Sundar Pichai on AI, the future of search, and what’s happening to the web

Today, I’m talking with Google and Alphabet CEO Sundar Pichai, in a conversation we recorded just after the Google I/O developer conference. This is the fifth year Sundar and I have sat down after I/O, and it’s become one of my favorite Decoder traditions.

There’s always a lot of news at I/O, and this year was no exception — Google has powerful new Gemini models, it’s putting AI agents in everything, and it’s making huge changes to Search on both the web and YouTube that will once again reshape the information ecosystem.

That’s a lot to talk about, and Sundar and I got into all of it. But I also realized it’s been a long time since I’d asked Sundar the Decoder questions about structure and decision making, so I started there. You’ll hear Sundar say he realized he needed to rethink how Google worked a few years ago in response to ChatGPT, and he made a lot of executive changes and big decisions to get the company in a more aggressive posture.

Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.

Of course, we also talked about all those search changes, and how it seems obvious that the real future of Google Search is bringing things like the new intelligent search box together with the company’s new Gemini Spark agent platform. That way, searches can set off tasks, not just deliver results. That’s exciting, but it seems likely to yet again change the dynamics of the open web.

If you’re a Decoder listener, you’ll know that I coined the term Google Zero a few years ago — that’s the idea that Google traffic to websites would fall to zero as the company answered more and more queries directly on the search results page. That’s gone from an idea Sundar batted away in previous interviews to something the entire media industry is grappling with. Even the CEOs of major publishers like Condé Nast are now publicly saying they’re planning for a world of zero search traffic from now on.

Google is also training its models on YouTube videos, and changing YouTube search to summarize and index videos so you get dropped right into the relevant parts. That’s sure to cause some creator angst, so I asked Sundar if he’s ready to fight the same battles with YouTubers as he currently is with publishers.

Finally, I asked Sundar about Google DeepMind CEO Demis Hassabis ending the I/O keynote by saying we’re “in the foothills of the singularity.” It’s no surprise that Sundar agrees with Demis, but his thoughts on the timeline to AGI are worth paying attention to.

Like I said, it’s one of my favorite episodes to do every year, because Sundar is always game to actually take the questions — and even look at search results on my phone with me. I think you’re really going to like this year’s conversation.

Okay: Sundar Pichai, CEO of Alphabet and Google. Here we go.

This interview has been lightly edited for length and clarity.

Sundar Pichai, you’re the CEO of Alphabet and of Google. Welcome back to Decoder.

It’s great to be here. Nice to see you again, Nilay.

This is one of my favorite yearly conversations. I think we’ve done it at I/O now five times.

Wow. I didn’t quite realize it’s been five times, but I enjoy it. Thanks again.

I want to start with a little bit of a lightning round. I was thinking about this. We’ve talked a lot. We always get deep into the weeds of the web and search and big, heady ideas, and I realize I have not asked you the Decoder questions in quite some time.

I was just looking back at our previous conversations, and at Google itself, and you’ve made quite a lot of changes to Google. I think a number of your direct reports have changed over time. You’ve obviously restructured DeepMind, platforms and devices, and Android. Tell me how Google is structured right now.

Okay. It is Google and Alphabet. Obviously we have Alphabet as well, but broadly I think about it as there are three main businesses in Google: Search, YouTube, and Google Cloud. There are enormous platforms we run, which is Android, Chrome, and the whole area to do with it. And powering it all is all these important technology areas, which is AI and our infrastructure work. And then you have the functions to go with it.

But at a high level, you can think of it as Search, YouTube, Google Cloud, and then our big computing platforms. Those are the main groups, and obviously powered by Google DeepMind and our infrastructure teams. That’s one simple way to get a mental model around it. And of course, we have other bets beyond that, Waymo being the most prominent of them all, but there are many, many other bets, like Isomorphic Labs and so on.

I want to stay focused on the Google of it. I feel like we could do an entire hour on Alphabet and how that’s structured and how that works as a public company with many bets. But just to stay focused on Google for one second, the knock on Google historically is this is a company that ships lots and lots of products. You can’t sell lots of products. There’s not tons of focus. There are thousands of names of different products that are overlapping in different ways.

Where that comes from, at least in my view, is that you do have these big infrastructure bets. You have all these capabilities, and the people running the businesses can use those capabilities to spin up products. And there’s maybe not a lot of overlap or central planning like, “Did we launch two of the same thing?” How do you resolve that tension? It does seem like Google has gotten a little more focused, but that is the company’s culture: “We’re going to make a lot of bets and see which ones work.” How does that resolve for you?

There’s a lot of intent in what we do too. I think it’s not an accident we have 13 products with a billion users each, and we’ve been committed to those products longer term. You can go back and think about when Gmail launched or Maps launched or Google Docs launched or Search launched or Chrome launched. We’ve been deep and consistent in many, many areas over a long period of time as well.

One way I’ve internalized it in the AI moment is for the first time, we have such a common infrastructure powering all of them with our Gemini models and the underlying AI infrastructure. So we are more able to, with intent, do things which cut across things. Personal intelligence is a great example of it. It’s one effort. Users get a choice to turn it on in each of the products, but it’s built with one common infrastructure so that it works consistently across our products.

The underlying Gemini model itself is an example of it. We are able to bring that model in the context of the products, like Ask Maps in the context of the Maps product. But a lot of the technology powering it — the voice tech, the model, the intelligence — is all one work, which is why I think the AI moment offers us a new way to think about it, and not just across Google, but across Alphabet too over time. What makes this moment so uniquely powerful is that you can invest so much in R&D and infrastructure and develop a technology, which then you can apply across all these areas, obviously in a context in which they are useful for users, but the underlying technology platform is common. There’s a lot of intent that way and so on.

You have to give room for innovation, so allowing room for innovation where teams on the margin are able to ship some new features. Sometimes you later work to harmonize them. Take NotebookLM. Notebooks are now showing up in Gemini, and it’s effectively projects as Notebooks. And you can create a Notebook in Gemini, you can go to NotebookLM, you will see the same Notebooks, vice-versa. So that’s an example of where you innovate it first, and then you’re harmonizing later.

I was watching the keynote yesterday and I saw a lot of intent and confidence from Google: “We have this core technology. We can express it in lots of ways. It’s still essentially Google-y.” There are lots of products, lots of Gemini words. I’m going to figure them all out, I promise.

I would contrast that with… I don’t know, three, four years ago when there was the ChatGPT moment, everyone worried about what Google would do. Could OpenAI show up and take your market share in search away? Between that and now, you have changed Google. You have restructured it. There are new people in leadership roles. Connect those dots for me. How did you think about, “I need to actually change how the company works,” with the competitive moment you were in that got you here?

That’s a great question. I always internalize that moment. It was tough to convey it outside, but I pivoted the company to be AI-first. We had all the ingredients, so in some ways I felt like the Overton window had changed. People were adopting these technologies faster than we had expected. To me it was a way to go and actually express ourselves through our products, but I realized we had to organize ourselves for it. And going back to my earlier point, I realized we need a core model and a core infrastructure team to power everything we are doing across Google. A lot of my initial energy was to go set that up.

To get one AI team, we had world-class research teams in Brain and DeepMind and brought those together as Google DeepMind, which was harder than it sounds because it’s like saying, “Go put Stanford and MIT together and create a department out of it or a university out of it.” So I think we’re doing that well. At that time I also set up with Amin Vahdat, who’s now our SVP of AI infrastructure, a centralized infrastructure team, which has paid great dividends. Another evolution was realizing we need a chief AI architect to architect this technology across Google, and Koray Kavukcuoglu took on that role as well. Those were important changes.

Search needed to move faster, and Search was split across many leaders, so we put it under Elizabeth Reid, with Nick Fox being responsible for the overall area, Josh Woodward coming to help with our Labs product and working on Gemini later and driving innovation. I have other extraordinary leaders in the company as well, leaders like Philipp Schindler who runs all our operations and so on. So it is stepping back, and thinking end to end about the structure and making sure we are set up well for this moment where we need to move faster as a company, which means we need to make faster decisions.

I set up these new product reviews once a week. They were AI product reviews, making sure we are intentional about how we apply this technology, where we apply it, and to review everything firsthand, that anything to do with AI, which we were shipping to users, went through that channel. I spent time directly with whoever was working on it.

The other Decoder question I ask everybody is about decisions. You’re describing a lot of big decisions, some of them uncomfortable as you change people around. How do you make decisions? What’s your framework?

A big part of my framework is over time understanding that there are very, very few decisions which are really consequential, and most decisions aren’t. What matters much more is that you make the decision, because that’s what determines the velocity of an organization. The more you’re able to make those decisions and keep the company moving forward, you’re generally better off.

Of course, there are a few decisions like combining and setting up Google DeepMind that are more consequential, and you want to take your time deliberating and doing it. But a lot of decision-making is about just making them. The more you’re able to do that, the more you do develop over time some pattern matching and you’ve seen a version of the problem before. So I think it’s good to rely on that and separate the signal from the noise so that the signal is that this is a really important decision and you want to really deliberate around it versus it may look big, but it i

[truncated for AI cost control]