Skip to main content

How Amazon runs Alexa, with Dave Limp

The senior vice president of devices and services gets into the weeds

Share this story

Photo Illustration by Grayson Blackmon / The Verge

My guest today is Dave Limp, the SVP of devices and services at Amazon — or, more simply, the guy in charge of Alexa. Dave’s group at Amazon also includes the Kindle e-reader, the Ring and Blink security camera systems, the Eero Wi-Fi router, and a host of other products that connect to Amazon services. 

Amazon just announced a slew of new products across all those categories and since we covered all those specific products at length already, I wanted to use this time with Dave to ask bigger questions: how does he decide what products Amazon should make? What does success look like? Ring and Blink and Eero are all companies that Amazon has acquired — how did he manage making them work together? How does Amazon manage having Dave’s team work with other parts of Amazon’s huge empire — it owns Whole Foods. Does Dave have to think about how to work with Whole Foods?

And I wanted to know what the business behind Alexa looks like — Amazon sells Echo products at basically break even, it runs the Alexa for all of them for free, and it employs thousands of engineers who work on it. How does that make money? How might it make money in the future? How should we think about Alexa competing with other smart assistants, and for what kinds of business? The answers were not what I expected.

This transcript has been lightly edited for clarity.

Dave Limp, you are the senior vice president of devices and services at Amazon. Welcome to Decoder.

It is great to be here. And, Nilay, before we start, I should just say early congratulations; you guys are about to turn 10, I think, and it is stunning what you and your team have pulled off. Your business is hard, and to do as much as you’ve done, and influence as many people as you have, just let me pass on my congratulations.

Oh, thank you. I really appreciate it. When I think about our rivals, it’s only Amazon on our list of true competitors. No, I’m kidding. It’s been a remarkable ride for 10 years, but one thing I do think about The Verge all the time is, we were lucky. A huge part of our success is that we started at the right time, next to a bunch of categories that were very early. Smartphones is the one that I think about the most, for obvious reasons, but Alexa and the Echo range of devices came out right on top of us. You’ve had a 10-year career in Amazon. We have been aligned with a bunch of big categories that have grown up over that time. And there’s an amount of luck to all of this that I don’t want to discount.

I think that’s always true in all our careers, but I would also just say that, you had a point of view, you stuck with it. It was different in the right ways for a lot of customers and that’s showing. So, I like your modesty, but still nevertheless, I think you and the team have done great work.

I think we’re the angstiest tech website. So, let’s get into our feelings.

Amazon announced a bunch of new stuff this past week, tons of products. You always do events with lots of products. I do want to talk about all of those, but Decoder is kind of a more expansive show. So before we do that, I just want to situate where devices and services live within Amazon — what you are in charge of. Because Amazon is a pretty unique company in how it’s structured and how it’s formed. So, senior vice president of devices and services, what are you in charge of at Amazon?

Well, broadly, all of the consumer electronics-based products, and our strategy is to deeply couple those with services. It starts with our heritage, which was the Kindle. So it started with Kindle, and it’s evolved to a large number of other products; Fire TV, and tablets, and Echo and Alexa, as you mentioned, and many of the services that are coupled with those, also sit in the organization. We try to keep the organizations together, so that you can invent very quickly. And then, I have a couple other projects that roll into my organization. Our low-Earth orbit satellite, called Kuiper, is in my organization. And then, our self-driving taxi efforts through the company we acquired called Zoox, Aicha Evans runs that, but that also rolls into the org.

How much time do you spend on new Kindles, and Echo devices, and Alexa? How much time do you spend on low-Earth orbit satellites?

Well, it varies depending on what milestones are kind of ahead of us. So, there’s no one day that’s the same, but they all get a good amount of time. But I tend to try to live in the future a bit; I’m probably at my happiest when I’m living in the future. And so, those are also the things that have more ambiguity, and might need a little bit more input, and decision-making. But the best part of the job is — and I do have, I think, the greatest job in the world — that you can context-switch like that. And I can be deep diving on the right reading font on a Kindle in one meeting, and the next meeting, we can be talking about launch capacity for satellites. It’s super fun.

One of the things I think about a lot is timelines. You mentioned The Verge at the beginning, I’ll just use The Verge as an example. We have teams at The Verge that live on a timeline of 20 minutes; something happens in the world, our news team has to make some decisions and ship a news article within 20 minutes. That’s usually our standard. We have some timelines that are two years of reporting and research and investigation. What kind of timelines do you work on? I imagine that the satellites are a multi-decade timeline, and I imagine shipping the next Kindle might be a year. What’s your shortest one, and what’s your longest one?

It depends on the product, as you said, but if we’re doing software, it’s agile. We use an agile process and while we’ve been on this call for five minutes, there’s probably been an update to Alexa that’s happened in that timeframe. Some piece of code has rolled out, and that’s happening continuously on the software and the service side of the fence. 

If it’s a piece of durable hardware where you’re rolling out an iterative kind of product that we know how to do today, I think when we’re at our fastest, we could probably do something interesting in six to eight months. Sometimes it’s a year, but that’s kind of the cycle on those kinds of products. If invention is involved, then all bets are off. Because unfortunately invention doesn’t work on the timeline that you necessarily dictate. Often you have to come up with a breakthrough, or you have to figure out a problem that requires multiple iterations, or multiple different tries at it. And often those end in failure and you gotta restart again. So those are harder to predict.

And then, like you said, there’s some long-term efforts. Kuiper is an example, self-driving taxis is another example, and Alexa’s North Star vision of being the Star Trek computer is a third. Those projects are measured in, certainly multiple years, if not decades, in terms of how they’re going to get to the North Star.

How often do you know what’s going on? I mean, it seems like you run a vast organization. How often do you know specifically what your teams are doing next? You said, some update to Alexa is just shipped out. I’m assuming there’s an email that you may or may not read that comes through your inbox and says what’s going to happen, but how often are you the one who pushes the button to make something happen?

I don’t necessarily push the buttons. It’s been a long time since I’ve written code, even though I was classically trained to do so, but I don’t push a button to deploy code any longer. I’d be getting in the way, by the way. There are lots of smarter people on the team that can do that better. But generally, where I’m best utilized, I think, is resource allocation; it’s a reasonably large organization and it needs the proper allocation of resources, and I think that I can help with that. And then secondarily, decisions that are — in Amazon speak, we would call them, “One-way door decisions.” Decisions that require judgment, that you want to take a step back. You want to think harder about them and you want to contemplate them, because they’re not impossible to reverse, but they have ramifications if you reverse them, especially as they relate to customers. And so, those would be decisions that I’m involved in.

And then, finally, the invention process, which has its own decision framework. That’s typically where I think I can add some value. We have a huge set of inventors in the organization, and throughout Amazon, but I think that I, on the edges, over the course of time, have probably made enough mistakes that I can try to pattern match against those and not do them again.

So I ask every executive this question. You are uniquely primed to answer it, because the reason I ask every executive who comes on Decoder about their decision-making frameworks is that I know Amazon has the type one, type two decision-making framework; some decisions are easier to reverse, some decisions are hard to reverse. What is your decision-making framework?

Well, I just laid out the three different versions of when you’re making decisions, and for each of those buckets, it’s different. An operational issue that is a type one or type two (or one-way door or two-way door decision), that framework is about sitting down with high-judgment individuals, inspecting the data that is in around the decision, really putting the customer front and center. The customer, by the way, isn’t always necessarily an end consumer; it could be a developer, it could be a partner, there’s lots of different aspects. And then, walking through the recommendations the team usually tees up and saying, “Okay, which one does the data lead you to that’s going to have the best customer outcome?” That’s the framework for those high-judgment decisions.

“You never want to have somebody inventing part-time”

And by the way, if I go through a day and can make two of those and they’re right, it’s a very good day. But the invention process is very different. It has a different decision-making framework, we use a thing we call the “working backwards process,” where we don’t use PowerPoint, or Keynote, that much inside of our business. We write narratives. And the narratives are six pages long, and a new product, any new product inside of Amazon, the first page of that product, that narrative is a press release, as if you were launching the product tomorrow. And then the next five pages are frequently asked questions; how is this going to be differentiated? How would it be priced? What invention do you have to solve to be able to do this, etc., etc. And there’s a bunch of those questions that go in.

And those docs come, and we review them, but that process is much messier. It is very rare that I would see a working backwards document like that for any new product — could be the original Echo or the original Kindle — that we would approve the first time we saw it. Normally it takes many iterations of that. And what’s critical about that is that because we really want that to be as good as we possibly can make it, because as soon as we agree on that document, the decision is made. That project is green-lighted. The next step is to find a single threaded leader to run that project. Somebody that wakes up full-time, every day, their job is to make that product happen, because you never want to have somebody inventing part-time, that’s a very important thing. And so, that’s why that process tends to be more iterative. We still try to do it quickly, obviously, but it’s definitely messier and slower than that type one, type two decision-making.

How big is your team?

We haven’t said specifically, but we have said things like, there’s well over 10,000 people just working on Alexa in the organization. So that gives you a sense of the kind of scale that we’re at and what we’re trying to do.

I talk to everybody about the chip shortage. It is affecting everyone. Is it affecting Amazon?

I think, as you said, it’s affecting everyone. I think we’re all responding in our own ways. I think people oversimplify the problem and they say it’s a chip shortage. It’s not really just a chip shortage, because different components for different reasons are facing supply chain issues. And that ebbs and flows and changes every week and it’s a very dynamic situation. We see it, we’re not immune to it. I think we’ve been able to keep the vast majority of our products in stock for customers.

But that is a lot of work by a lot of teams to chase parts because of some of the shortages. By the way, it isn’t just parts. If a factory gets shut down because they have a COVID outbreak, you’ve got to figure out how to mitigate around a factory situation. So, I wouldn’t want to oversimplify this problem, but we’re not immune. I think we’ve largely bobbed and weaved thanks to the hard work of my ops team.

How long do you think the chip shortage will last?

Again, it depends on the components. There will be some components that have already freed up and you’ve already started seeing them get back to a more baseline usage. There’s some that are very long lead because literally new fabs had to be built. So that’s a two- to three- year process to come up to full scale on a new fab. So, I’d be oversimplifying it by giving you one date. It’s just impossible to answer that question that way.

You just had an event, you announced a bunch of products. Some of them are from companies that you acquired, like Ring and Eero. I know those founders, Jamie Siminoff [founder and CEO of Ring] and Nick Weaver [co-founder and CEO of Eero], they’re smart, capable leaders. They were the CEOs of those companies before you bought them and they still retain the title of CEO at Amazon, inside of your organization. Do you make them work the Amazon way? Are they out running little fiefdoms? How does that operate?

By the way, I think Nick and Jamie are great, first of all. I would say that when we acquire a company and they become part of the family, the thing that you don’t want to do is stifle their innovation, because you’re spending real dollars on these companies. I sort of look at it both ways; what do they do that should inspire us to change our processes, our decision frameworks, our mechanisms internally, and we should be inspired by those, and glom them on. But I also said the exact same thing to both Nick and Jamie, when they joined the family, I said, “Listen, at your pace, you should look inside of Amazon and look at it as like an a la carte menu, and adopt the things that you think will help mature your business. Keep your growth going fast, continue to be scrappy, continue to invent on behalf of customers. But I’m very certain, there are things inside the rest of the organization that will be able to help you.”

And so, now a couple years later in both those cases, very often, when I see a new product that comes from Ring, or Blink, or Eero, it comes in the form of a working backwards document. Not all of them. There’s still ideas that Jamie wakes up in the morning, and decides to go fund and make happen. But they use some of that because they like that. You know, we obviously ship a lot of product every year, and we do that smoothly with transportation, and manufacturing, and reverse logistics, and reliability, and all those kinds of things, those get better at scale and you get better at it. And our companies have taken advantage of some of those capabilities too. So I wish there was kind of a one-size-fits-all answer, but it does feel more like an a la carte menu.

But the nice thing about all those companies too, by the way, that I would just add to that, and it’s kind of part of what we think about when we do acquisitions, and we look at companies is, first, do they fit with our culture already, right? So when you looked at Jamie’s company culture [Ring], it’s neighbor-focused culture — they call their customers “neighbors.” It was as passionate as Amazon’s customer-centric culture. And that helps because then you’re really kind of talking the same language, maybe with a slightly different vocabulary, but it’s very, very similar. And also, those teams, the other thing that they had in common is, I personally loved all their products. I just love them. And when you’re trying to delight customers, it’s such a good start when you have a great product. You can delight customers with okay products, but if you have great products, you’re immediately running at full speed, and all of those companies had great products.

At this last event, you announced Ring Pro Bridge, Ring Bridge Pro, some order of those words, that integrates Ring and Eero technology for the first time. Is that you saying, “All right, you have to work together, kids.” Is that organic? How do you make that integration happen?

Yeah, it’s Ring Alarm Pro.

Ring Alarm Pro, see, I knew it was three words. You gotta give me credit for that.

That’s good. I do. I give you a lot of credit. The way you phrased it made it sound like there’s some dictatorship. What happened was exactly what I described in the invention process, which is, we wrote down a working backwards document basically about what the next version of the Ring Base Station would look like. And one of the things that came out of that was this idea, I remember having the conversation, this idea that you might want your internet to have backup. I live in Washington, we have wind, I have a generator because my power goes down pretty often. And so, I have power backup, but the internet is equally as important to me as my power, as we’ve gone through the past decade. I don’t know if it’s a human right, but it sure feels really important.

The idea of being able to back up the internet was, in the case of that product, to have LTE built into it so it can flip over from wired connection to cellular. And once you have to flip over, then you kind of go, “Well, how do you do that seamlessly?” Well, it made all the sense in the world that it would be on the back of a router, and we have a company, Eero, building world-class mesh routers. And so, it really started with the customer problem first, and led to a “better together” scenario. And I think that’s the better way to do it than just say, “Eero must work with Ring. Ring must work with Blink.” In my experience, those work in the short term but in the long term the organization will rebel against it because it is often not right for customers.

I’ll just ask you. I have Ring cameras at my house, I have Blink cameras at my house. I also have an Eero router at my house — I feel like you’re in my house a lot. Why don’t my Ring and my Blink cameras work together?

Yeah. I think that’s order of operations. Over time, they will. You start seeing some of it happening already, the Ring Alarm Pro with Eero, back and forth — that’s a good example. When you have these companies join Amazon, they have their own roadmap already. When you buy a high-growth, innovative company, the recipe for failure in any acquisition like that is: screw with the passionate founder that is inventing and growing this brand. Don’t do that. Let them do what they do best, which is invent. But over time, you can start seeing the things come together. So this year, Alarm Pro; last year, we had Alexa integrated, so that Alexa could answer your doorbell. Now on Echo Shows and Fire TVs, you can see your doorbell on your screen right in front of you. That’s integrated and plumbed through. So over the course of the years, we are making more progress on all of these and that will continue. So yes, there will be a day, I hope, where a Blink customer and a Ring customer can intermix them. Today, you can have them both connect to Echo Shows. They come together through those APIs. But over time, we can do even more.

This problem seems kind of fractal at Amazon, which again is a loose... What’s the phrase? It’s loosely coupled, tightly aligned.

I think we’re loosely coupled organizationally, but as you said, we’re tightly aligned strategically.

Right. Amazon’s management ethos is as disciplined as any company I’ve ever seen. At the same time, you run devices and services inside of your organization. There are multiple companies starting with an Echo Dot that looks like Mickey Mouse to a satellite broadband system. And then you scale out to Amazon. Amazon runs a grocery store. Amazon obviously runs a retail business. They run AWS, the largest cloud computing system. Is it the same set of operating principles, one step up? Do you ever get commanded to work with Whole Foods in any way, or do you just run into each other in the office and say, “We should find a way to do groceries on Alexa.” How does that work?

I’ve been here 11 and a half years, I’m not sure I’ve ever been commanded to do something like that. I’m highly incented to do stuff like that though. Because things work best for all sorts of constituents, it could be customers, developers, other types of people, is when we do connect the dots around Amazon. A good example of that is Prime. Prime started as a two-day shipping benefit, super important and really good. But now you can get Amazon Originals, Wheel of Time is coming up, you can get prepared for that.

That’s a good plug.

I’m here to help. Actually, I’m just excited about Wheel of Time to be perfectly honest with you. 

Prime also has a reading benefit and it has a music benefit. As each of those happen, the experience for a Prime customer gets better, but it doesn’t necessarily end there. I’ll use Kindle as an example, our oldest product line, still one of my favorites. But when you start connecting those dots and you add a Prime benefit, then you can bring in other parts of the virtuous circle, in that case, authors. My mother-in-law was an author. She was actually lucky enough to have an agent and she writes cozy mysteries, but so many authors didn’t. Now there are hundreds of thousands of authors that are making a real living off of Kindle because they can directly publish. It’s one button and you publish. And then all of a sudden, now customers have more selection. And then therefore, they come back to read more and literacy goes up.

So when you start connecting the dots, there are really positive elements to those flywheels that are kind of unseen, but I can give you dozens of examples of those throughout Amazon. And so it’s not that you’re forced to do it as a leader at Amazon, but you see the delight and you see the oversized impact when it works. By the way, a lot of times we’ve tried it and it doesn’t always work. There are plenty of examples of failures, but when it does, then you start seeing the real benefit.

Let’s talk about all the products you announced last week. You tend to have rapid-fire product announcements. So last week, you announced some new Ring cameras, the Ring Alarm Pro, I got it right that time. There was a Ring home drone. There’s an Alexa robot. There’s an Alexa device for kids to video chat with grandparents. There’s a new Echo Show. Why announce all the products at once? It’s very different than other tech companies, which try to focus you on one at a time. Why do the rapid-fire events?

Well, the fall is when we tend to launch a lot of products. We don’t dictate this, but we generally try to do some things in the spring and then we do a bunch of stuff in the fall. Those tend to be just the cycle we’re on in terms of invention. Doesn’t mean lots of stuff doesn’t happen in between, but the big rocks tend to do it that way. So the fall was not just, it’s the most recent, but we also had a couple other launches as well. We had a brand new Kindle that came the week before, and three weeks before that we had a whole host of TV-based products, including our own TVs for the first time, and a new 4K Fire TV stick.

So it’s really about these things coming together. And then why do we put them into one event like that? We debate that a lot, to be honest with you. It’s a function of, if we dribble them all out, then I think it’s also just harder for a customer — and by the way, you are a customer of that. Because your job is to report, good, bad, and indifferent, the news. And for us to kind of dribble it out when we know you’re busy with other companies’ events and lots of other things going on, it feels like it has some efficiencies to it. I’m not sure it’s perfected though.

But when you look at the horizontal nature of what we’re trying to do, we have a lot of different product lines and I’m very proud of all of them, but that also just adds up. We have a lot of teams inventing and it adds up and some are net new and they’re a little bit more audacious, and sometimes they surprise people. And some of them are more iterative, but they are based on the fact that we wake up every day and we have a 360 feedback loop from our customers. They put reviews up and we read them and they call our customer service, 7 by 24, 365 days a year, and we listen to those calls and we know what we have to fix in our products. We know what we have to do for the next generation of our products and for a lot of our iterative work. And so you add all that up and it tends to be a busy one hour.

Yeah. I’m always curious how tech companies plan their marketing events of this kind. Because there’s actually only a handful of strategies and most people, to be candid, have drifted towards copying Apple’s strategy. But there, I think the way you all do it is very different. I’ve always been curious about what drives, is it, “We just don’t want to look like everyone else is starting to look,” or is it, “Man, we got a lot of stuff. We just got to get it out the door?”

It’s been a long, long time since I was at Apple. So I don’t know their strategy. I watch their events, certainly, but I do feel like they have their viewpoint on what they think you and the end customer want to hear about. I think I can only speak to our viewpoint, which is, what we want to try to get across is that we’re inventing on behalf of customers. That is the number one thing I would say. So do we have the most polished video when we start our events? No, we generally use UGC [user generated content] based video of stuff that we think is fun, that customers send in. And it’s a little grainy and it’s not as polished. but it’s real. Somebody is proposing on the front doorstep in front of a Ring camera and they sent it in. And it’s like, “Oh my God, what a joyful moment.” And an Echo is helping somebody that has limited eyesight listen to a book and that’s just a joyful moment.

So I think that, not speaking for others, that’s the lens that we want to introduce our products around. And then when you have things net new to the world, you mentioned Amazon Glow for trying to get kids and grandparents together, or a robot, or Always Home Cam in the case of drones, then you have to do a little bit more explaining because it’s like, “Well, why did you do this?” And so there’s a little teaching and storytelling as part of those, because they’re net new to the world and you do want to put them in the context, right or wrong, our context, the reason why we invented them, because we don’t yet have customer feedback on them, to be able to tell you the stories through the customer’s eyes.

Well, let’s start there. That’s a perfect segue. The drone, announced last year, you’ve taken preorders now. You’ve gotten a year of feedback, some very positive, some very excited, some very cynical. Do you have any numbers to share, are people preordering this thing?

Yeah. It’s off to a great start. It’s invites. It’s not preorders yet.

It’s invites.

It’s invites, but yeah, the invites are strong. I was having this discussion with Jamie the other night, it’s much stronger than I would have predicted.

Amazon is famous for kind of squishy language instead of numbers. Is there a number that you can share?

I go to school for squishy. No, I can’t share a number. But it’s stronger. But that doesn’t really matter. The number is the number. What matters is when we ship it and customers start reviewing it, and are they using it? That’s all that matters. That’s 99 out of 100 points. If we get that right, then good things will come. If we don’t get that right, then we need to quickly pivot and iterate and get it right. But as invites as a measure, it’s off to a better start than I would have predicted. The fact of the matter is any new-to-the-world product is going to have its cynics. By the way, I would postulate that if you don’t have vocal cynics, you didn’t take enough risk in your invention. It’s just another “me too” thing. Like it’s not new to the world. And so I would say, for a good new product that took a bunch of risk, there’s always going to be that noise.

So the Ring drone is a Ring camera that can detect noise, motion, take off, fly over your house, show you what’s happening. You have the Astro Robot, which you also announced. Not a Ring product, an Amazon product, that is a robot that can roll around your house with a camera on a periscope and a screen and look at stuff. Why is the drone a Ring product that can look at things and the Astro not a Ring product that can look at things?

If you were drawing the Venn diagram of Always Home Cam and Astro, they would overlap in terms of a mobile camera. So there is an area in the middle, but the Always Home Cam drone is dedicated to that task. And when it is not doing that task, it is sitting in its dock, camera obfuscated. It’s inside the dock. It’s not even out. And it does nothing else. That is its utility. And by the way, I think there’s a lot of utility there. So, from a brand perspective, that fits unbelievably well under the umbrella of “make your neighborhoods safer, your home safer,” which is Ring’s brand promise. And so that was a no-brainer. It’s a Ring product.

And so then, of what Astro does, there’s a small segment of it which is a security use case, for sure. It can do more than the Always Home Cam in many ways. It can see different places that the drone may not be able to see. It can do multiple patrols. It can patrol when you’re home. The drone will be too loud, you don’t want it to patrol when you’re home and sleeping. But it also has a bunch of other functionality. It has Alexa functionality. You can listen to music through it. It can bring a call to you. One of the things we pointed out at the launch event was for elder care, a loved one, that you might feel more comfortable putting it with them so that it can give you peace of mind. And so that’s a much more horizontal thing. So what we did is, going back to our first part of our conversation, is we deeply integrated Ring into Astro.

So it works with Ring. And if I brought up my Astro app right now and I turned on the camera, I would see it in my Ring app and my mosaic of cameras. But the brand was better for something completely net new. By the way, we also didn’t call it Echo. Because it’s more than just an Echo. And so it made a lot of sense to come up with something new. And in the end, like all naming meetings, we landed on something and some people will love Astro and some people won’t.

The value of the camera is usually for the person on the other end of the camera, especially for Ring cameras and Astro cameras. The person who the Ring camera is looking at rarely wants to be the subject of a Ring video. It’s the person who’s opening the Ring app and seeing what’s happening on the camera. That gets Ring into trouble a lot, right? There’s a lot of criticism that Ring is integrated with police departments, there’s privacy concerns, there’s opt-in, opt-out. Most of Ring’s cameras are designed for the exterior of your home. Now you’re bringing those cameras inside the home. They’re mobile. How do you think about that relationship between who gets to be seen, how this video goes to authorities, how Ring acts as that intermediary? Does that come up a lot in these conversations?

What does come up a lot is the use case for a camera that is focused in your home versus out of your home. Cameras focused out of your home, you want to make sure it doesn’t touch on your neighbor’s property. So you can draw a polygon and have keep-out zones. You want to think about how you get notified on it because it’s often a notification that drives, it might just be a doorbell event, but it could also be somebody walking past your spotlight camera, or more often than not, it’s you want to scare away a deer. That’s a real use case, it’s eating your garden or something. So outdoor cameras have their set of use cases.

Indoor cameras have a very different set of use cases. By the way, that’s what inspires something like the Always Home Cam, which is some customers will not want to put a camera in every room. They just won’t. 

I am one of those customers.

Yeah, exactly. There are some that will want to put them in some rooms. They want to keep track of their dog, and their dog or cat normally sleeps in one place and they want to have that experience to be able to drop in from time to time on their pet. So for the customer that doesn’t want a camera in every room, but when they’re away, they’re on vacation or they’re at work, and something triggers and they want to see what’s going on, is it a false alarm? Or is it something actually real? Being able to have a drone go straight there, but not have to have the idea that you have cameras in every room, it seems like a really good solution for that.

Now, the macro question you asked — the first thing is, our mental model is things that are indoors or inward-facing versus outward-facing, we have a very different mental model. But there is a horizontal level of privacy that’s underneath all of those. You do have to listen to the feedback from customers, pundits, and I think we have. What was our original first responder program for police departments, but also fire departments and other things, we heard the feedback from customers and we’ve totally changed that program over the last two years because of that feedback. I think what would be irresponsible would be not to listen to people. So now, a police department has to have a valid case number. They can’t just ask for the videos. They have to put a request out. Customers have to opt into it. It’s a much different program than it once was.

Another example of that was some customers just don’t want anybody to ever have their videos. It was not easy, and we’re still rolling this out by camera, but a bunch of the Ring cameras now allow you to turn on end-to-end encryption. Then, even Amazon, we don’t have the key. We couldn’t give the videos to anybody if we tried because we don’t have the ability to unlock them. We rolled out the first versions of that, maybe six months ago or so. Nobody else is doing that. So I think our job is to stay ahead and when we do roll things out that customers react negatively to, or pundits, that we should listen to that feedback and when they’re right, we should be willing to change our minds. And I think you saw us do that in that case. 

How long do you think that cycle takes? So Astro, let’s use Astro as an example. It’s announced, there’s an endless amount of feedback, positive, negative, that you can find about it. But no one has it yet. No one’s experienced it yet. Do you wait until some X number of people have it and they start leaving reviews before you’re like, “Okay, we have to start incorporating feedback,” or is it, “Hey, this thing isn’t shipped yet. We’re seeing a lot of feedback. We can adjust the product before it goes out the door?”

Well, first of all, there are people that have it. They’re just not people that aren’t Amazonians. So we don’t just build it and then ship it off. I’ve had a version of Astro off and on in my house for over a year. 

Is it any good?

I love it. Yeah.

Be a product reviewer. What’s good and what’s bad?

Well, it depends on when you had asked me that question. If you’d asked me that question a year ago, I would’ve told you it can’t find its way back to its dock every time. But now, we’ve solved that problem. And so over the course of time, software gets better and that’s why we put these things in people’s homes and try them out. And it doesn’t mean every customer loves it, that is beta testing it. But we don’t ship it until we think there are enough customers that love it, that we can find those customers out in the wild, real consumers that will also share that conviction.

And I have an anecdote: Every week, because we rotate units from beta tester to beta tester,

“Many of the beta testers send a note to me or directly to the Astro team, and they say, ‘Don’t take it away.’”

Now, obviously, they didn’t pay for the device, so that doesn’t guarantee success, but it’s a good signal that people love it for certain types of use cases.

But to answer the specific question — I don’t mean to duck it, I just wanted to give you a sense of our process. Once we get over a threshold that we think external people will love it, customers, then we will start getting that feedback instantaneously. But we’re also not naive. I read your articles. I read other articles. I am on Twitter. I read what people are saying about it. But I look at those comments through an optimistic eye and also understand that everybody right now is working off limited information. If I had showed you that exact same group of Twitter feeds, Reddit feeds, articles the week after we announced Echo, the cynicism would be dripping. It was like, “Oh my God, how could we possibly have this Pringles can with microphones on it in people’s kitchens? Amazon’s insane. This is destined for failure. And it’s like, six months later, here’s their next Fire Phone.” That was basically the sentiment.

I was very pro Echo, by the way.

Well, I like that. But nobody had it yet. Nobody had it in their house. In that case I probably had an Echo in my home for a year and a half. And they didn’t know how delightful it could be. By the way, it only did 13 things when we shipped it so it had a long way to go.

I want to talk about Echo in a second, but let me ask this last question on Astro. You’re familiar with my colleague, Dieter Bohn.

Sure.

Dieter and I have a million philosophical debates every week in our other show, The Vergecast. After Astro was announced, we had the following debate: What is a robot? Because you have been quoted saying, on a long enough timeline, every household in America will have a robot in it. Astro’s your attempt to begin this process. Candidly, I think Astro’s neat. I like things that move around. I like robots. I like things with smiley faces on them. I don’t know what the utility of it is. It doesn’t have arms. What is the task it’s going to automate for me in my home the way that I might think of a Roomba? A Roomba automates the task of vacuum cleaning. What do you think the specific utility of Astro is that makes it a robot? Because that to me is a philosophical question.

I will answer the question. But I would not limit the definition of robot to only direct problem-solving of a chore. Those are very good robots, by the way. And our fulfillment centers are filled to the brim with robots that have made our associates safer and more productive. We still hire lots of people, but the more mundane tasks can be picked up by robotics. Those are very good use cases. And a vacuum obviously has been a very successful home robot.

But there’s also just the delightfulness of having something around that is a companion too. That, I would also say, could be a quality of a good robot. I’ve bought every version of Aibo that’s come out. And that is a robot and it’s delightful. It doesn’t solve any problems. And you could argue it didn’t get big because it didn’t do that. That’s a different philosophical argument. But it was delightful in its own way because it had a personality and that kind of thing. And so I think what Astro, getting back to the root of the question is, it does have some practical things it solves. I am also one of these people that does not put a camera in every room of my house. I have a camera inside my house, a non-Echo Show camera, I have one dedicated camera; on my app, it’s called “liquor cabinet cam.” I have teenage kids and I decided instead of a lock and key, which is inconvenient for me, I just pointed a camera straight down on my liquor cabinet and it’s unbelievably effective.

That’s hilarious, I have to say. If there’s one thing everyone can take away from this, it’s liquor cabinet cam.

There you go. My salient parent advice, there it is. But when I travel, I do care about security. I really do. And so this product has changed that for me because it does go around the house and it does a routine. And I feel better about that. It’s peace of mind. That is an absolute use case.

The other thing that I would say is

“Moving a camera around your house on wheels at a cat- or dog-eye view is strangely delightful and more useful than I would’ve predicted a year ago.”

By the way, that use is, go and talk to my kids when I’m at work. I haven’t been staying in the basement quite as much as maybe you. I’m at the office. We’re building some things, so I’ve been here more. It might also be me checking on the dogs. I use that example in the launch. I use that more than I thought I would. It has utility for me. And there’s some urgent use cases, which are “you left the stove on,” “you left the iron on.” And those kinds of things. Certainly we’re limited to one floor right now. But those are real use cases.

And then the other one that has resonated with folks is something I talked about earlier, which is peace of mind for a loved one. If you get slightly less mobile, in my case I want to talk to my dad, and I want to communicate with him, the robot can take the call to him. The robot can also just be around. It has these gravity wheels. It figures out where you are in the house and hangs out where you are. That’s also very comforting in a strange way. It allows, then, for me to know that my dad was up and about every day. And so that also is peace of mind.

And then finally, I would just say that there is something about Alexa near at hand that is also useful. Not required. You could put an Echo Show in every room and it’s fine. I do have a house that’s very Alexafied. It does have that unique quality. But if you don’t, having Alexa, that just is right there and ready to answer the debate over what actor is on the TV at that particular point in time, or what might be going on in world events, is also very useful.

And by the way, I know you said you want to get back to Echo, but it will be like Echo too. The second we ship it, somebody will figure out something we totally didn’t catch in beta. And the best example on Echo that I use a lot is Smarthome. When we shipped Echo, it didn’t do Smarthome at all. And how did people make it work? Well, they hacked our to-do list. They did an HTML scrape on the to-do page that was up on amazon.com slash whatever Alexa to-dos and then they used IFTTT to trigger something. And so as soon as we saw that, we’re like, “Oh boy, how did we miss Smarthome? That was a mistake.” And we fixed it.

Let’s talk about Echo in the context of Smarthome. Another well-anticipated segue from Dave. But Alexa has gone from kind of a science experiment to now this ubiquitous piece of technology. It has worldwide brand recognition. All the research I have seen still says people are mostly using these voice assistants to set timers in the kitchen and play some music. And those are the top two use cases by far, and everything else is single digits in whatever order. I use it for smarthome stuff, but single digits for everyone else. What does your research say? Is that basically the same? Is it different?

It’s different. It’s a much longer tale than what you just described. Is music popular? Yeah. It sort of brought music back into the house. But behind the scenes of Echo and Alexa, and we kind of started talking about this a bit at this year’s launch event, is this concept of ambient intelligence. A lot of people want to go and say, “Oh, Echo and Alexa is this operating system.” That is a 1960s view of what gets built on Alexa. The ambient intelligence looks more like a fabric, like a microprocessor fabric where there’s lots of on-ramps and lots of off-ramps. Sometimes Alexa is involved in the middle of that and sometimes it isn’t. But it’s kind of routing all this complicated thing that’s going on in your home, increasingly in your car, in a very intelligent way, often behind the scenes ambiently. And so that’s a way of saying that, yes, do people set timers? For sure. Do people do music? For sure. But people ask an unbelievably large number of questions to Alexa.

Do you think of Alexa as a general search engine in that way?

I don’t think I’d call it a general search engine. It’s an ambient search engine, whereas a general search engine, you’re often spearfishing for a specific end goal. The things that people tend to ask Alexa are the things that are top of mind and right around you at that moment. It might be just a quick help on homework, or it might be a quick fact, or it might be who is that person on the TV show? I think over time it will evolve to more search like that. But today, the phone’s pretty good at that. And so it takes a whole different form of search.

But I would just say that Q and A is really important. Smarthome is bigger than you would give it credit for. It’s not just you. Now, you probably are more advanced down the threshold of making your home smart and we have a lot of work there to make it easier for customers. But the number of people that just have one smart plug or one light bulb is very high. And every single week, there are billions of interactions with Alexa, and music does not make up that number.

How does Alexa make money?

Well, there’s places it’s already making money and by the way, not only for us, but for others. We mentioned music and streaming music was around, but it had migrated to your phone basically. Everybody had headphones in, like this, which I’m doing right now to facilitate this podcast. But the music experience had gotten to a point where it was one-on-one. What happened is Alexa, and now smart speakers in general, has moved that to an ambient experience that you can share with your family and friends in the house. And that has increased home usage of music, many many-fold, and not just for Amazon Music, but for Spotify and Apple Music, etc. And all of those are making money, including, by the way, Amazon Music benefited from that.

Similarly, listening to audiobooks, which we have a service like that, which is Audible. A lot of people listen to audiobooks and we can make money on that. We mentioned Smarthome. Smarthome is up hundreds and hundreds of percent. You can just look at when we launched our Smarthome domain, the functionality of Smarthome on Alexa, and track Smarthome device sales. We look at it on Amazon, but we’re a fraction of the overall sales that are out there.

Yeah, but let me ask you this. I own an Alexa device, an Echo Show. I’ve got a Philips Hue Bulb for example. I pay you once for the Alexa device. I pay Philips once for the Hue Bulb. I say, “Turn on the light.” You don’t make any money when I say, “Turn on the light.” The money’s already made, and you’re just incurring the cost of running all the cloud services and the Alexa ambient fabric; where are you generating direct revenue from Alexa?

Well, I mentioned music, mentioned Audible.

Every time I say, “Play a song,” the Alexa division of Amazon gets money?

Well, Amazon Music gets money. We’ll take equal credit for that. We get some of that and Spotify gets a bunch of that and others do.

Let’s use the smart bulb example, because I want to play that one through. Which is, you buy that first smart bulb, hopefully from Amazon, but you might buy it from Walmart, but then you’re going to likely buy five more smart bulbs. And then those will burn out and you’ll need new ones over time. And over time you might then migrate into something else that is Alexa-controlled because that’s become your new normal and you expand that. And we’re not going to get every one of those sales on Amazon. But every time you buy a bulb on Amazon, we make some money. By the way, Walmart should make money if you buy from them too. I’m fine with that. That’s okay.

I’m just trying to assign direct revenue to Alexa in the most basic way I can think of. That makes sense. Alexa inspires me to buy more smart appliances, smart devices. Amazon has a retail store, okay, I buy it.

I’ll give you a more direct example, which is, we make it better when Ring and Echo are together. By the way, we do that for Arlo and others too. And now you might choose to take a Ring subscription, which is $3 a month. And maybe you’re just getting hung up on where the revenue flows directly, but then that at least rolls up into two divisions that roll into me.

Sure. I buy that too. I just, I’m comparing that to other gigantic platform vendors. I talk to a lot of other executives and creators in the show who are deeply aware of where the platform extracts rent from them. You are Twitter. You want to do Twitter subscriptions. Apple’s going to take 30 percent of every in-app purchase. This has led to lawsuits and angry regulators, the whole thing. None of that is occurring on the Alexa platform at this moment in time. And I’m wondering, is that a thing that you think about? Is it all there to support Amazon’s other divisions like music and the retail store? Or is it we’re going to run ads on it? What does that future look like, or does it never need that direct revenue because Amazon is so vast?

No, it’s not the latter. We’re obviously a for-profit company.

Amazon?

And we want to make money and return value to our shareholders while delighting customers. That’s the most important thing. I don’t really even think of Alexa as a platform, but other platforms as you described it. It often takes them a long time to settle out that business model. The first iPhone, best to my recollection, did not have an App Store on it.

Fair enough.

It had bookmarked HTML pages that were formatted for that screen size. And they found their way to an App Store, by the way, I think rightly so.

And so I think we’ve already found a bunch of things. You mentioned shopping on Alexa is super convenient and people do that all the time, and that benefits Amazon for sure. But we’ll also find new places. The most important thing, and it’s true of a lot of these other products, is you want to build something that is delightful and that becomes part of customers’ everyday lives. If you get those two right, monetization will follow in many different ways. And there are many music services on Echo today, including one of our own, that are ad-driven. It’s not as if there’s no ads on Echo. Podcasts, I can get your podcast on an Echo and it has ads in it. We’re not religiously against any business model necessarily, but we’re more focused on creating that delightful experience and also making sure that people want to use an Echo or an Alexa-enabled device every single day of their lives.

Let me just tie a bow on the music example, so I have it clearly. When I ask my Alexa to play a song from Spotify, Spotify has to pay the labels. Does Spotify have to pay Amazon?

Not on the non-ad based service. If they’re doing an ad-based service, it may be that we insert some ad inventory into that. And there might be some shares in that kind of case. But if it’s their subscription service, we don’t take anything from the subscription service.

That tracks with Fire TV, for example. If there’s a streaming service on a Fire TV device, and they have ads, you participate in that too, right?

Yeah. We participate in that. On Fire TV, it has a more typical app store. And so there, if somebody uses our payment mechanism, then it depends on the partner, but it has more app store-like economics associated with it.

I brought up Fire TV because it does have the traditional app store economics, so you can’t tell me, “you don’t want to compare yourself to another company.” Do you think Alexa will ever have app store economics the way that Fire TV does?

Well, we’re experimenting with that now because we’ve added in-purchase on Skills, and we’ve had that for almost a year now. And people can buy a dedicated sleep skill and those kinds of things. The percentages vary depending on where you are in the world and what it is, but those look more like what I’d call traditional app store economics.

If I had to offer you one direct criticism of the Echo devices versus the Google devices, is that the reason that I have Google Home Hubs in my house is because their integration with Google Photos is so sticky and Google Photos is such a great service. It would be great if I could plug Google Photos into my Echo devices. At the same time, it doesn’t seem like there’s a lot of pressure on the Amazon photo system to compete with Google fairly directly.

It’s nice. Google’s just way ahead and there’s not a lot of rapid iteration on the Amazon side. Is that something that’s under your control? Does Amazon Photos roll to you? Is that something you think about, or is it one day we’re just going to write a deal with Google and run Google Photos, too?

Those aren’t mutually exclusive, but to the question directly, Amazon Photos is in my organization, but I’d also love to entertain a deal with Google with their photos. That’d be great.

Has it ever come up?

I haven’t checked recently. I know we’ve talked to them in the past. I wouldn’t want to talk about any specific conversations, but we have a very long track record on Fire TV, our tablets, and I’ll add Echo to that mix, that we want to add other providers of services. We have Apple Music on Echo. And by the way, I love that. I pushed very hard to try to get that deal done because I think selection does matter to customers.

We offer Facebook Photos. We have our own photo service, but it should be sets of open APIs so that other people can participate. So, to me, it’s also a function. It’s a two-way street for any of the other partners. They should hold up their hand and say, “We’d love to participate,” but be it smart home or music or any of these, what we call domains, we want to have as much selection for customers as possible, and I think we’ve shown that time and time again by onboarding lots of external services, even when we often might have a competitive service, but that’s fine. That’s what’s right by customers.

I feel like a big function of Decoder is corporate matchmaking. Next time I have a Google executive on, I’ll ask them to talk to you and you can get this done.

Yeah.

I want to end with an expansive question about Alexa. And you and I, in ancient history, have talked about this before, but it’s been a long time. We’ve learned a lot about microphones, voice assistants, in our home and what they’re like. There’s a part of me that says, “Smart speakers, microphones, to a certain extent cameras, are just hardware. They can just be in all kinds of devices, and that the service backing them, whether it’s Alexa or Google Assistant or Siri, is actually the thing that makes them valuable.” But right now, all of those things are integrated into single devices I have to buy and potentially throw away if I want to switch ecosystems. And then, I’ve got to switch my smart home ecosystem. And then, I’ve got to buy a whole bunch of other stuff. Is there any movement towards, “We should just make the house smart and enable it for ambient computing? And then, you can select an ambient computing provider, or you can run two of them at once, because you talk about locking on the phone.

When I think about my smart home, I’m legitimately confused as to what would happen if I ever sold my house. I don’t know how to explain the absolute house of cards and hacks that operates my house to another human being, but it’s full of microphones and cameras and speakers, and they all work together. Is there a thought to, this hardware is actually a commodity and what we’re providing is a service that makes it valuable, and the whole industry should align around it?

Well, it feels like you basically took a page out of our playbook by saying that. I don’t know if you’re teeing me up to tell you our strategy, but that is our strategy. The issue is we do believe in that future. There are not many universal truths in the world. Customers will always love lower prices. That’s a good universal truth. But one I’m sure of is, whatever technology you have in your home — whatever group of technology, broadly speaking when I say technology — the only thing I’m sure of is it’s not going to be homogenous. There is no one company that can build everything that you want. It’s just not going to happen.

And so, as soon as you come to that belief, that that is a universal truth, then the next thing you have to come to grips with is you better be working really hard on open interoperability. And so, everything that we do, from our smart home APIs to our voice initiative to have interoperability, multiple voice assistants on the same thing, is about that, which is you should be able to talk to the smart speaker of your choice and ask — you said of Google with photos — you should be able to ask Google for your photos, and you should be able to ask Alexa for what you think Alexa is good at. And same thing with Smarthome and all these other things.

I think we’ve put our money where our mouth is there. We’ve driven this initiative, we’ve driven standards, like Matter and other types of things for smart home standards. We’ve opened our smart home APIs, we don’t charge for those APIs. And just last week we put Disney’s voice assistant, or we announced we’ll do it early next year, but we’re going to put Disney’s voice assistant on our own devices, just to show that we can do that. We put Alexa on Facebook’s voice product, Portal, as another example. And they run simultaneously and they run just fine. Customers don’t get confused.

And so that is the future. And we just need to, over the course of time, convince the rest of the industry that that’s the right place to be, or they can convince us that we’re wrong in some way and we need to adapt our strategy. That might also be true.

Well, there are two companies to convince. Disney has a voice assistant. It’s a pretty targeted voice assistant, so it makes sense that you could run them in parallel. Facebook doesn’t have a voice assistant. They’re not trying to do voice assistant stuff with the Portal. It makes sense they would ask you. Apple’s trying to do what you’re doing. Google is trying to do what you’re doing. Have you talked to them directly about, “Hey, we should be able to run multiple assistants at once?”

I won’t talk about private discussions, but we’ve said publicly that we want to have everybody that is willing to join as part of the voice interoperability initiative. That’s one component of that. And as to date, those two companies have not joined, but we’d love to have them part of that. But there are places where we are working together. Matter, for example, for Wi-Fi-based smart home devices, that was lacking a standard. In those cases, those two companies have taken a leadership role. By the way, I applaud that. I love that, but on the voice assistant side, as to date, I won’t speak to their reasons, but they have not chosen to go that direction.

Do you think it is correct to assume that the hardware will get commoditized here in the values, in the service layer?

The hardware for an Echo Dot is $50. And then, on Black Friday, it’s something lower than that. The strategy was never to build really expensive hardware. It’s not like a high-end smartphone. It is about a thin, hopefully in our case, our strategy is to have that ambient experience disappear and just blend into your home. And the intelligence is either on the edge someplace, increasingly, because we have more silicon we can put there that’s low-cost, that can do the edge compute, or in the cloud. It’s all about that ambient intelligent fabric that is making all this happen. But that doesn’t mean, just because it’s not on a device, that it can’t be interoperable. Of course, it can.

I’ve always wondered this, the price of the Dot is pretty cheap. Eventually, it feels like you’ll be able to lock in a Whole Foods order and just put one in your cart for free. Do you make money on the sale of an Echo Dot?

Our business model for almost all our devices, there’s probably some notable exceptions, but almost all of them has been to try to sell our devices at breakeven. And listen, are there some days where it’s slightly below that because it’s on sale, but there are other days where it’s slightly above that. The goal is to try to average out at breakeven. And by the way, other people have very different business models and they do just fine, but we’re really about the service coupled with that device.

And if we sell you a Dot and then you never use it, you put it in the drawer or it just sits unused on mute the entire time, I believe we’ve failed. That we haven’t given you utility. And so, why should we make money? What we want to do is make money if you’re using it over the course of time. And that goes all the way back, again, to the first device with Kindle. We shouldn’t make money when we sell you that device, we should make a little bit of money every time you buy a book. And if you’re reading and enjoying it, we’re all going to benefit. The author is going to benefit. We’ll benefit a little bit, and you’ll be joyful because you get lost in the story.

And then, we’re also incented to upgrade those devices, to keep upgrading on new software. We don’t want to tell you every two years you need a new version of that device. On one of my TVs, I have the original Fire TV stick, not the original Fire TV box, but the stick. And it’s substantially similar to what we shipped in our latest 4K. It’s not as fast, it doesn’t have all the speeds and feeds, but the interface and what I can do with it is equal. We’ve kept the software up to pace, and the business model incents that, and I think it’s really aligned with customers.

I look at a lot of other tech companies and they are all chasing a form factor after the phone. They’re all chasing AR or VR or the metaverse or some other thing that is not the phone, because the phone has been won and lost and we know what that landscape looks like. And the next form factor is where the next winners might emerge.

Amazon has ambient computing, it could be the next form factor. But you’re also making TVs, you’re also winning in ways that matter to people right now, and aren’t pie in the sky. Is Alexa the next form factor? Is it the next thing that will happen, that will break the smartphone the way the smartphone broke the PC? Or are you just focused on delivering the value now?

Yeah, I just don’t believe that the next form factor breaks the previous one. I’m talking to you on a PC right now.

I think the pandemic has actually changed that debate. I’ll be honest with you.

I never got rid of my PC. If I have to write one of those aforementioned six-page documents, that QWERTY keyboard, as old and as obsolete as it should be, is still the best way to write a six-page document. Now, did smartphones move us to another level? Yes. In the same way, ambient computing, Alexa and Echo and this ambient intelligence, is not going to supplant the computer nor the smartphone. It’s not going to. It’s doing something different, but I would argue that it’s already here and here to stay. It’s at scale, people are using it, they love it, they are adopting it.

I’ve used this anecdote before, but for me, I never really remember living without television. I just always had it. My parents, they saw television happen, and it took a long time for that to become their new normal. Now, talking to devices, that took me a while for it to be normal. But my kids have never really lived in a house that they have not spoken to. And when they go to a house that they can’t speak to it, they feel like the house has a bug. It’s not working right. And so, that to me, is when you know that you’re seeing advancements in a different computing paradigm. And so, my PC’s going to be around, the phone’s going to be around, but so too is Echo and Alexa and ambient computing, for sure. I’m very convicted on that.

Very good. Well, Dave Limp, thank you so much for your time. It is a pleasure to talk to you.

Good to see you.

Decoder with Nilay Patel /

A podcast from The Verge about big ideas and other problems.

Subcribe now!