Good morning,
Today’s Stratechery Interview is with Google Senior Vice President of Devices & Services Rick Osterloh. Osterloh has worked in mobile his entire career, including Good Technology, Motorola, and Android; he founded Google’s Pixel smartphone program in 2016, and earlier this year took over leadership of Android and Chrome as well.
This interview was recorded a few hours after Osterloh presided over this year’s Made by Google event, where the company announced new AI capabilities for Android and a new line of Pixel devices. In this interview we not only cover those announcements, but also Osterloh’s entire career, and the lessons working on smartphones for two decades has taught him about what it takes to succeed in arguably the most important market in the world.
As a reminder, all Stratechery content, including interviews, is available as a podcast; click the link at the top of this email to add Stratechery to your podcast player.
On to the Interview:
An Interview with Google SVP Rick Osterloh About Pixel, Android, and Smartphone History
This interview is lightly edited for clarity.
Smartphone History
Rick Osterloh, welcome to Stratechery.
Rick Osterloh: Thanks for having me, Ben. Great to talk with you.
Well, it is great to talk to you as well. I mean, we are nominally here because you had a little event this morning, we’re running this on Thursday, so I guess it was a couple of days ago when people hear this, but the real reason I wanted to talk to you is you are a well-known Golden State Warriors fan, and we are coming off an Olympics last couple of games where your boy, Steph Curry, put on quite a show. How are you feeling?
RO: It was so cool that the entire USA got to feel how a Golden State Warriors fan has felt for the last few years, and I thought that was actually one of the coolest things about the Olympics, it was totally unifying that everyone was cheering for Steph. And I mean, I don’t know, that was peak basketball. I have not had more fun watching the last two games than I have in years, maybe since the Celtics finals, it was incredible.
Are you feeling a bit melancholic that this is the peak and maybe this is the end?
RO: I mean, having LeBron [James] run point and then having Steph and KD [Kevin Durant] and Devin Booker and [Anthony] Davis or [Joel] Embiid playing center, it’s like the smartest basketball players ever, all with pretty specific roles. I don’t think you could — I don’t know, since the Dream Team, I don’t think I’ve seen anything like it, and it felt a lot like the Dream Team. I was like a kid when that happened, and that was the first time I think the pros were able to play, so that was an incredible experience. This was very similar, it was awesome to see.
Yeah. Well, unfortunately, we did have this event, so you weren’t able to see it in person, you had to be stuck in the Bay Area. But before we get to the Made with Google event, what I usually do with these interviews is I like to do my usual routine of learning more about you and your background, and I think actually your background is actually quite pertinent to a lot of the questions I want to get into. I’m talking about things like Motorola, stuff we’ll get back to, sort of back in the day but even before then, you don’t have much of a profile online, so I don’t even know where you grew up. So take me back to the very beginning. How’d you get interested in technology? Let’s go back to the basics.
RO: Sure, yeah. I grew up in Los Angeles, and I was there until I was 18, and then I went to college at Stanford and I stuck around there ever since. Basically, I stayed in the Bay Area more or less with a couple of vignettes outside.
It was in college where I started to get really interested in technology and probably for the first three years after college, I was still trying to figure out what career I wanted to pursue, but I came across a bunch of work in the telco space when I was doing consulting work after college, and that’s when I got interested in mobile.
So I was kind of in that weird period where the Internet first started when I graduated from college and so I sort of was like, “Oh, I think I missed joining a real interesting company in the very early web days”, and I ended up going back to graduate school also at Stanford. That was when I realized, “Okay, I want to try to do something in a new emerging tech space”. It seemed like mobile was the right place to pursue that, and I worked briefly at Amazon as an intern in the middle of my grad school time and then when I graduated, I kind of was working at Kleiner Perkins and there was a company they were incubating, and I joined it as one of the first employees, it was called Good Technology.
Yeah, I’m interested in Good Technology. So this was mobile device management for smartphones, but this was before smartphones were smartphones, this is like the pre-iPhone, pre-Android era.
RO: (laughing) Really dumb smartphones.
But at the time it seemed like you were on the cutting edge.
RO: It was definitely cutting edge. The thing that was — people had actual cell phones, that’s what they called them back then. There were cell phones, but the emerging technology was using mobile data, and that was what BlackBerry kind of invented this, all-in-one integrated solution that was super useful. I don’t know, did you ever have a BlackBerry, Ben?
I did not have a BlackBerry, I was a big Nokia person, so I had Symbian, which in retrospect was just an absolute mess. I was outside the US at the time, so BlackBerry wasn’t really the thing here in Asia. But yeah, it’s pretty historic for sure.
RO: Well, BlackBerry used — I mean, I can’t believe this is not that long ago — but BlackBerry used these ancient data networks, and one of them was called Mobitex, and it had bandwidth of literally two kilobits per second, so it was just enough to have a small text message sent back and forth. But to their credit, BlackBerry invented this idea of connecting it to email servers, and so then suddenly email became always-on and you could always get emails, and that was a huge breakthrough because before it had to be, you’d pull up a laptop and connect to whatever terrible connection you had to download your email. So it was a session-based thing, and that shifted to always-on.
What Good was trying to accomplish was create that capability of always-on email, but also syncing your calendar and contacts over the air and do it to things like Windows Mobile and Palm OS at the time — the open standard OS’ is at the time. So this was 25 years ago is when we started starting this whole business.
I think ultimately Good was actually acquired by BlackBerry, but this was like 2015 when both were sort of irrelevant.
RO: It changed hands several times. It was first acquired by Motorola and then Motorola divested it, and then another company that was a competitor of Good called Visto bought it. They renamed themselves Good, and then BlackBerry ultimately bought it, which is the ultimate irony.
Got it. So is that how you ended up at Motorola?
RO: Yeah, I ended up at Motorola through the acquisition.
I want to ask you about Motorola, but a question just occurred to me is you’re dealing with this two kilobyte connection, to what extent was every company in that era — if you think about that, it’s sort of like a world of scarcity where you’re trying to get an always-on connection through the thinnest of straws, and you hear about when the iPhone came out, it was just amazement at the level of capability and the processing power that was available. But how much of the inability of the existing players to adjust or respond is that they grew up in this world of super scarce broadband and couldn’t even have the imagination of what happens if not just you have a connection all the time, but that connection is de facto unlimited? Edge or 3G is not unlimited, but the iPhone and later Android assumed it was.
RO: I think when the big game changing moment came with connectivity was with 4G, and so I’ll return to that point in just a second. I think with the legacy players like Nokia, like Motorola and so many others, and I’m talking about 2000s era Motorola as opposed to the one today, but all the players back then didn’t realize that phones were becoming computers, that that was the asset they didn’t have was the capability to really make computers.
So frankly, I think in retrospect, it isn’t much of a surprise that a company that could marry strong R&D capabilities with the ability to make a computer emerged as a smartphone player down the line, and I’m referring to Apple in this case. The ones that were able to marry really strong R&D capabilities with the ability to produce a computer were the ones that came through. And really, Apple and Samsung were the primary ones that emerged, but there are others of course, too.
But Samsung couldn’t make a computer. They depended on Google for that, right?
RO: Well, what they had was all the really amazing hardware components that were essential to making these things come to life, and so they were big in memory and displays and microprocessors and foundry.
They made the processor of the iPhone.
RO: Yeah, so they had the real R&D capability to produce a computing device, and very few companies did. I mean, Motorola did, they divested some of it. Unfortunately, I think in retrospect, that turned out to be a little bit of a mistake, but they were able to recover and still be a player in the smartphone business.
But when Motorola acquired Good, my role literally was to try to help them bridge to this next era of computing, so I started a small division at the time, it turned into everything we were working on, but it was a small division to build products on Android. I was close with a lot of the early Android employees and so I was able to kind of get that started, and then when Sanjay Jha became CEO, he really focused the company all around Android and building smart products, and so Motorola was able to kind of make it through that knothole, but it was a difficult transition for them and for anyone that tried to make it.
I think Motorola acquired Good, I have May 2007, was that a direct response to the January 2007 unveiling of the iPhone?
RO: No, the acquisition was announced in 2006, it actually closed in January of 2007, but there was a period where it was — but actually, I mean, ironically, as I remember at the time, it was really actually centrally about competing with BlackBerry that the acquisition originally happened.
As it turned out, Good Technology was working on a client product for both iOS and Android at the time and so it was like I had a front row seat into like, “Oh my God, this industry is going to completely change.” I mean, this “smartphone space” was really uninteresting before iOS and the Android. There was BlackBerry doing their thing, there’s Palm OS too, which looked like they were in trouble and ultimately were and the other player doing stuff was Microsoft with Windows Mobile, which was just like an awful product.
That’s the thing, everyone says Microsoft missed mobile. No, Microsoft was there in mobile, they were just terrible at it.
RO: Yes! That is rewriting history, Microsoft had been in this space, they saw the opportunity, they put an enormous amount of resources in it, and they just didn’t have a good product and so that’s the actual what happened. I worked on every single Windows Mobile phone that was made from 2002-2003 until 2007 and frankly, they were all pretty bad.
The great irony is, to your point, is the key thing about smartphones is that they were actually computers, and Microsoft’s flaw is they didn’t see it as a computer, they saw it as an adjunct to your computer, and that fundamental distinction is why it never had a chance and it’s just interesting that the most well-known computer company missed that point.
RO: Yeah, I am not sure what happened there, there was millions of little things. For instance, if you thought of it as a computer, you’d probably think like, “All right, this needs sophisticated networking”, but none of the phones supported UDP, so every single phone had to have assumed a TCP connection, which was impossible to get right in the mobile network era of 3G and 2.5G data. So it’s one example of how it was just a missed opportunity for so many people back then, I don’t want to pick on Microsoft too much, everyone was kind of missing it in that era before iOS and Android.
Here’s the question. It’s been a decade, maybe we can get some honesty here, why did Google acquire Motorola?
RO: I wasn’t there when, I had left Motorola at that point, so I missed—
You were at Skype at this point, I think?
RO: I was working at Skype, I was living in London, and I was doing something totally different than what I had been doing.
(laughing) This is even better, you’re not involved, you’re just a spectator like me!
RO: I wasn’t involved. It was principally around IP, and so I think at the time, there was a big battle between companies about IP, and I think there was principally a concern that maybe everything would get stuck, all innovation would stop and ground to a halt because of these IP issues. Motorola did have quite a substantial amount of it, and so that was a major component of the deal, Google did retain those IP rights. A decision was made at some point that it made sense to divest Motorola, but quite clearly looking back in retrospect, that seemed to be a temporary decision.
Where is your timing in this? So Motorola, they divested to Lenovo. You worked for Motorola for Lenovo. Were you back at Google before then and you went with that, or how’d that work out for you?
RO: Yeah, so in 2010, I left Motorola to go to Skype, then Google acquired Motorola, and I made the decision to come back. It was funny because I was bridging two worlds, I was friends with Hiroshi [Lockheimer] and all the Android people, and then I knew all the Motorola people. So I was an interesting hybrid hire, if you will, and so I was head of product management for Motorola when Google owned it during that period for 2012-2014. I really don’t know because I wasn’t involved why Google decided to sell the division, and so I just literally wasn’t involved in that decision.
Smartphone Integration
I’m more curious what you learned as being a part of Lenovo. How did your understanding shift where you are an OEM partner, which you were to an extent, even in the pre-Skype days, but now Android is definitely the thing, but what did you take away from that?
RO: First of all, one of the things I learned was how good a PC maker has to be at managing their supply chain and operations, and how different the PC business operates than the phone business. So, that was quite a stark realization. The efficiency requirements to be competitive in PCs, I mean, it’s extraordinary.
Because they’re so commoditized?
RO: The model that we grew up with was this industry layered structure is what wins. In a mature market, you’ve got people that are integrators at the PC and supply chain managers of the PC layer. Then you have Intel, and chipmakers, and OS maker, and that’s how the industry structure works, that was what we were taught as the PC industry model. But the phone business broke that, where it turned out to be, you actually really had to be strong at R&D and the integration components were so difficult that it looked much more like an R&D and engineering centric business, than it did an operations and supply chain centric business.
Is that still the case today or has that evolved a bit?
RO: Even more so.
So basically, if you like the Clayton Christensen framework, you’d say, “Integrations devolve into modularity”, but your argument is actually the difficulty of these devices is such that everyone’s actually super highly integrated, even if we think of Apple as being the integrated player, the reality is everyone is.
RO: Everyone is. That comment primarily pertains to what you might consider to be — I think the market’s really different in the premium end of smartphones, versus say, call it under $400, super different, the under $400 segment probably resembles the PC market more.
Interesting.
RO: Where you’re using standard software, you’re buying silicon from merchant silicon providers, you’re buying components, and you’re trying to do a good job integrating, being efficient, and also probably doing some cool stuff in the UI and brand to make consumers interested. However, the premium end, you’re solving novel problems frequently.
A recent example on the hardware side would be foldable phones. This is a really, really, really hard mechanical engineering problem to get right. Then, on the electrical side, creating phones that are really capable of running on-device AI, also a different problem. You’re now dedicating substantial area of your silicon to accelerators for neural nets and for other AI workloads. The AI technology itself and all the underlying algorithms and models, they’re rapidly evolving. I mean, two years ago, LLMs weren’t really a thing.
Right. Well, you said even more so and then, your example even more so is AI. Was there an extent where there was a standardization happening, and now we’ve leaned all the way back into R&D and integration? Or, is it just always been impossible to break things apart?
RO: I think, again, part of it ties back to what part of the market you want to operate in. But in the premium side, for sure, I think the leaders are going to end up being people with deep technical capabilities. It is the frontier space of computing in my view. And, because phones are with you all the time and they’re so heavily used, people want them to do everything. And so, there’s almost a sensational appetite for increasing capability within phones, which keeps pushing the envelope on what computing capability can you add to it to be able to accomplish the next task. And, I mean, I wouldn’t have thought a decade ago that people would ever be interested in taking continuous 4K video on this, and then being able to immediately upload it to a cloud. And, I don’t know, you wouldn’t have envisioned that necessarily.
I think now, phones are on the cusp of being able to, not only do stuff like that, but also become your wallet, become your keys, run advanced AI workloads, do stuff in the background for you. I mean, the amount of capabilities they have today is outrageous, and that’s only going to grow based on what I’m seeing now. Various times I thought maybe this work had plateaued, but that is absolutely not the case. I think they’re going to become more and more computer-like, and because they’re with you, they’ve got this place of importance that is difficult to overestimate.
I like what you’re saying because I always like when someone aligns with a take that I’ve had for a long time, but this bit about the phone being always with you and the most important device in your life, and I think what a lot of the integrated/modular discussion overlooked is this bit that you can’t overshoot on the quality of the user experience, you’re dealing with an arena where it’s literally impossible to be too good. So, in that case, that means to be integrated, retain significant advantages, and you’re dealing with customers where the buyer is the user, they care about the device that they have and, to your point, it seems not only is that not going away, with AI in particular, it’s really deepening.
RO: Totally, completely agree. Another example is, so we just announced that we’re making this thing which is the Google Pixel Pro 9 Fold, and there’s been others like it. But what breaks through for this thing to me is now about 10-and-a-half millimeters thick, the front screen is basically the same size as our Pixel 9 Pro small size, if you will, which is not super small, but it’s 6.3 inches instead of huge.
(laughing) It’s all relative.
RO: Yeah, and the inside screen is eight inches, so I probably open my computer half the time I used to. I definitely don’t use a tablet nearly as much as I used to. And it’s because, again, it’s almost the same size, it’s just barely thicker than a normal phone, and the same width and height. So now, I’m like, “Okay. Well, we’ve just integrated yet another capability into effectively a smartphone package”.
It’s this insatiable black hole. I remember, so this was a big thing, the whole Christensen thing, and I wrote a piece about What Clayton Christensen Got Wrong about this bit, talking about the user experience but what I wrote the following week was a less famous Article, I called it Obsoletive, and it was about the reason why you can sell a more expensive device and get away with it is the whole it’s pulling in the camera, it’s pulling in the calculator, you had the picture of all these devices that an iPhone is, and what you’re speaking to is, look, that has not stopped. All it’s doing is, the black hole is expanding and pulling in more and more devices.
RO: The black hole’s expanding, exactly. The black hole is definitely expanding, and it’s really hard for users to add another device to their portfolio. So, there’s a strong desire to simplify aspects of your life and I think that is one of the reasons why there’s been so much integration into the phone. The phone is probably the last major technology thing that’s been added that everyone has to have.
And it might be the last one.
RO: It might be. I mean, you might be able to argue tablets after that, but phones are much more important to people than tablets, I would say. I don’t think that’s a stretch to make that comment and so therefore, it’s got a place in everyone’s pocket, and everyone’s going to want more value out of it. I’d say maybe in the very early days of smartphones, the value was compounding at 50% a year, the phone would be 50% — I’m making these numbers up — but the gains were so obvious that you almost had to upgrade every year. Now, the gains, maybe there’s 15, 10, 20% better every year. But now, if you wait three years, that’s a lot.
The three-year CAGR is very high.
RO: Exactly, the three-year CAGR, you compound 15% for three years, it’s a big number. So, they keep getting better and better and with AI basically being a completely new way of computing — literally the computer that powers it is different, and the way it works is different than software engineering in the past, that is going to keep seeing big changes and differences, and it’s also going to modify how phones are made and how the computer that powers them works, and then fundamentally how the UI works on the phones. So you could just see this is definitely going to keep changing. It’s not like this market’s about to head for stasis, it’s going to keep evolving quickly and so, anyway, it’s a pretty exciting time in the industry.
Android and Pixel
It is an exciting time, I will get to AI. One more re-litigating the past question that you’re driving at, you go back a decade and everyone is convinced that Android is positioned to dominate, because it’s going to be Windows versus Mac all over again. Again, leaving aside the fact no one actually understood Windows versus Mac, it was actually Mac versus DOS, but whatever.
RO: Right.
The reason was this bit we’re talking about, that modular is going to beat the integrated play. Is it actually correct in retrospect that Google/Android was always behind the eight-ball because you had to be integrated to win? And, their challenge has been that you had to deliver an integrated device without actually controlling every aspect and that was never an advantage, it’s always been a disadvantage, and that’s still the case today?
RO: First of all, if you look at the people that are successful today, which is really Apple and Samsung, I would say, the formula for the success between those two companies has been phenomenal continued execution, and they both have brought to the table incredible capabilities across all levels of hardware and software and various different emphases between the different companies.
But the fundamental point is that everyone that’s been successful has focused on really innovating and doing a great job for consumers, and I think that what has been clear in this category is that the clear horizontal layers that were part of the early days of the PC, they’re not relevant, because of the importance of the phone and the expectation that users have that it’ll continue to get better. So, if your goal is to be on the very cutting edge of innovation, I think you do have to have a fully integrated play, I think it’s really important and it is the right approach on the premium side of this market.
I’m not sure that that’s the case in all facets, all price points, all different phones. You could probably do a much better job actually delivering $100 smartphone, which is a really key part of Android. I mean, one of the key design criteria for Android is to try to be available for everyone, and have lots of choices for users so that $100 phone is probably best served by picking the best of breed, best price value of all the components available in the world, as opposed to the most performant or the highest quality. So, anyway, that’s a major facet of Android is to be able to try to serve everyone and give them the capable smartphone.
Is this though the reason why Pixel exists? As good as Samsung has executed, at the end of the day, they don’t control the operating system. You don’t control Samsung and so is that the reason?
RO: Well, it brings it back to AI actually, this is where this all comes together. All the way back in 2016, first of all, [Google CEO] Sundar [Pichai] declared the company was going to focus on AI-first and also, when we started the hardware division, the whole reason to start it was, these models, the way of computing for AI is completely different. We felt like in order to both take advantage of the user experience differences and to make sure that we were ready on the technical side, we needed to build a dedicated focus in that area in all aspects of our business but definitely mobile was one of them. That’s when we felt it was important to start this capability. Now, we started from nothing. When I came back to Google, I asked to meet the Pixel hardware team and it was two guys.
“That’s your job, Rick.”
RO: (laughing) I was like, “Oh, okay. So we have some work to do”.
Is this why Pixel started out relatively low end? I don’t want to say it wasn’t a low end phone per se, but you had the $400, $500 models. Now, today, you’re shipping $1800 models. Is that a function of your buildup and capability and even being able to go up the user experience chain there?
RO: It’s a reflection of our increase in capabilities. About two years in, we decided to acquire HTC smartphone division so we got a number of people that were steeped in the industry.
I should back up and say something first — to make a phone is probably one of the hardest things to do, it is so hard, you’re trying to put in so much technology into a small form factor, it’s got really complicated regulatory authority and certification work, it’s got all sorts of standards you have to meet for RF transmission, it’s got all sorts of interesting computing capabilities that are required and in an ever-increasing set of consumer demand. So, it is very, very hard to make a phone, and because it’s so important in people’s lives, the amount of quality issues that users will tolerate is very close to zero.
So not only do you have to ship a new one every year, hit your targets for timing, hit your targets for cost and price points, but you also have to nail quality and you have to be innovative enough at the same time that people are interested. It’s really, really, really hard to be in that space, it’s why there’s only a handful of people that can do it, and why the people that have risen to the top in the premium end are folks that have done this very well repeatedly forever.
I mean, Samsung and Apple have been making hardware for a very long time and doing a very good job with it all along, so it takes a long time to build the capabilities just to get the fundamentals right, and it has taken us a while to do that with our internal divisions but we also are trying to pursue excellence at every layer of our stack. We want to have excellent silicon, we want to have excellent hardware and devices, we need to have an excellent OS that goes to everyone and then we need to have excellence in AI.
Another key part of this is coordination of activity across all these different layers and making sure they all are working very well together. So I finally feel great about our internal capabilities, I think in the past couple of years you’re starting to see the fruit of that come together, not only in their Pixel devices, but also how we’re starting to execute on all of our AI capabilities and today we made some really nice announcements in mobile AI that I’m very excited about.
You mentioned in passing there, “An OS that serves everyone”. How difficult has that been that Android, for a long time, we’ll get to your promotion earlier this year, but for a long time you were in charge of the Pixel division and the Android division was somewhere else and that makes sense from a Google partnership perspective, they can tell folks like saying, “Hey, don’t worry, it’s one team that serves everyone, we’re building up our hardware capabilities”, but did you feel frustrated or challenged in terms of building these high-end devices because ultimately the OS was not devoted to you? You were one of the customers.
RO: I wasn’t frustrated about it. Well, first of all, I should say the whole reason we merged these divisions together is because Hiroshi decided he wanted to do something different. Hiroshi [Lockheimer] and I have literally worked together since 1999, he’s a really good friend of mine, everyone wants to make up stories of what happened, but the reality is that he wanted to do something different, and it is the right time for us to bring the divisions together because we talked about it for the last three or four years, “Should we bring these together? What would be the reasons to do it?”.
Well, the reason to do it was because we needed to put all of our effort behind trying to innovate an AI as fast as we could. We also, when we merged the two groups together, pulled in a portion of Google Research into our organization. They’re working on Applied AI stuff and putting that all together just really makes sense now given our feeling of urgency to try to improve overall AI velocity for the company, and we felt like now is the time. The technology was at the state where you could start to build an AI OS, like you could really actually start to bake the models into your design thinking for how the OS functions and how it does things.
I think it’s going to dramatically change how applications are built and what they can do, so that was a driving reason to bring the groups together. When you have two different teams, you’re doing things differently and separately and you try to coordinate, but you can only coordinate 10,000 people in different divisions at the top so much so by bringing them together, it made it much easier to solve the coordination problems.
For me personally, I mean, certainly it’s super fun. I love this domain, I love working on this stuff, it is definitely a really key time for us to be very focused on getting AI and mobile computing pointed in one direction and really focused and probably the challenges were certainly talking to our partners.
What is this conversation with Samsung like? It’s like, “Yeah, Pixel and Android are now under the same guy who was the Pixel guy”.
RO: Yeah, it’s like, “Hey, the Pixel guy’s in charge now, what does that mean?”. Well, it doesn’t mean anything with respect to a change in strategy, we have not changed strategy at all, and so when we work with Samsung, we work with them very closely and we work with them as we did.
In fact, what I think it allows us to do better is coordinate our overall AI focus so that the Pixel folks aren’t doing their own thing, they can contribute more to the direction of travel with Android overall and I mean, the Pixel team’s going to do the very best job they can at trying to build beautiful devices that people look at and think are awesome and want to buy. They’re going to do the best job they can at trying to make world-class silicon, especially silicon that powers our latest AI models, and that’s similar to what our other partners are trying to do on their own, and our Android team tries to serve all of them.
One of the benefits of the Android ecosystem is that it can offer a lot of different choices to users. So with Samsung it’s like talking to them, engaging very closely with them, working with them in a way that hopefully builds trust and certainly one where we try to make them successful by all our work that we’re doing right now on Gemini and AI.
What’s the goal of Pixel? Is Pixel supposed to make a lot of money or is it supposed to be a route to monetization for AI? Which there are questions around that as it pertains to search. For example, you don’t have a user necessarily choosing an auction winner. Maybe you do, maybe you don’t, but now you have this, “You can make a product better with AI and sell that product”, is that the goal?
RO: First off, it’s to try to lead an AI innovation, that’s number one. It also is a business, so we want to grow it, and it has been growing and at this point, what’s kind of interesting to note, if you break the smartphone market into two different segments, let’s just choose $400, this is because of how we kind of look at it internally. But a $400 price point above and $400 and below, the market looks wildly different. So you’d say like, “Oh, Android has whatever 70ish percent share worldwide”, that’s a wildly different number if you look at worldwide $400 and above and $400 and below.
It’s a lot smaller in the high-end one.
RO: Way smaller, way smaller, and Apple’s share looks very large, especially in some developed markets, and they do operate very differently. A lot of $400 and less are sold at retail, a lot of $400 and above is sold through carriers with subsidies and financing and all sorts of different mechanisms and so those markets look very different.
In those markets, it’s pretty clear we had to try to add innovation to push back from and gain more share for Android in that space, Pixel provides another choice for that, it is our direct brand projection from Google. This is how we sort of intend Android to look for our users, and brings a lot of Google services exactly how we want them to be delivered for users without anything else.
Android, one of the benefits of it is it lets our OEMs project their brand the way they want it, and Samsung’s doing that with their brand and they have a lot of very loyal users that how they’re building products, but we wanted to have a Google branded experience too and make that another good choice for people looking for a premium phone.
At the end of day, aren’t Pixel users coming from Samsung?
RO: Actually very few of them.
Interesting. So where do they come from?
RO: They’re coming from a large number of people, some of whom who’ve left the market and then also from Apple but ultimately, we think what we’re trying to do is we’re trying to gain Android share overall and that’s my principal goal is to try to improve Android competitiveness, improve our innovation, improve the problems we’re able to solve for users at all tiers. But we’ve certainly had issues in the premium space that we want to try to address.
AI Smartphones
We talked about the mindset shift that a smartphone was actually a small computer, that was the paradigm shift that not that smartphones were a new idea, it was the conception of them was different, and that created the conditions for just a complete overhaul of the market where Apple and Samsung, to your point, are the two winners. Nokia basically doesn’t exist anymore, Motorola is there, but it’s not what it was sort of back in the day.
Is it fair to characterize or to what extent do you see AI as engineering a similar shift, and could we look back in 10 years and the dominant players are totally different? Or is it a bit of a shift and we’ll get share here and there. How extreme should we expect this effect to be?
RO: I think it’s a really big shift, but I think the world is very different than how it was in 2007. I remember when I was at Motorola in 2007, I remember people saying things like, “Making the phone is so hard, the RF is so hard, I don’t see how new players are going to be able to do this”, and nobody’s saying that about AI. Like, “Oh, AI is so hard, I don’t see anyone being successful at it” — no, I think everyone realizes this is really important, it’s going to be a very competitive space and everyone’s going to do their best to try to be competitive in this arena for users and do great things for users, and so instead of denying that people will be able to build the next thing or focus on it, everyone’s embracing it as, “Yeah, it’s pretty clearly a big shift”.
Now I do like where we’re positioned in this new world, you’ve pointed this out in much of your writing. I think our infrastructure capability is crucial to making this all work. If you want to power really advanced AI and solve problems for consumers and mobile like we’re doing in my group, but also for enterprises in our Cloud group, you have to have world-class infrastructure running at scale that is production-ready and hardened and can run AI workloads, which are very different from cloud workloads in the past. We feel like we’re in a super strong position to be able to do that for billions of users.
I also really like our research capabilities in DeepMind. We have quite a amazing set of researchers who’ve been working in this problem space for a while and have recently started producing the best-in-class models like the one that has recently been released for our 1.5 Pro developer preview model that’s now number one on the LMSYS Chatbot Arena. [Editor’s note: Gemini 1.5 Pro is now number two, behind the most recent update to GPT-4o].
On this point, on the scale in the cloud, today’s presentation was structured in three parts. You had Android first, then Pixel, then wearables, I’m focused on the first two here. One of the stars of the Android section was Gemini Live, which is both capable and it also operates entirely in the cloud. This cloud aspect, is that last decision, is that a function of the cloud being so much more capable and that being Google’s forte, or by being in the cloud you reach every Android device including those cheap ones? What drove that where it is, “We’re just going straight cloud”?
RO: Yeah, I’ll give you three. So we talk about this as a hybrid compute infrastructure, but what that means is we believe that there should be cloud components and device components, not in the old days, we call client server, but clients-
Thick thin, that’s what I was looking for.
RO: And clients network computers, yeah, I mean we’ve been through this.
This is the second Sun reference I’ve had today, I had one in a conversation earlier!
RO: (laughing) It’s quite funny. But Gemini, let me start with — I’ll kind of segment this a little bit to answer your question. Gemini app is an example of something that makes sense to run in the cloud because then we can support everyone. So the basic assistant capability that is now Gemini is now replacing, this is something that you want to be able to run on $100 phones, which we can do with the app right now and also it needs to be suitable for very high-end phones too.
So the Gemini Live capability, however, is a pretty resource-intensive capability that builds upon all the Gemini models we have, but it definitely requires more advanced models and a lot more compute capability to be able to have this live ongoing conversation that’s aware of the context and keeps going and responds really quickly. So that definitely is at first, we’re launching it as an advanced feature, and it will probably stay in that domain for a while from the things we’re adding to it.
And that’s a function of cost? Theoretically, you could run Gemini Live—
RO: That’s a function of cost, yes. That’s a function of computing resources to make it work, and we’re adding video into that shortly. We talked about that today a little bit, but also previewed it at I/O with Project Astra. We’re adding video into that, so you can use your camera to show Gemini what you’re looking at. We gave an example of math homework today, but it could be how to assemble something or a puzzle you’re trying to solve or a game you’re trying to figure out the next move in like a chess move or something, it’ll tell you what to do. That is a lot of compute, so that’s an example of something that will probably be more narrow.
Basically when you’re paying $1,799 for Google Fold or Pixel Fold, you’re not just paying for two screens, you’re also paying for Google server time to deliver this capability.
RO: Yeah, I mean, in that case, the Fold comes with a year’s worth of Google One with a AI premium SKU, so you can experience it all together, but yes, you could use it on another phone with the advanced AI subscription.
So how would that work if I want to get Gemini Live in the long run, is that something that’s going to be available for other OEMs?
RO: Yeah, that’ll be available for other OEMs.
Will there have to be a business development deal where they are actually paying you to have this capability, or if the compute cost is imputed into the Google or the Pixel price, how does that work for a third party?
RO: It’s offered to consumers, and also the third party can pay for the SKU to offer it on behalf of their users if they’d like.
Got it, makes sense.
RO: So either way, but it’s primarily consumer subscription is the primary focus.
Would anything about Gemini Live be different though if it was only ever produced for Pixel on the high-end? You have Apple doing this hybrid approach where I actually feel like Apple Intelligence, one way to think about it is it’s one of the first agents that’s going to ship at scale, and that agent capability is basically deciding what stuff should be done locally, what stuff should be done in the cloud, what stuff should be done sending it to a third party or whatever it might be, whereas you’re just going straight to the cloud. Is that an incorrect framing in how I’m thinking about that?
RO: Yes, that’s a little bit incorrect. All the Pixel 9’s ship with Gemini Nano on the device, and the reason why we increased RAM on all of them is so that the model itself could be resonant in RAM and always ready. So the models to run very quickly require a lot of memory.
Are those models involved in Gemini Live though or is Gemini Live a distinct thing from everything else?
RO: In that case, they’re not involved in Gemini Live right now, although with time they might be.
As the smaller local model gets more and more capable, we try to take advantage of it more and more, both for responsiveness and also to reduce cost and we certainly use them for some of our other capabilities. Call Notes is a good example of that where you can have a phone call with someone and then we can record it and then give you a transcript of it and also summarize it for you, so that would be an example, and we do the same thing with our Recordings app, we also showed some examples how we use it for Pixel Studio, screenshots. We use the local model a lot and it’s, again, like this multi-tier architecture I think is the right one for the long term. You want to make sure that you’re parsing the tasks appropriately, not only so our infrastructure can scale, but also so that we can make things more accessible for users at the best possible price points.
But right now, no doubt, AI computing infrastructure, and I’m referring in this case not only to the cloud, but also to the device, is more expensive, it is more high-end. You need to have more RAM, you need to have more capable silicon, you need to dedicate a substantial area in that silicon to these accelerators, AI accelerators. So yes, the cutting edge of this is going to end up being premium for a while. However, we absolutely want to cascade as much of the capabilities as we can, as broadly as we can and so things like Gemini App, we can bring very broadly to Android users all over.
A quick just dig into the details. On this call recorder or Call Notes bit, you have to invoke it every time and it will tell both users that it’s happening?
RO: Yeah, yes.
If the whole thing is this is happening on your device and it’s totally private, why? Why not just do it all the time?
RO: The far end of the call really needs to be okay with it, so you want it to be announced, that was the way we went about designing this and so you basically specify that you want to start taking Call Notes. It will announce to the other end, “Hey, this call is being recorded”, we felt like that was a pretty important aspect of the design as opposed to just starting the call and immediately recording it right away, that feels like a little bit of a stretch.
Pixel Trade-offs
One of the trade-offs of Gemini Live being in the cloud is you did lose basic Google Assistant functionality, like setting a timer or something along those lines. How did you think about that trade-off? “We’re going to have this incredible ability and Google Assistant is now gone, I can’t turn on a timer” — turning on a timer, that’s supposed to be Siri’s failure!
RO: Yeah, the overall approach with using AI to drive the Assistant is the right approach and so there have been some feature gaps to date, but we’re closing a lot of those, a lot of the legacy feature gaps are being closed very quickly. Also, you would never be able to do something like Gemini Live on the old technology we were basing Google Assistant off of.
For sure, I agree with you, I think it makes sense, how did you decide that right now, it’s an acceptable level of regressions that — this is a question I had about Google Assistant and Siri and all these types of things.
RO: Well frankly, this is another great aspect of Pixel. That’s not a billion user product, so we can make these kinds of trade-offs and really for the first rollout of Gemini, it was a choice. The user could decide to stay with Google system or switch to the Gemini App, there were some trade-offs, a lot of people elected to switch to the Gemini App. Some people were unhappy at first, quite frankly, that helped us make it better and we quickly followed up with improvements to the capability and obviously we’re still trying to make it as good as we possibly can. Right now, it’s now got a pretty high rating.
Is this an advantage you have versus say a Samsung where they have to worry a lot more about upsetting people and you don’t have much to lose?
RO: Well, I think, generically, it has both pros and cons to be a challenger and a new entrant versus someone that has a really large install base and has to be careful. Various aspects of Google have this dynamic too. It’s very, very difficult to change things at a billion user scale versus a much smaller scale for our newer products.
You mentioned that it’s really hard to make a phone. What is the relative difficulty of making a phone as compared to selling a phone?
RO: They’re both very difficult, but they’re also super related. Making a phone is a very difficult integration and technical challenge and as I mentioned before, there’s no forgiveness when you make a mistake, and because phones aren’t bought by accident, they’re really researched, any mistake is very broadly known quickly, so it is a difficult space to break into if you’re new. Also even if you’re in it, you have to continue to execute really, really well.
Once you’re there, if you have a position in the market that’s a well-known brand, then the challenges become different, then it does look more like a standard battle in a very highly competitive market, and what matters there is having prominent distribution, having really capable aftermarket support and sales support, building trust with your consumers and a reputation and those reputations are very difficult to win and very easy to lose. So it certainly is very difficult, and it’s a mature market now, so it’s not like it’s growing off the charts anymore, it is now a mature market, a lot of those reputations are locked in.
So it takes a long time, if you’re new, to chip away at it and to build a position and a foothold and if you’re doing well in that space, it’s a very lucrative business so people that are doing well in the space don’t want to give up that foothold, so they’re very willing to defend it, and the people that are successful are making great products. So it is a difficult, difficult market both to build the technology and build the products, and then also sell it and maintain a great position in the market.
We’ve talked about Pixel and Android coming together under one roof. How much consideration do you have to give in your position to Google’s overall fortunes? For example, do you have to think about Google customers on the iPhone? Or is that someone else at Google’s problems and, “I’m here to beat the iPhone, that’s what it is”? How do you balance that with getting support from other parts of Google? I would argue in the early Android days, Google went too far to support Android and actually neglected the iPhone for a few years to their detriment. Is that calculation different now just because you feel, because with AI, we’re coming from a position of strength, you are now the tip of the spear? Or is it still sort of a, “Well, it’s a big company, it’s complicated”?
RO: There’s a saying, I can’t remember what book it’s from, but it’s like, “As an executive of the company, my first role is to basically support the strategy of the company”, and from the very beginning it’s been rooted in our mission of organizing the world’s information and make it universally accessible and useful. I think that’s an amazing mission, and so that is the primary mission of the company.
My role is to try to support that mission, but also in general, I support the mission, so when it comes to organizing information like search, it wants to be everywhere and it’s going to be on iPhone, it’s going to be on Windows, it’s going to be everywhere we can get it to. Frankly, I think we’ve built an amazing product and so people want it everywhere as well, so in my particular role, certainly I would advocate to try to help improve Android and its competitiveness overall but nevertheless, when push comes to shove, I support the company’s mission overall to make sure that some of our best services are able to reach users wherever they are, and I think we take an approach of breaking down all the different layers of activity we have and we try to do the very best we can within those activities.
So I control Android and I look after Chrome, and I look after Pixel, and a handful of other things, and I try to do the very best job I can with those different activities for the users that are in that ecosystem. And so it’s—
It’s complicated.
RO: It’s complicated in the micro decisions, but the macro is pretty easy. We’re trying to serve the overall company mission.
You did this presentation live, which I love, I’m a huge advocate of live demos. I think it’s particularly helpful when you’re doing—
RO: We started planning this a long time ago actually, like February or March, and we’re like, “You know what? We’re going back to the future. We’re going back to live presentation, live demos. It might be messy”, and we wanted to really focus on things that are shipping like now.
I thought that was really a welcome thing. You’re out there saying, “Yeah, we mean this, it’s all shipping a couple of weeks”, which I think is something Google has not been very good about over the last little bit.
RO: And not only us. I would say not only us.
That’s fair, to be fair.
RO: There’s been a lot of pre-announcements recently, so we wanted to have it be stuff that’s shipping soon.
Hey, you had some good shots at Apple in the presentation, that’s a good one right there. So that said, you did have a demo fail. Was there any regret when you had the demo fail?
RO: That proves that it was real. Actually, what happened is we had an actual temporary service outage in production that we failed over, that’s why it worked third time.
Oh, that’s rough, not even your fault.
RO: So it wasn’t even our fault — I mean, it is our fault fundamentally as a company. But anyhow, thankfully it ultimately worked. I’m very surprised and happy that Dave [Citron] gave it a third go and I was thrilled when everything else after that worked well. It was almost as if we were doing that to prove to everyone that they were actually live demos.
(laughing) This really is live, yeah, this is not canned for sure. Well, I find it very compelling. I bought the first several generations of Pixel phones just because I always try to buy new phones, keep up on it and frankly, there was a year or a while I stopped doing it. It’s like, “I get Android, I understand it, I’m comfortable with an iPhone”, and now I feel like, yeah, I’ve got to go get one of these for sure. And also, Giannis Antetokounmpo is telling me to go buy a Pixel phone.
RO: I know!
It’s a pretty good point in your favor.
RO: I’m so glad we finally got to this. I was like, Giannis, your guy, uses a Pixel!
That’s right.
RO: He at least hopefully lends it some credibility for you, but I think if you want to really kind of dive in and see, try the Fold, it’s really cool.
Oh, for sure, no question.
RO: It’s really cool.
It’s like a double whammy. You get to get the full integration, like what is an AI-driven OS in practice? Of course it’s early days but also the Fold’s super compelling, I watch basketball games on my phone all the time and I like to do other stuff.
RO: Me too.
And the iPhone picture-in-picture is terrible, it drives me up the wall.
RO: (laughing) Well, I hope you like it, and thank you for doing that.
I look forward to trying it. Rick, nice to meet you, and I would love to talk again sometime.
RO: Yeah, great to meet you too, Ben. Thanks so much.
This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery.
The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly.
Thanks for being a supporter, and have a great day!