Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

{audio:0:02:25} Just as a quick intro, I'm a long-only investor, very long-term focused, managing money for a handful of families, long-term money focused on high quality companies. We're very interested in the streaming business, Netflix in particular, and we are also interested in the Cloud business, AWS, that is Microsoft, Google. You are a perfect interview partner, because you've been at the intersection of both basically. Take me back to your time at Netflix. What did you do there? What were your responsibilities? How did the day-to-day look like?

{audio:0:03:53} I ran three groups, the performance group, the operating systems group, and the demand engineering group, which you can think of as capacity planning, although we thought about it differently than that. In performance, our job was everything that happened when you went to Netflix.com. When you press play, we tried to make that go faster. We were essentially an in-house consulting group that work with the different microservices to help them tune their applications. 

{audio:0:04:26} In the operating systems group, we ran Ubuntu and so I had a small team of people who were responsible for making sure that that continued to work the way we needed it, since that was the basis for pretty much everything that we did inside the cloud. Then on demand engineering side, we were responsible for making sure that all of the microservices had the capacity they needed, when and where they were needed, for this to run as reliably as possible. That involved basically moving capacity around between the three regions that we were in, working with Amazon on order-to-buy capacity, planning, capacity and so on. Also working with the services to understand what their capacity needs were in their growth, and essentially taking all of those individual pieces and managing them as a whole. Again, there was an efficiency expense component to that. We obviously didn't want to spend a ton of money in the cloud, but we wanted to spend a sufficient amount.

{audio:0:06:00} How was that time at Netflix? What are your impressions? 

{audio:0:06:06} It was great. It was one of the best companies I worked for. It and Sun Microsystems are in the top two. There were incredibly smart people there; we did really fun things. There were really interesting problems to solve, most of which moved the industry forward in different ways. Both of those things are true at both Sun and Netflix. Sun basically invented this whole industry around engineering-based computing and, at Netflix, I watched the streaming industry grow, and it was a good place to be.

{audio:0:07:00} May I ask you why you left? 

{audio:0:07:08} You retired. And if you didn't, you would still want to work there. How has Netflix changed? If you compare 2016 to 2021?

{audio:0:07:29} It got bigger. A year or so after I started, we celebrated our 100 million subscriber number, and now the number is a 200 million or something like that. It's just explosive growth. The pandemic certainly helped with that a lot, so it grew a ton. Engineering probably doubled in size. I'd have to think deeply on the numbers, but everything got bigger. 

{audio:0:12:03} Let's go into more of the technical side, which is where you are coming from. How important is technology or the technology side of Netflix's business today? I ask this question because certainly Netflix was the innovator and created a category and all the innovations behind it. But how important is that edge today, versus Disney+ or others?

{audio:0:12:53} I'll talk about three different things. There's recommendations, which is the recommendation engine, effectively. There's actually underlying streaming infrastructure, and how that plays into things, and there's content creation. I don't really have a whole lot to say about content creation. I'll talk about content creation from a technological standpoint. I actually do have some thoughts on that.

{audio:0:13:41} Let's go through those three. How important are they today, versus in the past?

{audio:0:13:48} Recommendations will continue to be extremely important. The thing that you constantly heard about Netflix is, I found this show that got recommended to me by Netflix. It's amazing. The question is, how do you continue to do that in a world where there's more and more content? To a certain extent, the thing that Netflix was really good at – and this is not a secret or anything – but think of the number of people who say, I saw this amazing show, and when asked where you saw it, they say, I'm pretty sure it was on Netflix. Then you discover it was an Amazon Prime show. In some sense, Netflix was very good at convincing people that whenever they saw a really good show, it was a Netflix show, no matter where it came from. That's both a blessing and a curse. There's very strong brand affinity there, that people associate equality with Netflix. But there's also this brand muddiness and people are really confused about who they're getting content from.

{audio:0:15:04} Going back to the technology or the recommendation side, to me there's this interesting question of, how do you continue to recommend the best content to people in a way that builds strong brand awareness? I don't have any opinions about how you do that other than there are other parts of the company that worried about that. I think that you can see that Netflix has been probing at this space. They did this 'too dumb' thing where they were trying to build this conversation around Netflix content. Looks like it didn't work out as well as they hoped it would. But you can see them basically experimenting with this, or how do we build this stronger brand awareness? To the extent that they succeeded that, I think that that will be very good for them. In fact, I think that as you look forward, that would be a space to be paying attention to. 

{audio:0:16:06} I want to compare it to Disney+ or newer streamers. Should we do that in the end? Or should we do it for each technological part? 

{audio:0:16:17} Let's do it for each technology part.  Disney+ is interesting, because they already had this huge brand awareness. In comparison to Netflix, they have multiple channels for their brand. For instance, if you go to a theme park, and you see a Star Wars ride, you can now connect that back to the streaming brand. Whereas with Netflix, if you haven't seen it on the screen, then it's hard to encounter it outside of that. 

{audio:0:17:03} How do the others compete with Netflix? Is the Netflix recommendation engine superior to the other streamers? Or is it something the others have to catch up with, including Disney+?

{audio:0:17:13} I believe it's something the others need to catch up with. I do not spend a lot of time watching Disney+, so I don't know if I could speak about that knowledgeably. My sense with Disney a year and a half ago or whenever they launched, and actually to this day, is that they have a core set of properties, mostly in Star Wars and in animation. Outside of that, they're not really strong. Again, that's my personal perception. But the thing is that the Star Wars brand is incredibly strong. The animation brand is incredibly strong.

{audio:0:18:10} So part one is the recommendation engine. Netflix is strong, and the others have to catch up, or are not there yet, and it's important. Second is the underlying streaming infrastructure, what does that mean exactly?

{audio:0:18:31} All of your streams eventually come out through a content delivery network, and that content delivery network is some part of the cost of your business. Then the question is, how much does that cost? There's two parts of this. There's the cost, and Netflix spent a bunch of time reducing that cost, building our own hardware, running our own backbones and so on. Versus, for instance, going to a third party CDN, which is I think what Disney does, and I don't know what Paramount does, and HBO may also be a third party. 

{audio:0:19:13} Who is that third party, do you know?

{audio:0:19:14} I do not know. For Netflix, it's in-house. The higher bit there is, somewhere you need to pay margin dollars. Whether you pay them to yourself or to somebody else? If you pay them to yourself, you don't have to pay them.

{audio:0:19:39} How important is it? And how does Netflix compete with the others? 

{audio:0:19:51} Let me answer those two questions. The answer to the first question is, I don't really know, because that's a financial engineering question. Other than I know that we spent a lot of time trying to reduce that cost and that was at a high level, at the read level goal. Is this cost going down at this rate? Because as you grow, you continue to add more CDN nodes, and if you're really focused on the cost per bit, then you basically take that off the bottom line. It's just cost. From your perspective, looking at the financials of a company, what percentage of their expenses is this thing? Are they reducing it? Is this a knob they could turn if they needed to? 

{audio:0:20:47} I'll flip around to the other side of the cost of it, which is customer experience. We spent a lot of time trying to make streaming really reliable, such that when you press play, you would be able to watch all the way to the end of the thing that you want to watch. There's a part of me that thinks we've made it too reliable, because I've been at this running battle with my son to cut off his access. I have my routers and say, top sending bits to our Roku. But it will run for another 30 minutes or more. And it's like, you must be buffering all this stuff when I'm not looking. It got me to think about the fact that actually makes a lot of sense. If you want to keep streaming no matter what, you need to be buffering stuff up. There's a lot of stuff that Netflix did on the engineering side around reliability of streams. Again, anecdotally, the sense I get is that we have a very high reliability, relative to some of the other streaming services. Then the question becomes, is that a customer retention thing? Or is it a customer happiness thing? I thought that Netflix was always very good at that. 

{audio:0:22:21} The CDN? I guess the cloud infrastructure is in there too?

{audio:0:22:29} Yes, I would include the cloud infrastructure. I don't know how important this is to the outside world. Netflix really embraced the cloud and microservices, and as a side effect, that allowed us to do a bunch of fairly clever things in terms of responding to demand. We added 23 million subs in a quarter. Yes, it was tense, the infrastructure was creaky, but it did not fail. Part of that was the architecture of how we looked at growth, how we dealt with microservices. One of the questions that comes up is, what happens when a big title launches? The answer is basically, not much, and that's because the infrastructure is really resilient. Whereas with some other providers, you do see these things where it's like, we launched the new Mandalorian and people couldn't stream because we were overwhelmed. Eventually, they manage to patch it up.

{audio:0:24:32} So you think what you hear in the media is basically this entire technological stuff like, this item is commoditized, everyone can stream today. You would disagree with this?

{audio:0:24:57} Everyone can stream today. That's like saying everyone can drive a car today, but if you drive 60-year-old car, you have a different level of performance than someone who's driving something brand new. I think that there's that same gradation. Going back to UI, as you go through these different services, some of their UI is just awful. I've had people say that, I go to Netflix, it's really easy to find that thing I was watching or to go back to where I was. I think it was HBO Max or something where I had to keep restarting the show from the beginning for a while because it wouldn't remember where I was when I paused it. So yes, there's different things.

{audio:0:25:56} The last one, the content creation, or the technical part of that? 

{audio:0:26:11} I was not on the content side, even on the engineering side, but I do know that there was a lot of work that was being done to standardize how we thought about content creation. The way to think about this is that a movie, or a TV show, is basically a small startup that forms for the length of the production and then disappears. As a side effect, you can imagine starting a company where it's like, we need to buy all of our hardware and all of our software and then in three years, we're going to throw it all away. The thing that Netflix looked at is, why don't we just buy all this hardware and software once and then reuse it from production to production? When you think about the world that way, you now start to see optimizations you could make around how you move data around, how you pick software. There's a lot about the film industry that was very superstition driven. I used this piece of software on my last picture, and it worked great, so I'm going use it again. On the technology side, there's this question of, can you get this economy of scale through producing all of these things through, essentially, a single studio? I believe the answer is yes. Indications, as I was leaving, for the answer to that was, yes. But it's a journey, because you're up against a set of long-lived beliefs in the industry. 

{audio:0:27:50} The commoditization we discussed, having a car doesn't mean you're going to be a good driver or have a performance car. Why does Netflix win? You named some competitors; why is Netflix the one that will continue to win? What is your view on that?

{audio:0:28:26} Why it's won is, because it was really good at staying focused on the single goal of presenting you with content that you would love when you were looking for something to watch. I don't know if you keep stuff in your queue of things to watch, but you'll find, for a lot of people, they have 30 things in this list of things I want to watch. Every few days, they turn on Netflix and go, here's an interesting show; I should save this for later. I have shows going back three years. I still need to watch The Crown. It's been in my queue forever. 

{audio:0:29:14} The other thing is that Netflix, for better or worse, is in this commoditized brand space which is, when you say streaming, people think Netflix. It seems the extent that Netflix can leverage that connection will continue to be a good thing. Whereas when you think Disney+, you think Star Wars for the most part. 

{audio:0:29:57} You're saying, one will be the content that they churn out or the property that they're able to create, and the other one will be how good are they at recommending stuff?

{audio:0:30:12} Yes. The third thing about Netflix that I think they are trying to get better at is figuring out how to leverage the international reach of their content. There are shows like Lupin that came out of France, that was this huge hit, and you compare that to, again, I'll pick Disney+, just because I'm thinking about them, but almost all of their big shows are, essentially, US based properties. The thing that is a question with Netflix is, do you believe the world is becoming more international, and I think that they're really well positioned for this. Or do you believe that the world is still very US-centric in terms of entertainment? I think that it's a challenge for Netflix to figure out how to make this work. Meaning it's challenging in the sense of 190 countries in the world all producing content and how do you navigate that? 

{audio:0:32:06} Going more to the technical side, Netflix has been one of the first big AWS customers, and I don't know how familiar you are with that decision, but do you know why they picked AWS over Microsoft or over Google?

{audio:0:32:26} Not really, other than I think AWS was the only game in town, really, because this decision was more than 10 years ago. I don't know, I wasn't around. I heard anecdotally internally, the sense I got it was because that was the only choice that made sense at the time.

{audio:0:32:59} With the business of metrics, the cloud is the only solution. It's extremely complex and expensive having everything on premise, in your own data centers.

{audio:0:33:34} It's a tradeoff of where you want to spend your engineering dollars. For on prem, it's like, I now have a bunch of people who are just hardware, I have a bunch of CIS admins. Could I now somehow take those dollars and get some portion of them back to go spend on engineers? The thing that that we always looked at as the important thing was innovation. The faster we can innovate, the faster we can go. If you look at an on-prem data center, let's say that you're an engineer, and you come to me and go, I have this idea for a new search engine, and I need 500 servers to test it. I'm like, I can order those servers, they'll be here in six months. Versus if you're in the cloud, and you can say, you can have, virtual machines tomorrow. You can prototype your idea. If it works, we can have 100 by the end of the week, so you can ramp these things up really quickly. 

{audio:0:34:35} I think, from the Netflix perspective, it wasn't a question of you can innovate much faster, because you could apply capacity where you needed it very quickly, versus having to go, we need to find rack space, need to find power, need to do all these other things. The other part about it, is that people do not think about it in terms of where they are on the growth curve for their data centers. If you have a data center with 100 machines in it, and you need to add 100 machines, that's a big deal. If you're Google and you have a million machines, and you need to add 100 machines. That's nothing. People say, if you were Google, this would be really obvious. But we weren't Google when we started.

{audio:0:35:23} One question I always ask myself as I research the hyperscalers is, how difficult is it for enterprise customers to switch from one cloud to another? How easy would it be for Netflix to move off AWS to Azure?

{audio:0:35:52} It's not impossible because nothing's impossible. It is hard, because all the cloud vendors have a bunch of things that wind up binding you to them in ways that are not obvious. For instance, here's our format for billing and that billing data comes to you once a month, and then you do something with it. Now we're going move to Azure; here's our format for billing. So now you have to change all of your back-end systems to deal with this new billing format. That's a simple one, but an obvious one. But it's also things like, here's what the interfaces are for launching new instances. We use GRPC; we use this other thing. So now you have to change all this infrastructure. A lot of it is less about the applications, because the way it comes to you is, we run containers, just bring your containers over here. But we still administer those containers; how do we go do that? While I think that this is an interesting threat that people bring up, we could move but the reality is, it's fairly arduous. 

{audio:0:37:35} Netflix is clearly an enterprise client by now, and this is deeply ingrained. The other thing is, I was kind of surprised that Netflix picked Microsoft for ads.

{audio:0:37:54} I wasn't around when that happened, so I don't know what drove it. Let's separate out business from architecture. I will pose this as a strength for Netflix. Your question is, basically, you have a business relationship with Amazon, so you should just do all these things through Amazon. I always flipped that around to say, you have an architecture that has really strong interface constraints and all you care about is an interface that allows you to go execute this thing, and whoever does the back end of it is unimportant to you? You can shop around for the person who's the most efficient, or the most cost benefit or whatever, I'm assuming, as part of this advertising discussion, and they went, here's the services we need. Who can provide those services? Can we do it in a way that's mostly independent of how it gets implemented so that we can go with the thing that's the best for us?

{audio:0:39:09} Linked to the infrastructure, what other tools have you been using in Netflix on the infrastructure side? Have you worked with Databricks or with Snowflake?

{audio:0:39:29} No. A lot of our infrastructure was home grown, and the reason for that was just the scale. Generally, what you'll find at most of these huge companies, certainly Google – Microsoft, I don't know about – is that almost all the services are in-house because their scale is much larger than a lot of these vendors can deal with and they have other constraints. Either they don't want to share with the vendors or they are off the vendors roadmap in ways that don't make sense. And that's the same for Netflix.

{audio:0:40:25} What tools do you use at Netflix? You wouldn't use something like a Datadog for monitoring and observability and stuff like that?

{audio:0:40:33} No, we have our own internal monitoring system that we built over time. 

{audio:0:40:48} I'm sure we did, although that was not on the engineering side.

{audio:0:40:51} There was nothing notable you could share that you were using in your engineering?

{audio:0:41:18} Again, almost all of the stuff we used were in-house tools that we built. A lot of them we open-sourced, so other people can use them. For instance, Chaos Monkey was a thing that we had open sourced, that allows you to do testing, inside the cloud, of new services and in various ways. Spinnaker was our system for doing deployment, and that we also open sourced. Part of it is again, this kind of entrapment you see with all these technologies. We built Spinnaker, it was really embedded in everything we did. So when people come along and say, here's another way to deploy software, it's like, but now we have to change all these other things. By the way, we can tune this exactly the way we want it to behave. Ultimately, the challenge is, how do you integrate these new things and how do you divert resources to integrate the new thing versus just keeping the old thing going? 

{audio:0:42:35} Out of interest, because I'm also soon in the cloud, do you know how much Netflix spends ballpark on the cloud? On AWS?

{audio:0:42:48} Yes, I do, but I can't speak about that. That was always proprietary.

{audio:0:43:20} I was just thinking a big chunk of the costs of earnings is the content, obviously, and the other one is tech infrastructure.

{audio:0:43:38} I think the financials, external financials or content spend is $14 billion or something like that. 

{audio:0:43:48} This year, it was $17 billion. 

{audio:0:43:50} I think they break out the engineering, which is way less than that. It may be $2 billion, $3 billion or something like that; somewhere in there is the Amazon spin. You can look at the number of engineers they have and do an approximation there. I don't know how they break out their financials.

{audio:0:44:09} The way I understood it, the AWS or the cloud spend that you have for the engineering team to test stuff, to research, to develop new things is the engineering part. But the part that is in the delivery of the products is in the cost of revenues. 

{audio:0:44:39} Now we're beyond my understanding of how the numbers are put together. What I will say is that, as part of the demand engineering team, we've spent a lot of time looking at the cloud spend, and at ways to reduce it. We had a set of targets, which was that the cloud spend had to grow slower than revenue. We were changing it as I was leaving. I want to say it had to grow slower than the rate at which we added subscribers. So basically, we had this goal to keep cloud spend growth below these other things, and we were having a discussion about whether or not we should try to make it grow slower than revenue, or slower than subs. Those were the two basic choices. Obviously, if your cloud spend is growing faster than revenue, that's bad. If it's growing faster than your subscribers are growing, that's probably bad. 

{audio:0:45:57} The question is, which of those two things you really care about? Because you can control revenue by changing pricing. We went back and forth between, we could do percent of revenue. The problem is, it's not clear that the numbers are really reflective of what's going on. We could do percent of the subs. And that actually made much more sense to us on engineering side, because we could see the number of subscribers and we knew that each subscriber required some amount of additional engineering cloud to support them. Having a goal like that forces you to be constantly looking at efficiency and ways to save. It's fair to say that in the all the time I was there, we always met or exceeded our goal.

{audio:0:47:01} There were three ways. Some of it was pricing, so we used long-term contracts, I think it was three years. Don't quote me on that. We had contracts with Amazon, and as we renewed them, right pricing would lead to price negotiation. New instance types, as they came online, they were more efficient, they were cheaper. Part of what the demand engineering team did was really manage that rollover of the fleet into cheaper instances. 

{audio:0:47:45} Thing number two, is my performance team. Actually, there's a blog post from four or five years ago where one of my engineers saved $1 million dollars by spending a couple of days with the product teams saying, if you make these few changes to your code here, your thing gets much more efficient. We would see stuff like that, where we would go swoop into some big service, find some efficiency and basically produce these seven-figure savings, which of course, just multiply over time because once you've saved it, the savings grow over time. 

{audio:0:48:25} Then the third thing we did was instance type selection. Inside Amazon, you have multiple different instance types that you can use. Some are memory intensive, some are compute intensive. There's a blend, and what we would do is work with different service owners to say, here's the instance type that makes the most sense for your service and that could turn it also into a big saving. For instance, an engineer may start by saying, I want the most compute intensive instance to prototype on. But then once you get into production, you don't actually need all those CP cycles; you could move to this instance type, which is half the price. What we would see is, they would prototype and do alpha testing, and you see this big growth in their cost, and then we come in and go, why did you switch instance types? 

{audio:0:49:28} We did a lot of tracking of cost. I saw a team that met weekly, bi weekly, that essentially looked at spend across the entire fleet. Looking for things such as, these people are growing faster than we expected. We should go talk to them.

{audio:0:49:46} Was that always tied back to subscriber growth, or not?

{audio:0:49:52} I'll say yes. We looked at it relative to subscriber growth.

{audio:0:50:01} Maybe another dimension, talent. What matters in your area is that you have the best people the right engineers and when you have the right people, there is more than a 10x difference. Netflix is famous for trying to pick those high performers; fewer people, but the best ones. Is that narrative true? How can I at least compete with Microsoft, Google, Amazon, Facebook in acquiring and retaining these people? 

{audio:0:50:44} Yes. It's just hard. You do it in two ways. First, really sharp people like to work with other really sharp people and, as you observed, we had this reputation of hiring these really good people. Actually, someone who worked in my team was an industry expert. He's one of these people where you'd go to conferences with him, and people would stop him for selfies and autographs and stuff, which was a little bit weird. I would interview people for the group, and they go, I want to come here, because I heard this person works here. There were two or three people in my team who were like, I'm here because this person is here, and I know if I'm near this person, that will be good for me. A large part of it is just being around people who seem to be setting the standard for a lot of things. 

{audio:0:52:01} Then the second thing was, we had these really interesting problems. I'll use my performance team. You could argue both sides of this. I had eight engineers; the equivalent team at Facebook was 30 people. That was the sense that we got, and the thing is that when you have groups that size, you work on this one little narrow thing, versus working on the broader problem. I think that people were attracted to the fact that they could actually work on significant broad problems.

{audio:0:52:53} Do you think this could change, now that the narrative around Netflix has changed?

{audio:0:53:00} There's a lot that's changing about Netflix. It's getting bigger. As any company gets bigger, they encounter growth pains. Netflix is encountering a set of growth pains. Certainly, before I started, there was a sense that you could get all of the significant parts of engineering in one big auditorium. By the time I left, that was impossible. When I started, we had these quarterly directors meetings where all the directors in the company would fly to one place, and we'd meet for three days. By the time I left, they were not quarterly anymore and they were not worldwide, because there were too many directors. The challenge is that as you get bigger, you start to lose those connections. At the beginning you can say, I see these people once a quarter, I have relationships with them. Now, it's like, I never see people from Europe because they do their own thing. The company restructured in how it thought about producing content, became more regional. All of those changes have these effects. But that being said, having been at Sun Microsystems for 25 years, all of this was very familiar to me. Well, here's how companies grow. I think the question for you is, how will Netflix navigate this inflection point, becoming the mid-size engineering company? I don't have answer to that.

{audio:0:54:57} I'd like to focus on the CDN, because you're a technical person. Could you explain to me what exactly the CDN is? For a layperson, what exactly is a CDN? Why does Netflix need it? I've read about it, but it would just be great to hear from you. Why does Netflix need it, and why is it so much better than the others?

{audio:0:55:22} CDN is short for content delivery network. It's used to get the content as close as possible to the end user. Basically, you build a network, and you distribute your content across that network, such that when you're in Singapore and you say, I want to watch something, there is a box in Singapore that has that content.

{audio:0:55:48} That's physical hardware, right?

{audio:0:55:51} Yes, hardware. So actually, the simplest description is big SSD, CPU in the middle, huge network pipe, and take bits off that SSD and send them out that network pipe as fast as possible.

{audio:0:56:07} Netflix was upgrading that itself, so you basically have thousands of mini data centers in cities where you'd have your customers, so you would rent those premises and you would manage those things?

{audio:0:56:27} We would manage the rental thing is and put rent in quotes. Because what you're doing is actually providing a service to the ISPs, because they don't now have to have a ton of data stream through their networks, because it's all already local. 

{audio:0:56:46} Who are the ISPs in the US, for instance?

{audio:0:56:50} Comcast, Verizon. They will give you space in their data centers, in exchange for reducing their bandwidth needs, so there's a benefit.

{audio:0:57:08} So there are there CDN's like Akamai or Cloudflare; I guess companies do need to reduce the latency. I guess to some extent, safety, as well. You said one reason is to be more efficient. One is being more efficient because their margins don't go away. But what are the reasons why Netflix did it themselves? 

{audio:0:57:50} Since we control the hardware, then we can now make hardware optimizations. For instance, we can go directly to Intel or AMD and say, here's what we need in a network card. And they may come back to us and go, yes, we can do that. Now it's turned out the CDN, people have their own operating systems group. They spend all their time tweaking the operating system for this stuff.

{audio:0:58:19} Why did Netflix do that with CDN, but not with cloud? Because you could argue the same with cloud; hat is a smaller expense, it's just less important. 

{audio:0:58:33} That, I don't know the answer to. It happened a long time ago. To be fair, I think the CDN thing was really obvious from the beginning. The cloud thing was not. Initially, they did have their own data centers, because they had their own data centers for the DVD business.

{audio:0:59:00} But the CDN was here before?

{audio:0:59:02} Because you start the streaming service, and you need something to stream off of. But again, this is all many years before I was there, so I don't know the exact timeline.

{audio:0:59:14} But could you rent that CDN to others? Or is that just for Netflix?

{audio:0:59:21} I suppose you could rent it. I don't know why they would ever do that, because it's all just for Netflix and it’s a value add to Netflix.

{audio:0:59:33} Is there something I should be asking you that I haven't?

{audio:0:59:46} I feel like we covered a bunch of stuff. Let's go back. You basically work for a bunch of family offices. You're trying to figure out what the future holds for Netflix next year. There's probably nothing you should ask me on the engineering side. There's a certain sort of content questions, that if you found someone to talk to them about might be interesting. I think that the big question I see that keeps coming up in the press on the content side is, can Netflix create something equivalent to the Marvel Universe, where you just have a set of these franchises that could go on for decades, as we've seen with Disney?

{audio:1:00:37} That's what I meant with "hit driven". The first time I came across Netflix, I didn't think that Netflix needed this because I saw it's just a machine that pumps out content and also has third party content. That's kind of switched or changed over the years, and it's more relying on its own content generation and as a result, it becomes more important what they create.

{audio:1:01:05} Which is why the content budget is $17 billion.

{audio:1:01:09} It is a more hit driven business than I initially thought. 

{audio:1:01:17} That's the hard part, and it's a more global business, too. I think that the challenge is, in the US, at least people continue to think of it as a US-based company. But it's really a global company. But because subs in the US have been flattening for a while, and even though Netflix keep saying, we're growing internationally, people keep going, but the US isn't growing.

{audio:1:01:49} Do you think that the ads make sense? Technologically, is that difficult to do?

{audio:1:01:56} I think it's hard to do well. Although if anyone could do it well, I think Netflix can. When I say that, have you watched any streaming services that are ad driven? You should try it. It's awful. And the problem is that the ads are inserted by algorithm, so you could be in the middle of some really important pivotal scene, and then suddenly, it will stop and you'll get three minutes of ads. 

{audio:1:02:30} That's a good point. Basically, you still think the best engineers in that domain are with Netflix, and if someone would be able to create a good product around it, it will be Netflix?

Netflix Engineering & AWS Relationship

Former Engineering Leader at Netflix

inpractise.com/articles/netflix-engineering-and-aws-relationship

Executive Bio

Former Engineering Leader at Netflix

Interview Transcript

With the business of metrics, the cloud is the only solution. It's extremely complex and expensive having everything on premise, in your own data centers.

Related Content

Amazon Advertising: Endemic & Non-Endemic Ads

Former Head of Sales at Amazon Advertising

Netflix Business Model & Economics

Former Director FP&A at Netflix & Former Head of Strategy & Finance at Opendoor

Amazon Retail: 1P vs 3P Economics

Former General Manager at Amazon

Lessons from Eight Years at Amazon, Inc.

Former CFO at Deliveroo & Finance Director, EMEA, Amazon

Partner Interview