When did Google decide it would need to make its own custom processor for phones? “I think I started April 16th and it was around the 17th, in 2016,” says Rick Osterloh, Google’s head of hardware. He says he sat down with Google CEO Sundar Pichai and the two agreed that custom silicon was in their future, but its exact form wasn’t yet clear. This year, as Google prepared to launch the latest Pixel that uses Google’s custom silicon, the Tensor SoC, I sat down with Osterloh and Pichai to talk about the phone, the chip, and the Android ecosystem.
Osterloh is five years in to his tenure as the head of Google’s hardware division. In those years, Google has acquired both HTC and Fitbit, pushed the idea of computational photography to new heights, and almost completely failed to pick up meaningful smartphone market share. Although the Pixel has been influential, the top-tier Pixels haven’t sold in big numbers — and the less expensive versions have fared only a little better.
“Part of the reason that I think the team has been more modest in their approach with Pixel over the past 18 months or so is because they’ve been waiting for Tensor,” Pichai says. Google has gone all out this year with a marketing blitz that began well ahead of today’s launch and continues with an NBA deal.
So now, with the Pixel 6 and 6 Pro, Osterloh hopes that the culmination of that chip decision paired with what his team has called Google’s “first real flagship” will make this year’s Pixel something different from all the phones that came before: something that will sell.
To get there, Google will need more than just a marketing blitz — it will need to have made a great phone. Answering that question will have to wait for our review, but there’s no question that Google’s ambition is to make a phone that can stand toe-to-toe with the best from Apple and Samsung.
For this week’s Vergecast, Pichai, Osterloh, and I got into all that as well as some of the antitrust issues facing Android. Listen in to hear, and stick around after the interview for some first-hand impressions of the phone from myself and Nilay Patel.
This transcript has been lightly edited for clarity.
Sundar Pichai, Rick Osterloh, thanks for coming on The Vergecast. I have asked both of you a version of this question at least a half a dozen times. Because the Pixel 6 is now announced and out, and you have your own chip, I’m going to ask it one more time. I’ll start with you, Sundar; why is it important that Google is in the hardware business?
Sundar Pichai: It’s a good question. I have answered it before, but I think it’s worth saying. If you’re committed to driving computing forward, I do think you need to think at the intersection of software services [and] hardware, to evolve it. I think we have a unique expression of it. We have always invested in certain deep technology to build helpful experiences with AI, we see a chance to express it in a unique way. So I think we will bring newer insights, newer experiences.
Second, it helps us drive the ecosystem forward. In my experience, every time we have done something well in a category, that category as a whole benefits in Android. Be it the Nexus phone or Nexus 7 as a tablet, and so on. And third, to do this well, you have to build a sustainable business. We are committed to it, we are building a business we want to grow and do well, as well.
I imagine that third answer is where I’m going to turn to you, Rick. How is that going?
Rick Osterloh: We’re still very early in this. I think the [hardware] organization was started about five years ago. And especially in such a mature market, that matters so much to users, it takes a long time to build the right capabilities to be leading. And we feel like finally, at this moment, we’ve gotten to the place where we feel really good about the innovation we’ve been able to build, the organization and capabilities we have in that towards building great phones. We think Pixel 6 is a big step change, and a nod to where we’re going in the future, which is really focused around AI innovation, really focused around building a portfolio that tries to address big parts of the market. We want to give our users the best of Google. So, that’s what we’re trying to do.
Tell me why you think the Pixel 6 is categorically different than other smartphones that are currently on the market. Or do you think that’s true?
RO: I think there are significant differences. The number one thing is that it really is a clear representation of Google’s brand. What we’re trying to deliver is the very best possible Google experience. We do that by bringing all of our services that many people use, pulling them together in a cohesive user experience on Android 12 and beautiful new material UI.
We also have been bringing a lot of our AI innovation. We’ve been working in tight coupling with our AI research teams. We’ve been building silicon together, and we put all that together in Pixel 6 so we can build the best possible experience. You see that in the work around speech. There are really exciting new improvements in that arena, we’ve always been strong in computational photography but we’re taking that a lot further with Pixel 6 and Tensor, and we’ve got a big jump in our video capabilities too. We’re just really excited about this being a moment where we pull it all together in its best possible form.
Do you think it’s fair to say you’re intending to go toe to toe with the best that Apple or Samsung have to offer in terms of their flagships?
RO: That is our intention, yes.
Are you intending to take market share from either of those companies?
RO: Certainly we want to grow. First of all, we acknowledge we’re a challenger. The market has been pretty ossified, there have been a couple of key leaders for a long time. We think we’re taking a different approach with AI-led innovation, and we think our time is now to be able to grow, and that we’ve taken a big step change in so many parts of the overall experience that it’s a good time for us to be investing a lot in distribution, in marketing, and of course, in the product and technology.
I’m bringing up the ecosystem because it’s Rick’s job to sell a bunch of phones. Sundar, it’s part of your job to make sure that the Android ecosystem is growing and secure and that all your partners are happy with it. When you’re sitting down with Samsung to talk about Android, are you talking about the Pixel line? How do those conversations go?
SP: They are our most important partner on Android. I think Samsung is a big partner for our devices and services team as well. There are many components [in this phone] from Samsung to make all of this work as well, so they’re a big partner there.
From an Android standpoint, we are very focused for example, on the work we are doing with Samsung on foldables, working hard to make that leap and make it successful; the partnership more recently on the watch, which we have done together with them. So I see Android as one of the more customizable and flexible OSes out there. It gives Samsung and other OEMs the chance to really innovate. And we have Google also doing that. I think some part of it will grow the pie overall. And I think it’s good as a consumer. It’s good we want to put out something unique in the market, and it’s more choice. So I think it’s healthy that way.
I’m actually wondering what both of you think about the market specifically with Android. You called it ossified. You are coming in as a challenger. You say you want to grow the pie, but I can’t imagine there’s a lot of new smartphone customers to get.
I’m wondering, going forward, what you think you need to do to be successful. You mentioned marketing, I’m wondering what else you might need to focus on to shake up the market like you seem to want to?
RO: I think first off is innovation. Our main focus as a company, and especially in our phone products with Pixel, is to try to offer as much help as we can to users across a lot of different arenas. People are very familiar with what Google does with search and Gmail and so many other categories. But we also think we can help people even further with AI as applied to photography, or videos, or speech, or natural language processing, or translation, and on and on. And we can put that together in the best way on Pixel. We also feel like there’s an opportunity to give people great value with a combination of these things. Our Pixel A-series has done a really good job with that to date, I believe, and we intend to continue forward with that. So there are both innovation and value-for-money opportunities in the market, and we hope to offer that for users.
Sundar, any thoughts on what you think the Android market needs these days?
SP: I take a slightly longer-term view, and if you look at computing as a whole, there are phones, you have seen hints of people doing stuff beyond their phones as well. There are watches, we’ve done Google TV with Chromecast, Google Home.
To me, this is as much about building a deep capability to innovate in computing for the very long run. Taking that 10, 20-year view. The day-to-day of the current position in high-end phones, I think that’s one snapshot in time, but if I zoom out, that’s what I meant by growing the pie and taking that long-term view.
RO: And if I could add one thing, it would be that we believe that we have a unique portfolio and that it covers a lot of the activity in the home. We recently acquired Fitbit, so we have a nice, robust capability on wearables now that we intend to grow and knit together. Then, of course, the phone is the most important part of people’s computing day-to-day usage. So we believe this unique set of things is the right portfolio to best express where Google can go in the future. We’ve been calling our computing vision for the future ambient computing, and that’s the way we think about it. We want to have people be able to use these devices wherever they are, whatever situation they’re in, whenever they need it, [and] be able to naturally interface with them. Google Assistant is a common interface across all these services. And we think it’s the leading assistant.
This combination is the right set of technologies for the future of where we feel computing needs to go for users. That is going to be our unique element in this, is the strength in the home, in mobile, and wearables. And we’re excited to start on that.
SP: And hopefully [we] do it in an open way. Hopefully all this computing doesn’t mean you have to be locked into one brand or one OEM. I think some of what Android brings is the ability to have many people plug in and participate. So I’m excited about that as well. And us doing all this also really pushes us hard to make sure different phones can work with different watches, etc., in a way that it doesn’t add complexity for the user and things can be seamless. I think that’s important as well.
I definitely want to talk about Tensor, but you mentioned Fitbit. How’s that going? I know they’re still releasing new products that are on the Fitbit platform. I’m waiting to see if they’re going to be doing hardware on your watch platform. The software story seems a little bit confusing. There’s some Fitbit stuff, but Google Health is still sitting around. So what’s going on with this?
RO: Well, they joined our company six or seven months ago now. So it’s pretty early in the integration. James [Park, Fitbit CEO] and team are part of my organization and we’re working really closely together. We’ve consolidated a lot of our health work into the Fitbit organization so that we can really have a focused consumer health approach for wearables and mobile. We announced at Google I/O that we’ll be working on Wear OS in our Fitbit team. So we’re really excited about where that’ll go. No doubt you’ll see them build wearables on Wear OS in the future. So we’re hard at work at that.
I asked too soon. We’ll check in on it next year, maybe.
RO: Yes. Next year. Let’s talk.
Let’s talk about Tensor. When did this project start? When did you decide, we’re making our own system on a chip?
RO: I think I started April 16th and it was around the 17th, in 2016.
So literally the day you started at Google, you walked in Sundar’s office...
RO: When Sundar and I were talking about starting this organization five years ago, this came up. We felt like it was inevitable that we need to start to invest here. Exactly the form wasn’t clear. But Moore’s Law and general computing laws started to break down around that time and a little bit earlier. So it’s pretty clear, if you wanted to be on the cutting edge of AI innovation, it would have to involve real complete system design. About a year in, 2017, we decided we actually really needed to build an SoC, because you couldn’t just build a single co-processor in order to really harness the full capabilities we needed across a diverse set of AI models and approaches. We discussed this a long time ago. I remember talking to Sundar, letting him know this was going to be a pretty big investment and take a long time. And we didn’t have an organization at that point to build it. So it’s been a long journey. But we’re really excited we’re here. I mean, this is why I came back to Google. Why we wanted to work together on this.
SP: I’ve always felt like you have to do deeper work. If you look at our services we’ve built on the consumer side, our data centers, we’ve never shied away from the fastest switches you needed for our networking side, subsea cables. So we have all these TPUs (Tensor processing units). On our data center side, we buy a lot and deeply use Nvidia GPUs and so on. But we have also tried to push the boundary with TPUs custom-built for our AI services. I think it’s important, recognizing that to do well here, particularly if you want to do it well from a AI standpoint, we need to evolve silicon with that focus. I think we had good DNA in thinking on the server side too, here. I think it’s a natural part and I appreciate the long-term focus. I clearly remember our conversation. It’s good to see it all come together.
Cast yourself back to this conversation. Was it obvious that you wanted to do what you’ve ended up doing or was it a little bit fuzzier? Did you say, “We’re going to need to do some AI chip, we’ll figure it out?” Or were you thinking, “Apple’s got their own chips, we’ve got to have our own chips.” What was that conversation decision process like?
RO: It wasn’t 100 percent clear what form our system design work would take in the very early days. It was only after really thinking through what the direction of travel in our various AI research teams was that we realized that we really needed to build the whole thing, simply because of how data moves in and out of our models and how, when we’re doing computational photography, we’re using not just an accelerator, but all parts of the chip to achieve what we need.
There were other elements too, like being able to have a very low power set of the system that could do things when the device was otherwise at sleep. So this combination made us realize we actually really need to build a full applications processor that includes a lot of different heterogeneous sub-components. That was evolutionary thinking. That took a while to fully come to that realization. And we had lots of discussions about it along the way. But it wasn’t one flash of one inspirational moment where that came to us. It was an evolution of thinking.
That’s the old history. I need to ask about more recent history. This morning as we were talking, Qualcomm decided to say that, “If a company’s making their own SoC that’s a red flag,” on Twitter. Wondering if y’all have anything you’d like to reply to Qualcomm with.
RO: Qualcomm has always been great partners of ours. We’re doing this so that we can work closely with our AI team.
SP: Qualcomm plays a really important role in the Android ecosystem and will continue to [do] so. That’s why I gave the server side example as well. We have done Tensor processing units on the server side, but we deeply use Nvidia GPUs across everything we do too. I think the market has done well overall. I think pushing the high end of silicon across all of us is good. But we are doing this for Pixel, and Qualcomm and others will play an important role in supplying the Android ecosystem.
There is a problem across all of Google communicating with customers. It’s a specific problem now with Tensor on the Pixel, of explaining the value of your AI systems, of getting people to understand what they can do, why it’s helpful, and have it be more than just, “it does this specific feature where you can erase something in the background.” I feel like a lot of the time when we talk about AI, it ends up just ending on jazz hands. And it’s supposed to be better. So I’m wondering how you’re thinking about communicating the capabilities that you’ve built for this phone to customers.
RO: I think it’s really difficult to reduce AI to a benchmark or any sort of simplified measure. The way we intend to try to communicate the value to users is through specific experiences that they’re likely to really get excited about. It does make things possible that weren’t possible before. For instance, on Tensor and Pixel 6, the speech recognizer is out of this world. We took a lot of engineering efforts to move our data center quality models, make them run on-device using the TPU. It wasn’t possible before Tensor. And what you get out of it is something that can recognize words at 200 words a minute. I mean, it can go faster than I can think.
And I think it’ll totally change how I use the phone, and I think people will see the same thing. It works across the whole device. That’s an example of how AI will affect everyone’s day-to-day usage of devices. We’ve been doing it for years in other areas, like with screening calls that might be unwanted or in computational photography areas. And we get to apply that more and more to harder problems. So we’ll show users these specific problems that we solve with it and hopefully they’ll find it helpful. Hopefully, they’ll see the benefits of AI overall.
When you’re thinking about what features to launch next, how much are you thinking about potential backlash? I think that with Duplex in particular, there was a very strong reaction, and it took a while for you all to figure out the best way to talk about it and how to get that out into the world. When you’re thinking about new features for the phone or even for Google search or something, how are you thinking about the rollout?
SP: Well, it’s a good question. We now have the experience of many features across our products people use and rely [on]. And they actually walked with their feet. Think about smart compose in Gmail. That’s a new thing to get used to. You have, effectively, AI suggesting phrases to you and stuff. Really, it resonates with users, right? People understand it. I would give a lot of credit to the people using our products. They understand, they are adapting, and they let us know if we get something wrong. I’ve seen [this] across search, across maps. When we do these features well, I think the bar doesn’t change. If it’s an AI feature or not, the bar is still the same; doing something which users find magical and useful. And if so, I think they will adopt it well.
RO: A wonderful attribute of Tensor is it enables us to do a lot of these capabilities closer and closer to the user. It keeps them in control. We took great pains to move our data center quality speech model to the phone. It’s running locally on-device. The information stays on the device. A lot of our computational capabilities are done there with our photography experience. And this is a direction of travel for us. We want to strike a great balance between making sure that we’ve got great capability for users and also making sure they’re comfortable with how data is being handled.
How are you balancing adding capabilities to the Pixel specifically? Maybe you’ve got a Pixel feature and then it goes into Android a year later or something? I could imagine a bunch of Android manufacturers would love to have your speech model on their phones. How are you thinking about the stuff that you do? Does it get put into the rest of Android in some way?
RO: It’s a balance that we work on with our platforms and ecosystems team and Hiroshi [Lockheimer]’s organization. But a lot of times Pixel is basically the vanguard of our innovation across the company. We can get it to market through that quickly. And then a lot of it does flow into our ecosystem. Some of it is harder for that to happen, in that a lot of the Pixel 6 innovation, for instance, is driven by new hardware capabilities that haven’t been possible so far. So those might take a while, if ever, to move over to the ecosystem. But in general, we want a lot of this capability available to people.
SP: Be it Google Assistant or Gboard, the Google keyboard, etc. There are many ways to do it. Over time also, and this is not directly answering your question, but I would say we are taking our speech APIs, computer vision APIs through GCP [Google Cloud Platform], through cloud, and providing to companies as well. So we’re giving it through APIs as well. But it’s a good question. I do think OEMs are interested in accomplishing an experience. I think we work hard to make it possible. So I’ve never seen that be the gating factor. If people have really wanted to make the camera better on their phones and they engage with us, I would say our teams work super hard to make their cameras better.
Since we’re talking about Google services on Android, I do need to ask, South Korea recently ruled you need to allow third-party payment systems in Google Play. It seems like you’re going to go along with that.
SP: Obviously I think our Google Play team is assessing it and understanding it. I think they are in dialogue with it and we’ll find the right way to comply and make sure we support the ecosystem well. It’s an important ecosystem for us. I do think it’s important to understand Android is different from others. We invest thousands of engineers, build the operating system, which we provide for free. We don’t take a share of the device sales, not a share of the carrier revenues. So in some way we have to sustain our ecosystem. We have a different model. Google Play is an important way. In fact, it’s the main source of revenue. It supports Android as a whole. I think we’ll make that viewpoint clear, but we’ll engage in conversations. I’ll leave it to the team to figure out the right next steps.
I guess similarly, there is also a ruling about your contracts often not allowing manufacturers to make forks of Android, if they want Google services on their Android phones. Do you know what your reaction to that ruling is going to be?
SP: I literally spent so many years in computing. I’ve never seen anything as open as Android. I buy a Peloton and I discover it’s built on Android. No one from Peloton, I’m sure, ever talked, I don’t know whether they talk[ed] to people at Google. Amazon ships on Android. Obviously one of the areas where Android gets criticized is at an end user level, the fragmentation as a cost, right? Developers complain of complexity. So we are constantly striving to create that balance. And as part of that, sometimes there may be feedback from regulators and I’ve taken the approach that we will understand it. I haven’t seen this as a major issue, but to the extent it is, I think we will address it too.
I guess I don’t know what addressing it means.
SP: It depends on what the specific thing is. I’m just basically pointing out there are many different ways people work on Android and ship Android. And in fact there are like, literally, I think if you [count] tablet, all the form factors, all the versions, there are forked versions of Android we ship all the time. So I do think we are trying to make sure at the end of the day we feel responsible for delivering a user experience on top. There are going to be areas where we are going to have viewpoints. I understand there’s a balance. And we’ll have to figure that out on a case-by-case basis.
To end, let’s talk about one more fun thing. Spending a ton of money, which you all have been doing on Pixel marketing. I have seen billboards. There are TV ads. The thing hasn’t even officially launched yet. You’ve got ads before we know the price. This is a blitz. Is it going to get bigger to try and sell this phone?
RO: Yes, we’re excited about the product and we want to make sure people know about it. We started our marketing early this year, which is a bit unusual. And the reason we did that was it’s a really big change. So we started that effort a couple of months ago and have been building it. You’ll continue to see it build throughout the year. We’re doing a terrific partnership we’re going to announce with the NBA, so we’re very excited about that. That’s just one of many things we’re doing, but you’ll see Pixel is the official phone of the NBA. And we’re very excited about that because I’m a huge basketball fan, so this is a fun topic.
Sundar, did you know this NBA deal happened? I saw you do a double-take.
SP: No, I realized the real thing that drives Rick to work hard on Pixel is so that he gets some courtside seats somewhere for NBA games.
What about carrier deals? Even with the iPhone, I think that a bunch of the recent success has been coming from getting huge discounts on trade-ins and huge discounts on — they’re not called subsidies anymore, but they might as well be, especially here in the US, where I think you’re really targeting — phone sales are driven by carrier, deals and subsidies. How is your thinking around that going?
RO: We’re going to be working very closely with carriers. You should expect to see Pixel on all major carriers going forward. In the past, we’ve had distribution with them, but not always the whole portfolio and not across all of them. That’s going to change going forward, and we’ve been talking to all the carriers in the US, and we expect to see that for Pixel 6 going forward.
Do you think you’re going to be able to make enough of them to meet demand?
RO: I hope so. I mean, we’re making a lot more than we have in the past. There is a minor supply chain issue happening in the world. We feel good about where we are right now for this year’s supply. And hopefully that will continue throughout next year, but we’ll see. I mean, we’re small right now, so it’s hard to know if you’ve gotten enough.
SP: There are many areas which we are a small player trying to do better. So we face the same challenges. We are competing against others who have economies of scale, better supply power and so on. And so I think the team has to work extra hard, be it on the go-to-market side, getting carrier deals and making sure people can see the product, or on the other side, making sure we can get supply. And I think those are all the challenges which make it even more exciting, but uncomfortably exciting.
There’s a cognitive dissonance when it comes to thinking about Google as a small player. You’re Google, but you’re a very small player in the actual phone market. I think it was four years ago, I asked you when you think the business will be big enough to start breaking it out in your financial results. And you told me five years, so you got a year left. Do you think you’re going to get there?
SP: I don’t want to make any forward statements from a financial standpoint, but broadly, I want to be clear. We want to build a sustainable business and there are other areas where we are a challenger, right? I think when we try to compete for enterprise contracts, we go up against big Microsoft bundle contracts all the time, trying to convince CIOs.
I think there are several aspects of business which have that characteristic and just like we have done with cloud or YouTube, I think to do well in technology, you have to build a sustainable business and they go hand-in-hand, otherwise you can’t. For someone like us, who believes in deep R&D investment, it is even more important to do it well. We are committed, and I take a very long-term view and you’re right, it’s year four or five, but I thought of this in a 10- to 20-year timeframe.
Was the plan to have this be the hockey stick moment for Pixel? It seems like you’ve been building up to this, and this is the moment when instead of having a little bit of success, you really want this phone to be the thing that launches it as a real competitor with Samsung and Apple?
RO: It’s definitely a step change moment for us. We see this as truly the starting line for Pixel, in that we’ve brought together all the elements that we think we need to be a real player in this space. But consumer electronics in general doesn’t usually [grow] like hockey sticks like you would in a web services business. It’s a steady build. And that’s how we’re approaching the next several years, a real steady build. And we’re excited about where it’s headed because what we’re building with Tensor, and with Pixel 6, is how we see the future as well.
SP: Tensor was a big part of it, so I think a common question I would ask Rick is about Tensor timelines and not where the status of the project is. Part of the reason that I think the team has been more modest in their approach with Pixel over the past 18 months or so is because they’ve been waiting for Tensor to time our investments this way.