Beyond The Keynote With Sundar Pichai

Welcome. Welcome, welcome to Google I.O. 2026. This is the dialogue stage. My name is Matt Berman. I'm the CEO of Ford Future. And today, I am super excited to share a conversation with the man who has been leading Google for the last 10 years. Please help me welcome Sundar Pichai. All right. Well, thanks for joining me. How are you? Pleasure to be here. Congratulations on all the announcements, of course. I kind of want to dive right into it. I actually recently found out that you're the first PM of Google Chrome. Oh. Yes. Yeah. You have a unique insight and probably very strong opinions about the future of the internet. And so that's where I want to start. It seems like the internet is being transformed right before our eyes. Agents are being built. They're infiltrating the internet. And I'm wondering, do you see the future as agents being the entry point to the internet for most people? First of all, great to be here. Thanks for doing it, Matt. And love your show. I love your content. And I appreciate all of you joining as well. Look, I do think agents are going to be a fundamental part of how we work. Because having used them, if you're living, maybe today the people who are on the frontier of how agents work are developers. And particularly with coding, two years ago, most developers started using these tools. They started giving them more and more auto-completions. And you were accepting them, et cetera. But over the last few months, developers are actually doing agentic workflows, the developers on the frontier. And they are actually deploying agents, orchestrating agent. You saw the demo in anti-gravity for building an OS. You're effectively in an agentic workflow. So I think once you get the taste of using something like that and the superpower that comes with it, I think they are genuinely adding value in a way. I think people will use them. I think it's important we build it in a way that users feel a sense of control and agency and transparency when they use agents. I think that's important. That's an important foundation. But I think, yes, I do expect agents to be a core part of how we use the web. But that doesn't mean it'll take away from people use the web for a lot of reasons. You're entertaining yourself. You're trying to do something meaningful at times. And it depends on what you're shopping for. If it's your weekly groceries versus you're trying to buy your loved one a gift. And so I think it'll allow humans to use the internet in ways that gives them joy and purpose and not always be forced to deal with, I have to fill these 18 form fields to renew a DMV license. So that's how agents will separate it out, I think. Yeah, I mean, I'm particularly excited to have agents actually do real-world tasks. You mentioned the DMV. That use case is incredible. But I also think about putting so much trust into agents to really be the arbiter of our information diet. And I'm wondering, how do we make sure that we are putting the trust in the right agents and the agents are deciding correctly what information that we should have? We are already. You are doing a version of that. If you use Gmail and there's a spam filter working on your behalf, filtering spam, in some ways you're trusting an agent. It is an agent, not the way we think of agents today. I think we've always, I look at it as working on Waymo. You have to make people trust sitting on the backseat of a Waymo and let the car go. That is an agent in some ways, but people are willing to trust. But that's because we've done the work over time to demonstrate to people, both with data, with how we have operated it, that it's fundamentally safe and it's freeing you up to enjoy the ride. And so I think it all depends on the value you deliver. That's why I think building the agency and trust with users is a shared journey, and we have to get that part right. Part of the reason in Gemini Spark, Gemini Spark is actually very powerful under the hood, but we are taking the deliberate, careful step of making it work with your first party surfaces, like Gmail, Calendar, et cetera, before we expose third party with MCP and full computer use and browser use. It can do all that, but we want to make sure users are in control and they feel comfortable. We are getting their feedback, improving the product as we give those capabilities. Sundar, if I'm crossing the street, you mentioned Waymo. I actually trust walking in front of a Waymo more so than I trust walking in front of a human driver. I don't know if other people agree with that, but it was almost an immediate trust that I had for Waymo. So I'm hoping agents are the same way. I'm old enough to remember the early days of the internet, which was an absolute wild west, but you got to raw information very quickly. And I think part of something I'm concerned about is that we're increasing the buffer between us and the raw internet. We saw it a little bit with the browser, a little bit with apps, and now more so with agents. How do we account for that? I think, look, people, in our experience with search and in YouTube, YouTube is a great example, right? People have a sense of connection with the creators they like and follow, right? And so that's a big part of what they're looking for. I think people are in different mindsets. There are times they want to discover content on the web. Shopping is delightful for a lot of people in a lot of moments. And so they're not trying to fully outsource that, right? That's why I try to distinguish between what feels like something which may be a chore at times and what feels like something which is delightful. Similarly, I think it could be in news. People have their trusted sources, at least at Google through search and YouTube and through agents. We think there will always be an incredible value from the ecosystem which users want to connect to, and the agents should be in the job of doing that. But you are right. There are times the agents are playing a role in the middle. Sometimes it's very good because it helps improve user satisfaction because the users are getting to what they want in a better way, but there's a layer of abstraction, too, and that's what you're talking about. But I think it's always been true with technology a bit, but I don't think ultimately it's also made at the same time it's also the tools with which you can create content are also exploding. People will also be creating more content. So I think there will be a new balance which we will find, but yeah, it is an interesting moment of evolution. Yeah, I like that. I think you're saying there's definitely going to be a strong place for agents to do that curation on our behalf, especially in the era of complete slop domination. But also that feeling of exploration can still be there. Yes. Because it's a fundamental human need that doesn't go away. Yeah, exactly. OK, so speaking of the Wild West, and I know you're probably feeling this pretty strongly at Google, there have been increasing cyber security, cyber attacks. The models are getting better at cyber. Obviously, Google has been thinking about cyber security for decades. Are you seeing cyber attacks, especially AI-enhanced cyber attacks, ramp up at Google? Look, we've been seeing, to be very clear, we've deeply cared about cyber because we've been working on frontier technologies for a while. Google pioneered many of important security frontiers like Zero Trust and so on. We have worked hard to keep the company at the frontier. And also, we operate many products and platforms around the world that touch billions of people. We've been pretty aggressive in deploying agentic workflows. Our internal security teams use agentic workflows to help detect vulnerabilities. And then how do you work to patch them? And we have steadily seen over the last two years, as the model capabilities have progressed, we are able to detect more vulnerabilities. And we've been working very hard to patch them. I think Mythos was a point of inflection of capturing that moment in time. They put out a model which is really well-built for that particular task. Right? But what we are excited about is part of the reason we are sharing maybe one of the undermentioned announcements today at iOS, Codemender. So Codemender is a product which we use internally and which we are building to share externally. It not only helps you identify the vulnerabilities, generate patches, test and verify that they work, and deploy them. And it's running 24-7. That's right. And real-time. We completed a recent acquisition of Viz. So Viz is state-of-the-art in being able to do this real-time monitoring of vulnerabilities, et cetera. So I think combination of what we have with Viz and Codemender, I think we are using it internally to stay at the frontier. And I think it's an important moment for the industry. I have to say I'm heartened by the cross-industry collaboration going on at this moment. I think one of the examples I would call out today, be it SynthID or watermarking, companies coming together around cyber, that is so important for this industry with this technology. So I'm encouraged by those trends as well. So you mentioned Mythos. So I want to talk about that for a moment. Obviously, Anthropic decided not to release Mythos publicly, just a handful of companies. We have OpenAI releasing GPT 5.5 Cyber. Which approach do you think is more appropriate? Which is right for Google? Is there some model that is just too good and you're going to hold it back? Or is the more iterative deployment strategy that OpenAI takes more aligned with what Google believes? I think it depends on where you feel. It depends on the model capability. If it is not fundamentally changing what's out there already in terms of the state of the art, I think it's definitely OK to put it out. But in the security world, there's a well-established practice. Google has done Project Zero for a long time. And we have teams of people who find vulnerabilities. Then we notify the vendor, give them 90 days to patch it before we acknowledge the vulnerability in the wild. And there are well-established practices in the security industry around how to do it. So I think it makes sense to me if you suddenly have something which dramatically changes the frontier, first of all, I think it's important to work closely with the government on that. And you approach it in a responsible way. So I think that is consistent with how the security industry works. But I do think it's important to also make sure enough people get access to it so they can patch their systems and so on. And they go hand in hand. So I think there's validity to that approach. So is there some threshold by which you would say, past that point, we can't release it? Or is it more, let's look at what the landscape of model capabilities are. Let's look at how cyber is right now. And let's make the determination on a model-per-model basis. That's what I would say. Is the next one you're introducing, does it dramatically change the frontier? Is it a 1% to 2% improvement over the current state of the art? Or are you taking these 20% jumps? That's where the judgment comes in. And I think that's how I would change my approach. So on the topic of model strategy, something near and dear to my heart is open source. Basically, Google and NVIDIA are the only companies with a real open source model strategy nowadays. Google's model is smaller, meant to run on edge devices. Why not release a large open source frontier model? First of all, Google, we've been big fans of open source. Google was built on a lot of open source systems. We have worked on many big things which are open source. I mean, I personally worked on Chromium and Android and so on, Kubernetes. And I can name many projects which Google has contributed pretty strongly to the world in open source. In AI, we've been building Jemma models. And we've been updating them year after year, right? And they're awesome. And I think the recent release of Jemma 4 was a great release. So we are pushing it. I think all of us are trying to make sure the frontier takes a lot of investment to get the frontier done. You've seen our CapEx dollars, right? And so you're working. You're putting a lot of R&D dollars to generate those incremental frontier models. And you're discovering new techniques as part of doing those models. So we all have to be mindful of that. But I think we are also committed to making sure there's an open source ecosystem which is able to develop. And so we take a balanced approach there. And I think we'll continue to take that balanced approach. That's how I see it. Yeah. And by the way, I do love the Jemma models. I run them locally at home. They are fantastic. So definitely, thank you. Obviously, a company of the size and the resources of Google, if you're making that decision of, OK, large closed source, large open source, we can't do both, probably most startups in the US also will struggle with that decision. What is your sense of the business model for open source in America? Is it viable right now? First of all, we not only do open source. Part of the reason we invest so much in flashlight and flash models is so that we are giving a range of options, right? And those models are workhorses, too, to support. But your question on open source, it's not that we don't know. We've been moving the open source frontier, too. There's a lot of very, very good open source models, particularly from China, which startups are adopting, too. Yeah, we're going to talk about that. So I think it depends, right? You go through moments in technology where the frontier moves so fast, maybe sometimes the open source may not be able to fully keep up with it. But then there are moments where open source will take leaps if the technology curve slows down or takes a break, right? So it's tough to predict it fully in the future. But I expect there to be a demand for a strong open source ecosystem. And I think we will definitely play a part in it. And I hope others do, too. There have been a number of open source ecosystems that have been really successful over many decades in technology. But I think just the upfront cost of baking a model makes it extremely difficult, especially if you're putting out the model. And then all of a sudden, your competitors are serving inference at a higher margin than you. But I am hopeful. I do love open source. And we're going to talk about the workhorse models in a moment. But I want to talk about China. And they have been putting out incredible open source models. If you can put yourself in the shoes of another enterprise CEO, and you're looking at the landscape of which AI model to choose for your business, and you're seeing deep-sea at a fraction of the cost but still near the frontier, why wouldn't America adopt Chinese open source AI? What's the argument to go with American AI? You're saying, why wouldn't you just use the best open source models available? Look, I think, at the end of the day, what are companies trying to do, right? They're trying to solve problems. These are, and they're trying to solve a problem with a solution. The question is, what are the solutions available? Remember, they are designing. Let's say you're doing something in customer service. You want predictability. You want reliability. You want consistency. So you want safety, security. So companies are optimizing for a lot of factors. I think that gives a place for both open source models. And there'll be providers who will take the open source models and build that ecosystem around it, which makes a lot of sense. There'll be closed source models. And it'll be an open marketplace. And people will have a lot of choice, I think. I am more OK if it is open source with the right licenses. It should matter less where it came from. I think, over time, there are good ways to inspect open source. I'm not saying that's exactly true, particularly with how the AI models are developed. But with open source comes a community which is responsible for it, cares about it. So if something wrong is happening in that software, it's not like it's going to go unnoticed. I think that creates a level of trust for people to adopt that technology. I think so. I worry less about, are we adopting open source models from China, or more that, are we doing enough in the US to make sure we are staying at the frontier? That's how I think about it. Now, I know Google is big on co-design, full stack, and just continuing on where the open source model is coming from. I've heard that argument that is, if it's open source, it doesn't necessarily matter. We're going to fine tune it or customize it for our needs. But ultimately, if we continue to build on top of China's open source, there's also an argument that they're going to optimize their models for their own chips, and then all of a sudden, we're kind of built on another country's technology. Is that an incorrect argument? Look, I think the fundamentals of AI, the way people should be building use cases on top is, because the models are changing so fast anyway, you need to build it in a way in which you're able to evolve the models underneath. I think that has got to be the way you're working through this moment. And so I think you have to be dynamic enough that you have to be able to adapt, because the model frontier, the model ecosystems are changing pretty sharply. And so that's the way I would think about it now. And it's too early to predict if this is a real concern or not. OK, talking about model strategy, one of my favorite things is watching the Frontier Labs and seeing how their model strategies play out. Anthropic and OpenAI seem almost exclusively focused on the absolute frontier. And Google has that. But you guys also put a lot of emphasis on what I'll call the workhorse class of models, the flash class of models. Talk a little bit about why. Why is that such a big part of Google's strategy? Look, I mean, in our mission statement, we have this thing to make technology universally accessible and useful. We've always deeply cared for what is the most important technology in our lifetimes that it diffuses as broadly as possible. And we get really excited at driving efficiency and making sure the best models can work in the fastest possible way cheaper. We need to do it for search, right? Because we have to give it to billions of people. We want to put it in Gemini. And so we want to give it to developers so that they can do powerful things with it. And we've had a lot of success with this strategy. And I think 3.5 Flash particularly made this point during the keynote. But I've heard anecdotally from a lot of CIOs who are so concerned about how much their companies are blowing through budgets. You can feel it talking to them. And I think the problem is going to get worse as we go through the year, right? And I think that's where I think the Flash model will really shine, because particularly in an agentic workflow, where you need these things to be repeatedly used and used a lot of times, I think it's so important to have a model which is very capable, but it's fast and efficient, right? And even accounting for token use, Flash is remarkably cost efficient, right? And so I'm really excited. We are finding it internally. We are using it as a blend of pro and Flash, right, internally. And I think most companies should learn to use it that way. To be very clear, we're super committed to being at the frontier on every category. I'm excited for pro, which we are working on. But I think Flash has a unique role to play in this compute-constrained time. Yeah, no, I agree completely, especially most companies are not solving math Olympiad problems. They're not at the absolute cutting edge of science. They need real work done. And that is truly where the Flash model shines. And not everybody is token maxing or unconcerned with the budget. So I definitely appreciate that. If you think about the future of AI, is it just truly a race to self-improving AI? Maybe just to play the devil's advocate for a second, the Flash family of models are great now, but ultimately whoever reaches self-improving AI first wins, and then nothing else matters. Do you think about it like that? Look, I think, first of all, there's a responsibility that comes with this technology. And I think we all need to be careful to avoid this race condition at all costs. And I think we owe it to humanity to make sure we deploy this technology responsibly. You are right in your question that there is this current moment where people feel like the curve is so steep and where you are in the curve matters. But just like a few months ago when we launched 3.0, people are like, oh, we are so in the frontier, no one will ever be able to catch up. I think at the Frontier Labs, it's very dynamic. The competition is fierce. We all have our strengths and weaknesses. We all also have different cadences of our pre-training release cycles. So the peaks don't exactly match. All this creates that perception gap which shifts widely in like four to six weeks. But I think a few labs are really at the frontier, and then there's a big gap. And I think there are scenarios in which things like recursive self-improvement come into play. But I think if they come into play with that no difference from the cyber moment, we all have to handle those moments far more responsibly than today, right? And so that goes hand in hand. And so I think the more AI becomes advanced, the more it's a societal conversation versus a single company conversation. Yeah, yeah, well said. So all of, you know, we're talking a lot about models, but all of it is downstream from compute. I'm always both kind of impressed and in awe of Google's ability to serve, you know, you're serving your own models, inference on the API. You're also powering your suite of products with Gemini to literally billions of users. You're also allowing your competitors to use your inference. You're also selling TPUs. I've heard the reason for this is because, I mean, Thomas told me, he said, we planned really well, which, you know, makes a lot of sense. You've been at this for 10 plus years. I think I've also heard that Google's revenue is literally constrained by compute. So I wanted to give you the opportunity. What is the state of compute at Google right now? Look, I, all of us are, I think we have made a set of, you know, bold, right decisions. So the last few years to invest in compute and scale it up aggressively. But having said that, I don't think any of us sit in the chair and say, we wish we had, you know, you look back and say, I wish I had done a little bit more. So we're living in one of those moments in time. And, you know, there are costs going up too. So like for a given budget, you may be getting less compute than you had previously planned for. Memory prices, what have you. You know, so the costs are going up. I think we plan, you know, we are able to plan long term for cloud separately from our, you know, own internal needs. And we do long range plans and we plan for it. I think it's good. And some of it is like, when you're on Google Cloud and you're supporting customers, your customers may look at something and say, well, I want access to that. Like they may look at a demo of 3.5 flash on anti-gravity and say, how are you exactly running it at 800 tokens per second? And like, could we get access to that? You know, so you're also supporting customers through those journeys and hence you're meeting them in what they're asking for. And, but it is not an easy balancing act. And, you know, we are constantly thinking as far ahead as possible. And we are making trade-offs like every other company right now. So do you have, maybe an obvious question, more demand than you have compute to serve it? Absolutely. Like, what is the scale of that? And hence the emphasis on something like 3.5 flash even more, right? And, you know, could we have done an even better Omni model? Yes. But like, you know, how can we do an Omni model which we can give to as many people as possible? So there's constantly, you're making trade-offs of trade-offs of like that, including on, you know, do you build a very large model which, you know, like an ultra-size model that will increase the capability frontier, but then who all can you give it to? So you're constantly, all of us are making these trade-offs and sometimes you make a trade-off, it looks like, well, maybe you've done that trade-off a bit differently. So we only have a few seconds left. One last question for you. What is the main bottleneck for compute for Google right now? Is it land? Is it political will? What is it? I think the way it works is by definition, bottlenecks work this way. If you think something is a bottleneck and you solve it, something else becomes a bottleneck, right? Like, you know, that's the whole, like definition of bottlenecks. I think at various times, I think you're having a few areas, your ability to physically permit and construct data centers, the power they need, and then very quickly you get into the core components for these systems. They are all the bottleneck and it's like you need all of that to work together to get a square set of a chip you need, right? And like, you know, and so I feel like there are a few parallel bottlenecks going through and it almost doesn't matter. There are like a few of them which are bottlenecks and at various times you may conclude it is memory, but then if everyone concludes it's memory, tomorrow you turn around and say, okay, no, no, no. It's actually this. And so, you know, but I think there are systemic bottlenecks across all layers of the stack now. Yeah, whatever the bottleneck is in the moment, that's the most compute you can have in that moment. Okay, everybody, please help me thank Sundar Pichai. Thank you so much. I appreciate it. Thank you.