot about making sure we have the best models. Are we pushing the frontier as a company?

I think the last few months have shown the breadth and range of what we are doing. We are there, and we have to continue to stay there. For me, you think and worry a lot more about execution from within. Are we executing? Are we moving fast? Are we innovating? Over the past 12 months, I think we've really picked up pace as a company to meet the moment. That's where I do spend a lot of time.

Look, as a CEO, one of the first things I did in 2015, in addition to being AI-first, was to really bet big on turning great products like YouTube, Workspace, and Cloud into robust businesses, as well as great products. Last year, we exited a combination of YouTube and Cloud at $110 billion. I think people don't internalize that Google is one of the largest enterprise software companies in the world now.

And the largest media company, in some ways, right? We're doing a podcast, and I think we're the largest podcasting service in the world. So I feel like, as a company, we are set up well. For the first time, you have this cross-cutting technology. Thinking of us as a deep computer science company, what better technology than AI, which horizontally can impact all aspects of our business—Search, YouTube, Cloud, and the other new things we are doing?

So it feels like an exciting time. Not only have we continued to do well in Search, we are doing well in these other businesses. To me, it feels like one of the biggest opportunities ahead as a company, too. I think the next decade ahead looks to me as exciting as the past decade.

As I think about my time at Google, right below us in the garage, they were building these super-secret shipping-container data centers. They had these data centers in a box that you could ship anywhere, as long as you had access to water and power. You could connect to the internet, and you could scale data center capacity all over the world. That was 20 years ago.

It's always seemed to me that one of Google's core, and not well-understood, advantages was its infrastructure advantage—something that Google has invested in at its core from the beginning. Can you tell me a little bit about where you view Google's infrastructure advantage playing out in the AI competitive landscape today? How does it translate into cost, speed, and product quality? Where do you guys think about investing the $70 billion of capex this year—in the chip layer, in networking, and in the data center?

We can unpack both—where our capex is going—but on your first point, one of the ways we look at the Pareto frontier of performance and cost, Google literally is on the Pareto frontier. We deliver the best models at the most cost-effective price point, right? The Gemini Flash series of models are a real workhorse in the industry, right?

Part of why we are able to do that is because we train and serve our models on our infrastructure, including TPUs. We are in our 7th generation of TPUs, and we built our 1st version in 2017. I remember talking about it at Google I/O. Probably people didn't pay attention to it because they were asking, "Why are you building a specific machine-learning-accelerated chip?"

It plays out everywhere. To your earlier question on cost per query in Search, the reason we feel comfortable we can serve it at that scale is because we are constantly innovating through each generation, including chips, which are really, really good at inference. Ironwood, which is the latest in our TPU series, has a single pod that is over 40 exaflops. The scale of these things is incredible.

We've thought about it all the way from subsea cables to the scale at which we do infrastructure, which is unparalleled. I've always viewed that full-stack approach—deep infrastructure, foundational, fundamental R&D on top of it, and then you build and innovate on top of that—and I think that approach will serve us well over time. Empirically, it really plays out in the cost at which we are able to provide our models.

Part of the reason we've had a lot of traction with the Gemini 2.5 series is not only are they great models, but we are offering them at a very attractive value. We can do that because we are driving our infrastructure cost down.

On the $75 billion in capex for 2025, obviously the majority of that goes into servers, data centers, and so on, with servers being the vast portion of it. I would say, looking at the compute part of the spend, half of that is going toward our cloud business in 2025, and obviously that is a very different business from Search and so on.

A lot of it is to power the innovations from Google DeepMind pushing the frontier, and we're doing it across many dimensions—not just large language models, but even there, doing it across not just text, images, and video, but also building world models. There's just a lot of innovation that we are pushing on the frontier, obviously to support our core products like Search, YouTube, and Gemini. But 50% of the compute goes toward Google Cloud.

Let's just talk about chips for a second. This is a big part of the conversation: NVIDIA has the real market monopoly in AI, is what everyone says. Do TPUs provide a wholesale replacement for your need for NVIDIA in the supply chain, or is NVIDIA still a core part of the mix in the data center—for training versus inference, in large language models versus other models? Maybe just share your understanding of where the mix evolves to for you guys.

First of all, at a high level, NVIDIA is a phenomenal company. Jensen is awesome. We have been working with NVIDIA now for a very, very long time, and we continue to do so. We serve a lot of the Gemini traffic on GPUs as well, right? And so we give customers choice.

Internally, we train our Gemini models on TPUs, and we serve them that way across our products. But we use both, and I do think everyone in the industry is going to try and do something like that. NVIDIA's R&D, its ability to drive innovation, and its software stack are world-class. So they have a lot of advantages as a company, and I have extraordinary respect for them.

We are committed; we are actually deploying GPUs internally as well. I like that flexibility, and we're also long-term committed to the TPU direction. So I think it's a good combination to have both, and I think we push each other and drive the frontier forward.

Just going back, there’s an infrastructure advantage inherent in all of the investment that’s been made for 20-plus years and the continued investment. A lot of folks have said that some of the performance in foundational LLMs is starting to plateau, and as a result, we’re seeing a less differentiated landscape amongst the competitors, and that should be a consideration for Google. That’s the outside narrative.

Can you share a little bit about—and then I want to come back to non-LLM models, where there are other advantages for Google in a minute—but maybe just on this point: how much more of an opportunity to continue to evolve LLMs is there? Where does Google’s advantage lie in maintaining better performance in the models over time?

I think maybe it was Andrej Karpathy who used the term AJI, which is like—he called it artificial jagged intelligence, right? So I think the progress is not going to be always smooth, right? You go through these periods: it looks like something is slow, and then you see a paradigm breakthrough, et cetera. It’s been going like that for a while.

Obviously, over the last couple of years, all of us scaled up on pre-training, and then there was a lot of momentum with post-training and then with inference compute. Now there’s progress with how you take all that and stitch it together in agentic workflows and so on. So I do think there’s a lot of progress, and it feels pretty continuous to me, right?

I think it’s both true that progress gets harder, which I think will distinguish the elite teams, at least on the foundational side. I think that might be a factor. I feel that the harder the problem is, the better set up we are for that. I think we are well set up for that.

I do think we are pushing the research frontier in a much broader way than most other people, beyond just LLMs and transformer-based models. I mean, diffusion-based models—all those areas—we are exploring in a deep, deep way, right? There’s always the chance that we may reach a point where you don’t quite get the returns from the additional compute you’re going to put in, but I quite haven’t seen it yet.

The progress looks maybe harder because you’re now dealing with a lot more compute. So you’re really running into the limits of, like, can I actually get as many electricians as I can to build the data centers at the speed? All that stuff. But I haven’t seen—or at least, talking to our researchers, I haven’t seen—anything fundamentally saying, “Hey, we’re not going to be able to move past this point,” or something like that.

Does Google have a data advantage with YouTube or other products or services? Are you able to train on that data in a way that others can’t?

I think we have the opportunity to create much better experiences for people. People use products like Gmail, Calendar, Docs, YouTube, Search, et cetera. So, with their permission, taking that personal context into account, I think we can deliver much better experiences. We are working on that, but it’s something on which we have to deliver.

I view that as one of the differentiated innovation opportunities we have ahead as a company, but it’s something we’re thoughtfully working on. We’ll make progress there.

That makes a lot of sense. If Search evolves—and I’ve been using a lot of voice AI tools—I find them incredible. I can have a conversation, access the news, dive deep on a topic. It’s just so incredible.

What do you view the future of human-computer interaction being 5 to 10 years from now as AI evolves? As computing evolves, am I looking at a screen? Am I typing in a chat? Am I using an AirPod and just getting audio? Am I doing audio plus a screen? Is it just a personalized interface, and there’s not even a concept of the web? What does the future look like for accessing information and pursuing my interests in life as a human using compute?

It’s a great question. I do think the answer has got to be—we humans have always adapted to computing, and it’s always been that way—but over time, the answer will be that you need to do less of the hard work, less of the adaptation, and computing kind of works for you, right? That’s the holy grail, I think, and we are making progress, right? Be it touch, be it voice, everything inches us towards this future.

For example, when I wear AR glasses, I already wear glasses, so it’s not that. The AR glasses aren’t quite as comfortable as my normal glasses, but they’re getting there. It’s obvious to me that that’ll push it to the next level of seamlessness, where it kind of is ambiently there and doing stuff for you. So I think that’s the arc there, you know, the arc of how it’ll be more seamless and just be there for you.

Will it be like Neuralink down the line? When I want to understand something, is it that seamless? I think all of that is a possibility, but I think in the immediate world, given you’re going to have really natively multimodal models, which can take audio, vision, language—all of that—and be there in your line of view, I think when AR really works, I think that’ll wow people.

I’m not talking about immersive displays. I’m talking more about AR glasses, right? I think that paradigm looks very interesting to me, having used it. You can kind of feel that next leap, right? I think we’ll all enjoy using it in a way, but you still have a few system-integration challenges to work through.

So we have maybe a couple of cycles to go to get to that sweet spot, like what smartphones were around 2006, 2007. So maybe that’s the next leap, right? And so probably that’s what’s exciting for me.

Are you spending a lot of time on hardware?

Yes, right. I think we are definitely excited about AR glasses, the next form factors, robotics as another area, all that. We obviously build Pixel phones, and we build vast data centers, so we are definitely in the physical world.

You can think of Waymo as a big robot. We’re driving around everywhere, so we’re making, with our partners, cars that way. So definitely, yes.

I just want to zoom out and look at this competitive landscape that’s emerged for Google. Maybe it’s always been challenging, maybe there’s always been competitors, but they’re getting a lot of money and they’re investing a lot of money to compete with Google.

How have the founders of Google—I’ve seen both of them recently. Sounds like Sergey’s spending time here. They both independently shared with me that this is the most exciting thing they’ve ever seen in computer science, and it’s transforming everything. How engaged are they? How much time do you spend with them, and what’s your relationship like there?

We are obviously fortunate to have both of them involved in their own unique ways, deeply. I talk to them all the time. Look, I think both Larry and Sergey—I credit them. They always envisioned where AI would be.

I swear I’ve had conversations maybe as early as 15, 20 years ago about moments like this with them. I think they would both argue that this is the most exciting time in the field, and they both engage in their own ways.

I think Sergey is definitely spending time with the Gemini team in a pretty hardcore way, sitting and coding and spending time with the engineers. That gives energy to the team, which I think is unparalleled, right? To have a founder sitting there looking at loss curves, giving feedback on model architectures, and asking how we can improve post-training, et cetera. I think it’s a rare, rare place to be.

My favorite conversations are sometimes when the 3 of us sit and talk. They are very nonlinear thinkers, so I feel like it expands the conversation into ways which you don’t always expect, and out of it come interesting ideas.

So I think I always have access to that, but I’ve worked with them for such a long time. There is friendship, respect, and mutual dialogue. We love doing that, and I think we’ll always have that.

Your competitors out there have active founders. OpenAI has Sam, xAI has Elon, Meta has Zuck, and Microsoft has Satya. Are you willing to share your perspectives on those 4 competitors, both the companies and the leaders?

Look, it’s obviously, by definition, a very impressive group, right? I think you’re talking about some of the best companies and some of the best entrepreneurs. It shows both how much progress we are going to see, because you’re basically talking about many people who are working hard to drive that progress, right?

So, to the earlier question, when you were talking about whether we’re going to see progress, the answer has got to be yes because of the unique types of people here pushing progress. Each of them—they’re different people. I’m fortunate to know all of them, and I think maybe only 1 of them has invited me to a dance, not the others, right?

I spent time with Elon maybe 2 weeks ago, when I talked to him, and his ability to will future technologies into existence, I think it’s just unparalleled. So, look, these are phenomenal people. I respect all of them.

There are partnerships involved, and there’s competition involved. But if I were to step back and say, at the end of the day, I love driving technology progress in a way that impacts people positively.

When you think about areas like healthcare and other important areas—education—and you know, we are now talking about this, this is why AI is so profound. So, the opportunity is what excites me. I think all of us are going to do well in this scenario. That’s how I think about it.

I think that’s what a lot of people don’t grok, and I think this is an important point. Everyone out there says there are competitors, there’s a winner, and everyone else is a loser, but this is an entirely new world that’s going to be a lot bigger than the world we had last year. Everyone’s building down their own path, but there’s going to be a lot of success. It’s not just about who’s going to beat whom in the marketplace.

When the internet happened, Google wasn’t even around. So, obviously, the other thing you can say is there are companies we don’t even know, that haven’t been started yet, whose names aren’t known, that might be extraordinarily big winners in the AI thing. It’s going to be a much bigger landscape—an opportunity landscape—than all the previous technologies we have known combined. So, I think it’s all about the companies that will end up doing well, or you will do well, because you’re able to innovate and execute with the best talent. That’s what ends up being the driver.

Well, let’s talk about that, and let’s talk about the unknown competitor. DeepSeek popped up. Tell me about your impression of the model, the performance, the rumors about the next model, and what does that tell you about what’s going on in China and what’s going on that we’re not seeing?

Look, I think the main moment from DeepSeek was this: if you follow AI research and scan through papers and read them, nobody who does that would underestimate China. When you look at the amount of research output from China, they have extraordinary talent. So, I do think all of us had to adjust our priors a little bit after the DeepSeek moment. It was like, wow, they are even closer to the frontier than most people maybe assumed. I think that was a moment.

Internally, for us, and externally, people are very impressed—and rightfully so—with how efficient their models were. Interestingly, for us internally, we benchmarked it to Flash, and Flash was as efficient, or you could argue in some ways better. To our earlier conversations, I do think this is more inside baseball for us. We were benchmarking it and saying, look, it’s good to see, because they had to work in a hardware-constrained way, which I think drove a lot of their innovations and efficiency improvements.

I was pleased with that, but it tells you that the frontier is evolving rapidly. There are more players closer to it than people fully realize, and it’s going to be a very dynamic moment in the industry. I think China will be very, very competitive on the AI frontier. That’s just what I always assumed.

Much of the narrative—and I think probably the fact—around the ability to deploy AI at scale is predicated on the availability of electricity. Even Elon—and I’ve been talking about this for a while on my podcast—but Elon this week is saying, “Hey, I need a terawatt of compute.” A terawatt is roughly the power production, or the electricity production capacity, of the entire United States.

The US is going from 1 to 2 between now and 2040. China’s going from 3 to 8, and there’s probably upside given all the new electricity production technologies that they’re rolling out now, which will be additive to that. How much is electricity generation going to play a role in who is going to economically benefit from AI over the next 10 to 15 years, and where is the US compared to China, and maybe where is Google?

Well, look, you are definitely hitting on what is—when you look at any system, you want to find where the constraint is, because that’s what gates the whole system. You are rightfully identifying the most likely constraint for AI progress and hence, by definition, GDP growth and all that stuff. I do worry about it a lot.

Sometimes you run into challenges that you have to solve. You’re running into physics barriers or something like that. This is not a problem like that. We already know the technologies that can work to supply the demand we need. So, to me, it’s more of an execution challenge.

I would phrase the energy problem as obviously multifaceted. But I think we shouldn’t have an innovator’s dilemma in the energy sector. We should lean into all the possible innovations ahead, and there are many of them. Obviously, I think people will perpetually underestimate solar. Solar plus batteries will end up being huge. The amount of innovation that’s going into nuclear, geothermal—all of that—are opportunities to embrace, and there are more I’m not mentioning.

I think upgrading the grid, solving for transmission, permitting to make all of that progress faster, and actually addressing a workforce constraint will be important. To my earlier point, if you look at the number of electricians leaving the workforce versus the demand that all of us are suddenly creating, and you project out this demand, there’s a huge mismatch. Literally, how do you make sure there are incentives and workforce development to address shortages like that over the next decade? Those will end up being important policies.

I think we are fortunate. People like Secretary Wright and Secretary Burgum are very deeply aware of the issue, and I think they are hitting the problem hard. I definitely think it’s solvable, but I think we all have to put our minds toward it.

But for your business today, you don’t see electricity constraining growth in the business at this moment or in the projectable future?

No, I won’t say that. For example, we are supply-constrained this year in our cloud business, and all of us are simultaneously looking to scale up data centers. We’re running into real constraints. The way the constraints play out today is delays in projects because of permitting or not having access to electricians. All of that is a reality all of us are dealing with.

If this trend line continues—the pace at which we are all ramping up—and obviously, for it to continue, we all have to generate the returns on it, and it has to really impact the economy in a more substantive way, so they go hand in hand. If the trend continues, these constraints will be much more visible. I think today we’re all working through these constraints, so there are real constraints today, but I expect us to be competitive with China, et cetera. I think we have to solve these constraints in the near future.

What does that look like, then? Fast-forward 15 years. The US has 25% of the electricity of China. Is China just a bigger GDP in that moment? Is the pie going to grow for everyone? How do we think about that? The way I’ve assumed it is that the US always meets these moments. There’s never been a time where the US just doesn’t meet these moments, right?

To me, I look at it and say it just means that the capitalist solutions will innovate through this moment. That’s why people are working hard to build SMRs and nuclear fusion, et cetera. I’ve assumed we will meet that moment. If we don’t, or if the lines don’t match, I think the conversations will get louder and louder until we meet the moment. That’s the way I internalize it.

There’s a history of Google investing in innovative technologies and being ignored or being told that they don’t make much sense. Good luck. The TPU is a great example. The acquisition of DeepMind is a great example. The investment in infrastructure is a great example. The insane, continued investment forever in Waymo is a great example. Suddenly, it looks like Waymo’s on track to be a $100 billion business, and this is actually going to work. Mind-blowing persistence and patience.

By the way, we are taking the same patient approach in many other areas.

That’s my next question. Quantum is one. Tell me about quantum. Everyone ignores quantum. You’ve had this investment for some time. Why is quantum so important? Again, I want to use the historical data: it does seem like a small bet. Good luck. But what does quantum evolve to from a compute perspective for humanity, and when does that happen, do you think?

Obviously, quantum has gotten a lot more attention in the last 12 months or so. But we’ve been working through these things, just like Waymo, whether there’s attention from the outside or not, because we’re working on these things out of conviction on the long-term trends. It comes from those first principles.

Obviously, the universe is fundamentally quantum. To do any kind of large-scale simulations in a way that truly represents nature, you would need some versions of quantum computing. To me, quantum feels like where AI was around 2015. So, I would say in a 5-year time frame, you would have that moment where some really useful, practical computation is done in a quantum way, far superior to classical computers. That’ll be that aha moment, which will really show the promise of the industry.

I’m absolutely confident that we will get there when I see the progress, and I can pattern-match it to progress in the other fundamental areas we have worked on.

So it really doesn't feel like—obviously, look, these are very challenging areas. You may hit a constraint. I do think a lot of people are making announcements in quantum, so in some ways it's tough to distinguish them. We had the same scenario in self-driving maybe 3 years ago. There were so many people doing self-driving. It looked like everyone was roughly the same, but they weren't. I could internally tell the difference—how far ahead Waymo was.

I feel that way about our quantum effort too. I think there are a lot of announcements, a lot of noise in the industry. There are a few good people, but I do think we are at the frontier there. I'm pretty excited about it in a 3-to-5-year time frame, but we'll be patient and get there.

Yeah. Do you want to speculate on a business in quantum?

Look, we are committed to—in almost all these cases, our goal would be to demonstrate more and more useful, practical algorithms, show progress on that, and give access to it through the cloud, right? I always say it's tough to project innovation on top of a platform, right? Nobody could say just because you had smartphones and GPS and payments, something like Uber would get invented. You couldn't linearly sit and project Uber from the underlying innovation. That's how the world works.

For me, quantum is that foundational layer. Again, just like AI, there are going to be extraordinary innovations on top of it. We don't know the algorithms yet. It's almost like trying to predict how people would use personal computers in 1977 or something.

That's right. We're very early, and some of the constraints in quantum are that there aren't quantum computers to test new algorithms on. There's a lot of theory in quantum algorithm development, but not a lot of testability or experimentation at this point.

We are working on all of that too. I think we'll have more exciting moments to share this year.

I think that's what's interesting. It will expand people's minds about the potential of what you can actually do. Right now, no one really knows how to think about quantum or where it's going to take us. But those announcements, I think, are going to be really prescient. I'm assuming all your friends will show up and say, “We've got a quantum effort now, too.”

Tell me about robotics. I think this was going to be the year of the robot. We see so many models being trained on simulation data or real-world observational data that are then being used to control physical systems. Call it physical AI, call it robotics—lots of startups, lots of big companies. Google bought Boston Dynamics and a bunch of other robotics companies. I think Andy Rubin was overseeing these for a while, and then you sold them off and decided it was too early. What's your point of view on the opportunity in robotics today? How does Google play here?

We are definitely in robotics. We again have probably one of the most advanced frontier R&D teams in the world now. The Gemini Robotics efforts around vision-language-action models, et cetera, are world-class.

I do think we are now thinking through how we either partner or where we actually bring products out. You are right—we tried the application layer too early, when I think robotics wasn't really being influenced by AI as much. But now it's really the combination of AI plus robotics that gives that next sweet spot, right? We are making plans there. Nothing to share today, but you will see us make more announcements in the space.

We are definitely foundationally driving the underlying models, and we are building state-of-the-art models there. We are working with partners and testing them. When I look at the progress of humanoid robots, et cetera, in the past I would say, “These are obviously—you can see how janky they are.” Now I have to take 5 seconds to look at it closely and say, “Is this fake, or is this an actual robot doing it?” You can see the progress in the field underway.

I think we are probably 2 to 3 years away from that magical moment in robotics too. That's the next exciting phase.

Is it a good way to think about it that Google could potentially develop the Android for robotics and ultimately have a broad play here?

Yeah, we have Intrinsic, so one of our bets is effectively doing that—supporting robotics manufacturers. We are committed to having Gemini as a model that will take all modalities into account and work very, very well for robotics. It's definitely something we are committed to. How we actually bring products out—first-party versus third-party, et cetera—is what we're thinking through.

I want to talk a little bit about culture, which seems to be a key differentiator on the competitive landscape. I go back to thinking about Google offering free food, massages at work, and 20% time as a way to attract and win in the early days of the talent wars in Silicon Valley, in the early 2000s, and that persisted.

But what happened is it grew and became more amenities, and the narrative is that Google ended up creating a culture that kind of moved away from more accountability and performance and was much more about coddling employees. Can you comment on your observations on the evolution of Google over the 20 years that you've been here and what you've tried to do lately as a leader? How do you think about the culture you want to foster, and what are you doing about it?

Look, I think it's important to step back and say the underpinnings of a culture in which you really invest in employees, empower them, and even provide some of the perks was to create a culture where it's positive and optimistic, you're in an innovation mindset, and people are talking to each other. Maybe by giving lunch here, people are all sitting and talking ideas through lunch and cross-pollinating. Imagine. That's the thesis of it—not that we are trying to give lunch to people, right?

To this day, I feel we still get a lot of innovation in the company, at all levels of the company. I think people wake up and say, “Well, I can go do this.” NotebookLM, et cetera, are great examples, right? People do that all the time. Empowering employees has been, is, and will be a source of strength for Google. We can attract higher-caliber people who feel like they have the agency to do that.

But that doesn't mean—I think people shouldn't confuse that with, for example, Google DeepMind. I think from Demis and others, there's an extraordinary leadership team: Koray, Jeff, Oriol, Noam, et cetera. All these leaders have strong opinions about how to drive that frontier forward, and that's happening too.

I think it's important to strike a balance between the two. When you empower employees a lot, in some ways we've allowed for more free speech than other companies. That's one way you can think about it. You're going to hear voices sometimes. You can hear from what is effectively 500 people in the company, but that doesn't represent the company as a whole. In some ways, we are different from other companies, and it can confuse people on the outside.

Overall, we have a clear sense of where we are going. We want to empower people, all in the service of our mission. Over the past few years—and you're right, there are moments, not just for us but as an industry, when some of the other things became more of the focus than the mission of the company and why we are all here. We are not all here in the company to resolve all our personal differences or something. We are here because we're excited about innovating in service of the mission of the company and the impact we can have. Bringing that focus back is something I've been very deliberate about for the past few years, and I think it needs reinforcing.

One of the lessons for me was that we all grew so much that you assumed everyone always understood those underpinnings. But then, when you added so many people, you realized you have to go back and repeat that a lot to help people internalize it. We've done that, and we do that all the time. Moments like this help a lot too.

The current moment is genuinely both so exciting and so intense. It actually reminds me a lot of early Google, right? When I walk into the Google DeepMind building, some of our earliest engineers are all sitting there working together. People come in 5 days a week at a minimum, right? You have that intensity and that excitement, and I feel that same sense of optimism.

That's what I'm focused on. To me, that's the hardcoreness that matters, right? Are smart people really working with a passion? That's where that intensity comes from, and you have to work hard to create that. There are pockets of the company where that doesn't happen, and you figure out what changes you need to make to do that.

Sometimes, for example, I recreated the notion of labs, because I said, “Well, there are things that are possible with 10-person teams, and so we need to go and do that again.” There are quite a few projects that we have shipped and that are underway, which will be an outcome of those efforts as well.

Your values are enduring. Culture is something you're constantly tweaking to make sure you're true to your values, and so by definition there's going to be drift. You work hard to snap it back.

Was there a moment in the last 10 years where you said, “I’ve got to spend more time on this?”

Oh, for sure. Look, I think COVID was such a big distortion to our way of working. Fundamentally, Google was designed to be a culture in which people were seeing each other and engaging with each other. Losing that continuity definitely impacted our culture.

When we got people back in a 3-2 model—and some teams are working beyond that—I think it was important. I’ve spent time getting those connections back. For example, with Google DeepMind, we were intentional about creating a physical space where we could get all of them back in the same building, both in London and in Mountain View. Taking our newest building, with that kind of tent-like roof structure, and putting all the people in it—being intentional about it—has made a massive difference.

Have you found a shift in your ability to recruit top talent? A lot of great talent has started other great companies, and other great companies in Silicon Valley have recruited folks. I know there’s always a talent war going on, but has there been a shift in the tenor for Google in the last period of time because of some of the underlying advantages in AI or some of the cultural changes that are underway?

The talent market—we go through these fierce moments for talent. AI is one of them. Whenever there are these moments, Google is obviously fortunate to have some of the most talented employees, so we are a source of talent. I’m equally proud of the fact that I think Googlers have left to start over 2,000 companies. There is a virtuous cycle: people come back, we acquire companies, and all of that keeps the company fresh.

But in the current AI moment, I think we are both holding on to critical talent and recruiting. I always look at the tip of the tree: Are we able to attract the best PhD researchers coming out of the top programs? The answer is yes. There are people who have left who have come back, so I feel good about the position we’re in. But you work at it hard every week, every month, and so on.

Do you think this is going to change in the future with how we do education and how AI plays a role in education? Are you going to be able to identify, recruit, and then teach and train talent out of high schools and at an earlier age? Is the traditional kind of college education system going to change because of AI and on-the-job training?

There’s a lot of potential for change. There’s a part of me that feels maybe we’ve all misunderstood what colleges are about. Maybe colleges are about that community, people getting together, and exchanging ideas. There may be intangibles which would still make it more valuable than we all perceive it to be.

But the way I think about it is, you’re going to get extraordinary talent in more places around the world. People have access to AI, so you don’t need to be in a few certain places to be that great talent. I think the nature of that changes.

By the way, I think it’s an important thing to internalize. We often talk about talent. We’ve always been able to recruit the best talent in the country, but now there’s extraordinary talent emerging in other parts of the world, too. I think it’s something not to lose sight of. Maybe that’s the way I would think about it.

Just taking a step back, zooming back: I had a conversation 10 years ago with Larry Page where he talked about the transition from Google to Alphabet. Alphabet was going to be this holding company that would discover or develop the next $100 billion revenue business. At the time, I think Google wasn’t quite at $100 billion.

There have been a lot of these investments and other bets since that time. Do you still think about Alphabet as a holding company? Are there still multiple businesses that you want to stand up and foster, with this kind of holding-company model? Does that still hold, or is Google really the core engine that’s going to continue to evolve and have ancillary businesses that are somewhat adjacent to Google?

I’ll answer it two ways. I think we are not a holding company in the sense that we’re not just looking to invest capital in other attractive businesses. That’s not who we are. From a foundational technology basis, if you can take the technology and the R&D we do and identify problems in which we can innovate and bring a differentiated value proposition, we’ll do that.

That’s the way we approach it, and the structure is an outcome of that. It means you will have businesses that, on paper, may look very disparate, but there’s a common strand underneath them. Waymo is going to keep getting better because of the same work we do in Gemini and AI, as will Google Cloud, Search, YouTube, Isomorphic, robotics, and so on. That is the unifying layer.

It’s a continuum. Is Google Cloud a Google business or an Alphabet business? We segment it out. The branding matters less, I think. We’ll have a range of companies. Some of them will leave and IPO because maybe that’s the best way they can make progress. All of that is a possibility.

But what I think founders think about is the underlying innovation by which we create these businesses. We think at the units of quantum: AlphaFold and, hence, Isomorphic; self-driving and building the Waymo Driver, and hence all the businesses on top of it. Maybe that’s how we think about it.

Does X still play a big role in driving innovation, and do you continue to invest there?

Yeah, look, if anything, over time a lot of innovations did come out of X, including Waymo and the early incarnations of Google Brain. X, as an incubator, allows us to push the boundaries. They’re thinking about Tapestry, thinking about the grid problem—things that are extraordinary—but it’s all rooted in computer science, physics, and deep technology R&D. I think that’s the foundation across everything we do.

As we wrap up, I want to ask you one last question to hopefully frame your experience of the last 10 years as a CEO: biggest regret, biggest mistake, and what you’re most proud of.

Proud is obvious. Look, I think we have, as a company, something that not that many companies can do: push the technology frontier. You don’t hear of companies winning Nobel Prizes often. That level of foundational R&D, and then applying it to create businesses and value—I think we’ve done an extraordinary job at that, and we aspire to do that. I’m really proud of it. I think we’re pretty unique as a company that way.

There are a lot of small regrets, by nature. I tend to look forward and learn from the mistakes we make. But there are acquisitions we debated hard, came close to, and some of them are just—give me one name or get in trouble. Maybe Netflix. We debated Netflix at some point super-intensely inside.

You go through these moments, and I wouldn’t call them regrets, but you always look back and think, in a world of butterfly effects, there were alternate paths. Maybe they’re in a different part of the multiverse.

Yes. I always tell people I think they underappreciate the role that Bell Labs played in driving innovation and ultimately human prosperity in the early 20th century. I do think a lot of people underappreciate the role that Alphabet is playing in driving innovation across so many different lanes, which drives prosperity, businesses, and competition. All that stuff aside, the innovation being driven out of Alphabet continues to impress and benefit us all.

I want to thank you for your leadership and your time.

Thanks, David. Real pleasure.