Fastly, Inc. (FSLY) Earnings Call Transcript & Summary

All right. Great. I think we're getting ready to go here. So thank you very much for everybody being here. My name is Frank Louthan. I'm the senior wireline analyst here at Raymond James covering CDNs, data centers, telco, everything in that genre. Really pleased to have Fastly join us here for the -- for our conference. And so we're going to jump in a little bit. Maybe turn this over to either Josh or Ron or whoever wants to jump in.

But let's talk a little bit about -- set the stage a bit, talk to us a little bit about Fastly. Where do you fit in, in the CDN world with edge and security? And maybe a little bit about how you're different from your peers?

Sure. Frank, it's nice to be here. We appreciate it. I think, for us, we look at the world with -- in 3 components: we've got delivery, we have compute, we have security. And ultimately, they all form this virtuous cycle of engagement. So if you look at an example, like an e-customer -- an e-commerce example, like you have a customer whose fundamental -- fundamentally, they understand that performance matters. The faster their website gets delivered to a customer, the more you buy, the more you put in your cart, the more you -- the happier you are, the more you come back. They also know that they need to deliver their product and make sure that it works across all regions of the world and at scale. Because in these holiday shopping season that we just are coming out of, they need to make sure that no matter what that scale is, you get that fast experience. So when customers migrate to Fastly, they get great performance, they get effortless scale, and then they discover the ability to bring application logic to the edge. And they graduate -- as they graduate through this process, they bring application logic to the edge, and they build entire applications focused around the edge, which means, again, virtuous cycle. It gets faster if scale is better. And then we move into the security side. They discover along the way that security can also be effortless, that security can be one of these things, not this encumbered legacy product hardware world. They can have security that works and that is combined into that ecosystem. And as they build more on us, they discover more places they can use it, and you start getting into this virtuous cycle. One of the things that we really differentiate on at Fastly is customer satisfaction. If you look at our NPS scores, if you look at our customer sat and, probably most important, if you look at our revenue churn, you see that we are leading. And when customers, the names of which we have, when they come to you, they come with high expectations. When they stay on you and they continue to build, that is a real differentiator. So we begin our journey with the premise right now that the Internet is a different place than when the legacy providers started in this industry. And the success of the major cloud vendors has really accelerated this to some extent. So software-defined, we have this unified service. All of our software runs on all of our servers for delivery, security and compute. And we shed all these constraints of expensive routers, custom hardware. Really, we're a software company. I mean, that's what we do. So if you look at the analyst community, Gartner, Forrester, they are really recognizing us for the incredible work. And our differentiation is about speed, it's about scale, it's about programmability, it's about giving people control. Those are the reasons that people come to us and stay with us.

All right. Great. Right, so let's walk through -- you mentioned on the call, you set out a growth path there. What do you need to kind of reach that goal that you laid out?

Sure. So we had 3 main areas that we targeted when we were developing our 2025 goals, which are to not only bring more requests to our new compute platform, but bring more enterprise developers on to the platform to drive those requests and ultimately drive revenues. We've made the bet that developers would be the new decision-makers when it comes to how applications are delivered and secured because they're the ones driving the transformation. So our goal is to make their jobs easier and more exciting by actually giving them full visibility control. And as Joshua talked about early, security built in that doesn't slow them down, but actually speeds them up. And so one of the things that we're doing to that goal is the commitment to reduce the friction of adoption by giving them a sandbox to innovate and free of charge. So we launched a fully functional free trial version of Compute@Edge, which we've announced. With innovation and development, we think the requests will come through. I think the next area really surrounds the security business. We have a meaningful security business today, but believe strongly that it will continue to grow at scale in the future and reach 10x what it is today by 2024. I think this is very much based on the evolution we're seeing in the market. Security is becoming more and more important, but even more importantly, part of everyday conversation. You can't have delivery or compute without security. So it's about having security as part of compute and as part of delivery built in. And then I think lastly, if you take a look kind of at a macro level, the $1 billion revenue was based on our history of this 30% -- of the 30% CAGR of growth that we set out during the IPO. And while we're exiting a tough comparison in 2021 due to the elevated growth we saw in 2020 due to the pandemic, we're now more focused than ever on executing this goal through innovation and setting the right people in place. It's the transformation in the sales organization that Brett Shirk has put in place. We recently named Lakshmi Sharma as our new Chief Product Officer, who comes from the Google Cloud organization and comes in with decades of experience building dynamic product organizations.

So how much reaching that goal is organic versus inorganic?

So essentially, the -- we can achieve this goal organically. I think there's potential for a few acqui-hires built in there. But essentially, it's -- we can achieve this, we believe, organically. Now if we saw an opportunity for a solution that was a fit like Signal Sciences, we're very open to doing M&A that we believe can successfully accelerate our business.

Okay. Great. So let's talk a little bit about -- you mentioned getting the growth there and the requests for a second up and so forth. Let's talk a little bit about the traffic. So we raised our rating last week looking at some of our proprietary work and looking at what -- your traffic and so forth. It implied to us that the trends are improving. It looks like you -- and according to our model, you may be able to even exceed the high end of the range of the guidance. Talk to us about how things are tracking in the quarter from what you're seeing and what would have to happen here to beat your range of guidance of revenue for the quarter?

Yes. So I think, first and foremost, I think we're happy with how the quarter is progressing. I think we really kind of appreciate the work you've done here. I think at this point in time, we stand by our existing revenue guidance. As we've talked about, today, a big part of our business is usage-based, and there's always going to be some inherent volatility. And so our guidance will continue to reflect that inherent volatility in our business and reflecting the outlook that we have high confidence in.

All right. I'm trying to get [ Vernon off follow ] an 8-K here, but you're not biting. Okay. All right. So looking out over the next 12 to 18 months, you do expect some customer concentration to smooth out a bit. What could kind of help with that to drive some more consistent results as you look out 12, 18 months or so?

Yes. No, we do expect it to smooth out. We've talked about some variability inherent in the business today, but we do see that being reducing over time. We have some customer concentration, which is really kind of that driver of that variability. And I think there's 3 main sort of drivers or business dynamics that we're seeing in our business today that are going to reduce that over time. First, as we accelerate product innovation and with the transformation that Brett is doing on the sales team, we're growing our enterprise customers. In Q3, we had the largest organic increase in new enterprise customers since our IPO. And as we increase the number of enterprise customers, we reduce our dependence on individual customers. Second, we're increasing our focus on the SMB market through leveraging our partner market. This further reduces the reliance on large customers. And then thirdly, through our SaaS-based security revenue, and as these recurring revenue streams grow and grow at a faster rate than maybe core, this is going to drive a further reduction in our volatility. So taken together, we see our volatility or variability decreasing over time.

Right. Great. Maybe talk to us a little bit about pricing over the next 12 months. How should we think about that?

Yes. I think what I would look to as a guide is really kind of what we've seen over the past year in terms of pricing. We actually saw a deceleration of pricing pressure in '21. Pricing declines in '21 were actually lower than what we saw in 2019 and 2020. And so I would expect, from a trend perspective, to use that as the best guide in terms of going forward, i.e., not an acceleration in pricing declines, but kind of in line with what we're seeing recently.

Great. All right. I have a question for the audience, gets into another topic that we were going to address on edge, but again an investor question. How is fastly Compute@Edge product differentiated versus telco or tower company edge offerings that they seem to discussing more often. Edge, obviously, a very, very well used term, I mean, there's a lot of different aspects of it. But talk to us about what you mean when you talk about edge and specifically the Compute@Edge product?

Yes, it's a great question. I think there's a lot of both hype in the market and confusion. So when we talk about -- so I guess, start -- let's start. There are many edges. We don't claim to be the only edge. You've got a mobile device, and you've got points all the way back to the core central server. So let's just be clear. There are multiple edges. We think of ours as the developers' edge. So we think of this through the eyes of the customer. If you think of developers who we work with every day, they need a place where -- which is close to users, but a place where they have control over their data and they can ensure that it is secure and that they have control. So think about things like PCI audits or SOX on it. So Fastly is the last place that our developers can put code knowing it's in a safe environment. And they also need, of course, performance and the ability to leverage the density of compute and storage in order to make this work. So think about this as you need a number of ingredients to get the developers' edge to work. To take these complicated applications, move them from a central location to the edge, you need a number of factors to all work together. One of the most important factors actually, given how much of the content across the Internet is secure, is you need to have the TLS keys, right? You have to be able to decrypt this information in order to work on it. And that's what we are trusted with. When you start moving out into the telco locations or the towers, you lose that ability or it's very difficult to have those 3 ingredients, dense compute, dense storage and the certificates in order to unpack that data and do work. So a lot of the conversation right now in the telco space is around doing work with packets where you can't actually open them up and see what's going on. And as we move to a full TLS or a full SSL or solution, HTTPS, what you see the lock in your browser. What you'll see is that actually, we're moving away from the days where raw unencrypted bit, you can actually have a lot of impact on them. So we're seeing that the necessary ingredient to actually have the encryption keys in order to do the work is critical in order to add any high value. And today, most organizations aren't interested in building relationships with the hundreds of telcos around the world, having their keys be at hundreds of locations, being forced to do all of those -- or hundreds of different companies and thousands or tens of thousands of locations and being forced to do all the auditing. So that's why they trust us. So I think what we're seeing is there's 2 camps here. We've got telcos and towers that can move unencrypted bits. And we've got the edges that are going to be able to add value to those bits because we know what they are and we know how to add value. And it's 2 very different camps. And I would say the camps, which actually have the data and have the keys and have the compute horsepower to do it are going to just add a lot more value. So it's not that there isn't value to add on, taking a bite if you don't know what it is and doing something to it. You just can't do a lot to it. And I think that's where the developers' edge is so important in the evolution of edge computing. As I say, it's not the only edge, but it's a very, very important aspect of this conversation. We can talk for hours about this when I think it's a fascinating discussion.

Yes. Well, we don't have hours, but I got -- let me dig in that just a little bit more. So walk us to maybe what's the architecture on the edge. Because I think folks can understand your architecture and where you place your nodes inside the data centers and where the connectivity is. But what's the architecture that you're adding to add these additional edge components? And how does that give you an advantage over your peers who are also claiming some kind of proficiency with edge?

So -- I mean I think there's 2 aspects to that. The first aspect is we work in compute-dense POPs at the major intersections of the Internet, essentially the backbone. And I think one of the things that is misunderstood about the Internet is how it's changed over the last 20 years. It's almost like a highway system in United States, for example, now where you've got all roads lead to Los Angeles, all roads lead to Seattle, all roads lead to New York. And the reality is everything hubs through these major locations. There are 16 to 18 locations in the United States, where almost everything is a hub and spoke through. So let me give you an example. If you're sitting on an AT&T phone in Santa Fe, you are actually going to get taken over AT&T's proprietary network and exited in Dallas. So it doesn't matter. If I've got a POP in Santa Fe, which is what the legacy providers, they've sneezed POPs all over the United States, the reality is I'm not going to hit that POP until I go from the phone to the tower over proprietary network to Dallas, then I'm going to go back over the Internet to Santa Fe. So it doesn't make you any faster anymore. So the reality is what we need is very dense, very powerful locations at these intersections of the Internet. The rest of them are actually irrelevant for the most part in terms of how the Internet works. And this primitive, this prototype of how the Internet works is being adopted in all countries in the world. So we actually have about 100 locations where the Internet transits through and you need to be in those locations. And that strategy of sort of sneezing things all over the Internet is not one that produces performance. So I would say, one, all of our edges are in these dense locations. And all of our edges, all the edge technology we're talking about are, as I said earlier, are all on the same servers doing the same work. So if you look at the architecture, it's density where the Internet highways connect. And that's why we're able to be the most performant solution on the Internet is because of that strategy. Now there's a lot that is then built in software. So unlike legacy providers where they go and buy large expensive routers, everything Fastly does is in software. And our software-defined network, we talked a lot about this in our IPO process in the S-1, is a huge differentiator for us and it's what allows us to be so incredibly CapEx-efficient, given the scale and growth that we have.

Okay. All right. That's great. So how does edge help -- how does that -- how do you see that -- you talked about the intersection of all these different products earlier. How does edge fit in to kind of -- to increase the amount of CDN and security products that you sell? And how is that going to be a driver going forward?

There are 2 aspects to edge, which are important. One, for us, edge is a platform. So we build things on top of it. And these dense locations and all of our software sits on top of it allows us to build products. So for example, the next-generation WAF product that we are bringing to the edge with our acquisition of Signal Sciences, that is a great example of built on our own edge capability. We also have edge or Compute@Edge as a product our customers can buy to build their own things or to take one of the recipes or primitives that we have and build their own solution. So we really have 2 aspects of this. And I think if you look at why it's important, programmability has always been important. Our customer, our leader, our champion, the developers of the world, they're the new decision makers. They're the ones who control budget. They're the ones who design us in at the basic level, right before -- right as you're having, in this imagination, you get built into the framework. So I'd start there. But I think if you look at our Compute@Edge products, the ability for you to go right at our edges, it's really differentiated across 3 main parameters. The first is performance. If I am going to do work on a request that's flowing through the network, I can hold and wait on it and just have it sort of let me -- give me 100 milliseconds, I'll get back to you as I start up. That's unacceptable. So start-up times, huge differentiator for Fastly. We're measured in nanoseconds. Our competitors are measured in milliseconds. We're talking about orders of magnitude difference. So that's one area. The other 2 are equally important. The first is scale. So we don't -- we're not out there saying, "Oh, you can only do 10 requests a second on this." The reality is the technology that we built needs to be able to scale to our current network and well beyond. And so many of the solutions that are being used in the market today inherently have really significant scale limitations. You can only do a couple of hundred requests or a couple of thousand requests. Imagine, when I go to one of our largest customers, and I say, "Hey, you're doing 100,000 requests a second, but you can only do 50 of them on compute." I mean it's a nonstarter. You're not going to start imagining the future leveraging the edge. And the last is security. And one of the challenges with inherently in a solution where you share customers on the same servers, you need to make sure it's safe. And one of the challenges of safety, if you start intermingling requests from different customers together, then a customer who maybe is a bad actor can go and access some of the information that a good actor has. So for us, each of our requests and responses are isolated in its own environment. Many of the solutions on the market today co-mingle those and that can lead to very bad outcomes. You've already seen that in the market -- in the history of our market where we've seen solutions that bleed data across customers. And that's not okay for our enterprise customers. That's a big challenge. So I think if you look at the edge, it's really about performance, it's about that scale question, and it's about the security question. And that's what differentiates our edge from others.

Okay. And you mentioned more of a capital-light model, but what kind of capital investment is required as you build out edge and as customers come on? And is that saving them some CapEx? So how should we think about that?

Yes. So I think Josh already talked about, we have a very efficient sort of capital investment. However, we will discuss the outlook, I think, in more depth on our Q4 earnings call. But in short, these investments are already planned and built into our existing models alongside the organic growth of traffic and revenue. I think it's important to note that our CapEx investments are impacted by the timing of these investments. And as we've discussed, we need to continue to invest in the network ahead of expected traffic. We believe investing in our edge network is important for driving the growth that we're talking about to achieve. And so that's how I would sort of look at that investment level. It's efficient, but we are increasing our investment levels to build out this network.

All right. Great. Folks, if you have any questions out there in the audience, feel free to send them along. They should pop up in my e-mail. With that, we'll just kind of keep going. So talk to us about the sales process for edge. Is this a separate sale process? I mean how is this incorporated and packaged with CDN or security? And you mentioned the developers there. Is that a -- how different is that group of customers that you're trying to win over and champion this for you than who you're usually targeting within the organizations?

Yes, it's a great question. And the edge is continuing to evolve. I think if you look at the Compute@Edge platform, so again, this is the platform that allows you as a developer, as a customer to innovate. What we see is 2 things. Often you're also looking for delivery and security. So there's often this process where you come to this -- as I said, where you come to this after solving a scale or a performance problem. And that's important. So that kind of hovers around the CDN ecosystem. In security, I would say it's the same thing. One of the areas that's really exciting, however, is that you can come at Compute@Edge without looking at a traditional CDN use case. And in fact, CDN has been around for 25 years. And I talked to a lot of CTOs who were like, this is boring, like I don't even want to talk about this. So what's really cool about edge from our perspective is the ability to start a totally new conversation. This isn't a conversation where you've had a vendor for 20 years, you're unhappy with that vendor, you want to switch, you have inertia. This is about greenfield. This is about opportunities that you are trying to solve through other means that are mission-critical. So I would sort of say there is and can be a separate sales process. What is great when we talk to procurement about compute is it's just a different category, and that really helps us. But it also as often is bundled together with our CDN and security discussion. And the sales process is the same as it's always been for Fastly. It starts with getting our product in at the moment of innovation, taking all the enterprise sales bureaucracy and scaffolding away so that at 2:00 a.m., when you are inspired as a developer to build a future of your company, you can do that. And that's critical. Now we also need -- because we are so mission-critical to our customers, we also need a top-down motion. More enterprise sellers are going in to provide the confidence that a company can put their entire destiny, online destiny on us because that's really what customers are doing. So when you look at the types of brands that trust Fastly, they are trusting us deeply, and we need to make sure that the C-level executives also understand that. So it's really the 2 motions that come together, and compute actually follows that. We've released -- in the last quarter, we talked about our playground and free trial account for developers, so they can just start doing that, which is amazing. We talked about the fact that we're also giving up to $1 million of credits to our enterprise customers to play with it and see how it works. And those 2 programs, we're seeing a tremendous amount of interest in there. They're going well. Both of those, I think, are very exciting. And what we have seen, based on our history, have to be in place in order to get the kind of momentum that we need.

So maybe walk us through that promotion a little bit because that's got some investors' attention, particularly the credits. How does that work? Is that just -- does that mean you just play with it until you would have theoretically build $1 million? Or is that a real credit against -- it's going to go against future billing? Or how should we think about it?

Yes. So we really want people to experiment. There are different phases to this. The first phase is, hey, come for 3 months and just try this out. And then it sort of graduates where you then -- we sort of share in that process through the next phase of the engagement. But really what's important to developers is they want to know as they try new things that they're not going to embarrass themselves in front of someone like Ron, the CFO, right? They don't want to have this crazy bill where Ron was like, we didn't budget for that. And so this is a process. As you start looking at innovative technology and new solutions, you have to give people the cushion to experiment. And that's really what we're seeing. That's really what we're trying to do. So it's -- initially, you're good. You can see what it feels like. And then we'll start sharing in that over the process of the next few months.

Okay. That's great. All right. Let's switch gears a little bit to security and talk about that a bit. What specifically do you -- on -- as we talk about -- you talked about the 10x growth on security Again, we talked about this a little bit earlier. But talk to us about organic versus inorganic and internal growth. And lot of investor questions on your ability to -- why you threw out a number like that and your ability to reach that goal. Walk us through what you see about the security part of the business that can really hit that kind of growth rate over the next few years?

Yes, certainly. So I think as we talked about earlier, our growth expectations around security are primarily based on organic growth with a few acqui-hires. As we said, if a solution comes our way that makes sense, like the Signal Sciences acquisition did, we remain very open to M&A opportunities. I think when you look at kind of the drivers of growth and building on a number of the things that Joshua has said, it's been true for a while. We hear from customers that these disparate solutions that they have to deploy to drive security are challenging for them. And whether that's buying them from different companies or whether that's other companies who have brought products together through acquisitions that are not really integrated. And so we look at the centralization of security, bringing that together, making it a core part of delivery as a massive driver for the growth we see in the future. I think -- further, I think we don't in digital world of security or -- excuse me -- of delivery or compute without security being a core piece to that.

Okay. And so what role does sales play in all of that? How should we think about what sort of investments you need on the sales side to really to reach that?

Yes. So I think, one, if you look at the integration of Signal Sciences, I think we've taken the strengths of both the Fastly and Signal Sciences team and we've merged them together. We did that both on the sales front where they've learned from us and we've learned from them, but particularly on the product front. So what you'll see almost a year later now is it's almost inseparable to think about what is Signal Science and what's Fastly? Why we're presenting -- a lot of our metrics are combined because we've brought the business together. I think if you look at -- excuse me. I think if you bring it all together, I think, to this integrated solution, that ties into what customers want. I think as for growth, I mean, as for team member expansion, growing our sales force, I mean, to some extent, that's a little bit of math. As we scale the business, we're going to scale our sales force. We're going to have to grow our sales team. We're going to leverage the learnings to develop that team as we grow the sales team to deliver these targets.

Okay. All right. All right. So how much do you think as far as the growth like -- versus -- expanding Signal Sciences versus today versus others? I mean besides some of the capabilities that they have, how should we think about -- if we look at that mix of business when you hit that 10x growth rate, what do you think is going to be -- is going to make up the bulk of that?

I mean, from our perspective, the Signal Sciences investment was foundational. So I see all of this as expanding upon Signal Sciences or the vast majority of it. So the Signal Sciences product really brought to the market some core differentiators. One of them was ease of use. Another was ease of implementation. Those are the same things people are looking at across all factors of the security ecosystem. It's too complex. You don't have enough foundational visibility and control, again, similar attributes that Fastly brought to the table. And so I think a lot of this is going to come as offshoots of the core investment in Signal Sciences. Obviously, you've got the WAF business, and the next-generation WAF is obviously an important element of this. But we have the bot market. We've got the TLS aspect of this. We've got the denial-of-service protection. All of these are, in many ways, offshoots of the core investment. So I think I would look at it as sort of the fertile ground, our ability to bring even more investment to the table. I think it's going to form a huge part of that.

All right. Great. So your competitors have similar things. When customers are coming to Fastly for security, how are you winning in the marketplace? What are they choosing to use you for over your competitors?

There are a few attributes that really resonate with our customers. The first is the -- actually the quality of protection. What would be amazing to most listeners on this probably is that, in some cases, a majority of web application firewalls are in logging mode for other vendors, which means they tell you after the fact that somebody has broken into your house. You usually know that, but they tell you after the fact because you actually -- in the morning, you go through and you review all these logs. The reason they don't go into blocking mode, which is what you'd expect from an app firewall is because many of them have so many false negatives or false positives that you actually are spending most -- a lot of your time blocking good people, and that's terrible. No sales executive wants to hear that you've sort of barred the doors on your store from all the good people that want to come in. So one of the great attributes of Signal Sciences, one of the reasons that 90% of customers run it in block mode is because it gets that right. And although that seems simple, this is a very hard technological problem, which is how do you make sure that bad folks are blocked and the good people are let through. So I would say, one is just the efficiency, the quality of the protection is just vastly different. I think the other element of this is really this DevOps workflow, right? Engineers love the product, developers love the product because they want to use it, and it gets integrated into their workflow. One of the other strengths that we see is its ability to be deployed in many different forms. So we have the cloud aspect of it. We also have agents that can sit on servers. We've got tools that work with the major clouds. So all of the implementation side and the flexibility there really mirrors the complexity that our enterprise customers face. To walk to everyone and say, hey -- as a start-up, you can say, okay, everything is going to cloud. This is easy. Many of our customers have their own data centers. They've got disparate environments, maybe by country. They need the ability to have very flexible deployments. Those are things that are very different from the competitive set.

Okay. Great. That's great. All right. Folks, you've got a few more minutes if someone has some other questions. Otherwise, maybe let's go back a little bit to pricing. Let's talk a little bit about pricing, say, for the next 12 months. According to kind of our data, we're tracking, it implies that your -- the pricing declines are actually improving a bit, meaning they're declining a little bit less than they were earlier in the year. How should we think about that? And what's your outlook for pricing going forward?

Yes. So I think as we had articulated, I think that, precisely as you noted, we have seen a deceleration of pricing pressure in 2021. Those -- the declines we've seen in 2021 were lower than what we saw in '19 and '20. And so I think our outlook is such that we would expect to continue to see that trend. We don't see an acceleration in the rate of declines from what we've been seeing in '21.

Okay. Great. All right. I've got a question from an investor. It's a good one here. So your revenue growth over the years has been strong, but there has been no leverage in the model. When will we see evidence of this? And can it be a cash flow positive business? So walk us through the operating leverage that you expect to get with all this growth.

Yes. So I think a couple of things. One, as we talked about on the capital side, we're investing to build a global network, an edge network to support compute and the revenue opportunity. We believe there's an opportunity to drive growth with increased investment in the business. And we investing for growth does drive long-term leverage in our business. One of the things to look at is in terms of gross margin. I think 2020 provides a really good example that when you see traffic align with the capacity of our network, you see significant gross margin accretion, which we saw in 2020. What you're seeing this year is that investment that we're making for our edge network that the international expansion is having an impact on revenue. I think secondly, when you look at margins, as you look at Compute@Edge, as you look at security, those are products that also have higher gross margins. So part of this leverage in the business, this improving results over time is growth in security. It's over time, as the capacity aligns with our investments in the network, allows us to drive leverage and ultimately a business with positive free cash flow. But we believe the opportunities for investment in the near term are big.

All right. Great. All right. A couple of other things. So I think it's either on our call back we had with you, maybe you said it on the earnings call, but you've gotten the traffic back after the Q2 outage without really -- without any pricing concessions. It's good to hear. Walk us through kind of how that went? And what confidence you have that you're not going to see any more lingering effects of that outage on your customer base and all that's gotten resolved?

Yes, as you mentioned, we did have an outage, and that was a challenging time for some of our customers. I think the fact that 99%-plus of our customers came back as soon as we came back is a testament to the relationship and the trust. We also had one top 10 customer that needed to go through its own process. Pricing concession doesn't really come into the discussion. When you are a unique product that's highly valued by a customer who also has outages, right, I mean all of our customers also have outages, I think they understand that this is part of the world that we live in. And really their goal is to understand, do we have the processes in place and did we mitigate this properly? And have we mitigated this class of issue in the future. And I think those are the questions that came to mind. So pricing wasn't really a conversation that was being had when you have something that's this mission critical. The reality of why we're so mission-critical comes back to the same premise. If -- performance really matters, right? If you are in the e-commerce sector, you know, as we've talked about earlier, that the faster a site is, the more somebody is going to buy, the more they're going to put in the cart, the more they're going to come back. And that is fundamentally dollars, right? That's really what it means. And so when you're making a choice about more dollars or less dollars, most of our customers want more dollars, and that's why they come to us is because we work with them and, as they gain, we gain. So really this is about deep-seated partnership. It's about differentiated technology. And as for the lingering effects, I think we've come out of those moments stronger, I mean, as a business, because we take them seriously, because we acknowledge to our customers the severity of the issue. Many of the vendors in this space have historically underplayed outages. And you'll hear this all the time where they say, "Well, wasn't that big a deal. It didn't affect customers." If you're a customer and your site is down, that really hurts your credibility, hurts your credibility within the organization. So this downplaying of issues in the past by legacy vendors, I think, has really hurt the trust. We work with developers. Developers are -- have a certain mindset. They care deeply about integrity. They care deeply about honesty and they want transparency. And those are things that are core to our values. And so in a strange way, when we have outages and we acknowledge them and we deal with them in a transparent way, it does increase trust with the types of people that we deal with. And I think that we've come out of it with increased trust from our customers because they know that we take this seriously, and we're willing to stand up and accept responsibility.

I think -- that's great. I think this transitioned a little bit on my last question. We continue to see the rise in self-provisioning. It's always major media players building some -- with some -- building their data center facilities and so forth and their own networks. How do you remain relevant to them? You talked about the customers trusting you. Is that part of it? How do you remain relevant to these guys over time rather than them putting on their own networks?

Sure. So we've talked about this in the past, and this can be -- you can talk about this as private CDN or you can talk about this as managed CDN. The reality is that there are about 50 customers in the world who have the combination of attributes that make it attractive to build some of the network themselves. So you have to, in almost all cases, B2B2C provider. You also have to have something that actually really helps the telcos. So imagine, for example, when you go buy a phone in -- from -- let's say, from Telefónica, they'll say, "Hey, you get free Netflix." So when there's one of those symbiotic relationships with the telco and then you have to have enough global scale and presence, that's about 50 customers in the world. So I think it's important to sort of say what the market is. If you then expand and say, "Okay, well, what are the primitives? How has this been done in the past?" About 10 years ago, you saw 1 or 2 really clear examples of how this was going to be done. You look at Netflix and you look at Apple, and they basically had no choice. They were being told by legacy providers and being held -- basically being held hostage by the legacy provider, and they had no choice but to do it all themselves. In the last 10 years, we actually haven't seen almost any example of anyone doing it all themselves. And that's because this idea of hybrid is really the primitive or what we see everyone doing. So if you're in one of these top 50 customers and you have the scale, you're not going to go build this all yourself, you're going to partner with a company. One of the things we came out with from day 1 was a private or managed CDN offering because we know that the economics work for certain customers, and we also know that they need the scale, the performance, the control, the flexibility that Fastly brings. So if you actually look at why this trend is the way it is, in many ways, it's because of us. We are providing to these vendors this capability. So I would say this is about partnership, and that's really what we've seen. And so this is a trend that we encourage. We think it's great. And ultimately, we're relevant because we help people in this process. But they're not moving only to manage, they're actually moving to a hybrid solution where some of it is their own servers, some of it is not, and that is absolutely the future. So we've seen everyone do in the last 10 years.

Funny. The carriers use the hybrid solution. If they -- maybe if they thought differently about their future, they wouldn't have chased everybody to build it themselves. But all right, folks. So that kind of runs out of time here. Really appreciate you guys being here with us. Hope everybody is having a great conference. We've got a couple more panel discussions, got one in about an hour. We'll be talking about more of the networking side of things and data and infrastructure. I hope everybody has a great week and look forward to hearing good feedback from the conference. So thanks, everybody. We'll talk to you soon.

Thank you.

Take care.