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

Good afternoon, everyone, and thank you for coming to day 1 of the Morgan Stanley TMT conference. I'm Sanjit Singh, part of the Morgan Stanley software research team, and we're especially pleased to have the management team from Fastly. We have Joshua Bixby, CEO of Fastly; and Adriel Lares, Chief Financial Officer. Thank you both for joining us this afternoon.

It's a pleasure to be here. Thank you.

So looking forward to hitting on all the exciting aspects of the Fastly story. First, before we get there, let me go through some quick research disclosures. For important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. And if you have any questions, please feel free to reach out to your Morgan Stanley sales representative. So with that out of the way, I wanted to start just really talking about Fastly and how it relates to edge computing and what that all means, right? And so at a basic construct, right, we have data centers, we have hyperscale cloud computing platforms and we have the devices. All these things are growing at a pretty nice clip. In terms of nailing down what the edge means in the architecture and where that provides the opportunity of Fastly, Joshua, how would you sort of frame that out for us in terms of sort of explaining the opportunity for someone who's coming to the story relatively new?

Yes, it's a great question. I think we're in a time of incredible innovation and change. So it wouldn't be surprising that people who aren't at the sort of edge of this debate don't understand this because we're right in the middle of moments of change, certainly spurred by COVID and the digital transformation. So I think the easiest way to think about this is the edge is in on-ramp due to major clouds and the hybrid environments that we exist in. And when I use the term on-ramp, what I mean is the edge right next to users. But because it's right next to all users, it's also right next the clouds. And so I think a bit more of it's a bridge where there's an on point, there's an offering, and that is in many, many different directions. So what's happening is that as organizations rethink this paradox, and you're absolutely right. I've got a client, a mobile app, let's say, I've got a web server, I've got an application server, I've got a database server, that's been the paradigm through which developers' lenses have always worked. What we are seeing is that the primitives of the most innovative companies in the world are realizing there's another dimension here, and that is this network. This network that spans globally, and it's close to users and it's close to the 3 major [ clients ]. And what that network can provide is a place to do some of the critical compute. And the beauty of doing it close to users, of course, is it improves the performance, it improves the scale, it also includes -- improves the security. So this is about rethinking how the applications are built, and we're at the start of that. It's a very exciting time.

That's a great way to sort of introduce the topic and the conversation. And one of the questions that we get when it comes to modern platforms like you guys is, what was sort of the aha moments? When you look at your founder and he's looking at the CDNs in 2010, 2011, what was the sort of architectural shift that he saw or the technology that he saw that allowed him to build a platform with far less points of presence than your traditional incumbents? Can you sort of pinpoint the architecture shift? I think that would be useful to understand.

Yes. I think there are 2 elements. There's certainly the architectural shift, which is the Internet, over the last 30 years, have evolved. Whereas Internet used to be much more local, it's now much more hub-and-spoke interrelated. So if you are at 20 major connection points in the U.S., you can have, as we show every day, better performance than somebody that has the [ cost ]. And that's just because of the way the highways work on the Internet. So if I'm on a cellphone in Santa Fe, on AT&T, I'm going to get tunneled over the AT&T network to Dallas. So I can have a POP in Santa Fe, I can have a POP in Albuquerque. None of that matters. What matters is that the POP that I have in Dallas is able to deliver all of the content, a long tail content to the Internet and has the compute that's there. The real aha moment for Arthur and the team when they built this was you need to compute next to the data, and you need to be able to control it in real time. So this is really a story of control. Arthur started out building Wikis, and everyone knows Wikipedia and Wikia in this entire world. It might be the same for months. Until something happens, you might have hundreds of users that are coming to make changes. And so how do you create a system where all of that experience can be delivered to the edge? And how do you control it programmatically such that you can make sure people don't see stale content? And the idea of this is actually transcended well beyond the Internet and the traditional way we thought about the Internet. This is about edges all over the world that have the capacity to deliver, secure and scale every byte that goes across the Internet, which is more and more content every day for all of us.

It's a great point. And when I think about your revenue and where Fastly is today and I look at your product portfolio, content delivery, image optimization, a little bit of video and streaming, security services such as WAF and DDOS, load balancing, et cetera. A lot of these are established product markets, right, where there's established budgets. And so the question is, is it about just building a better mouse trap, i.e., minimizing latency for these established services? What's the sort of source of value for this set of capabilities versus some of your competitors?

Yes. So I think we do sit in this incredible richness of multiple billion-dollar markets that are all being turned on their head as we speak. So I think you are right in saying all of these markets are ripe for disruption. A lot of these are appliance markets. And as the data center goes away or as it transforms to hybrid, those appliances have nowhere to live. I think fundamentally, this is, yes, about bringing a product experience and a customer experience that is superior. But it's also about capturing the innovation of developers. We are at a time where developers are defining the future. It's not the world that we used to be in where you would build -- where you would buy differentiation. It's no longer the case. Every company in the world has to be a technical innovator. In order to be a technical innovator, they need to rely on developers, and developers need an entirely rich tool set in order to work and apply their craft. And ultimately, this entire concept of digital transformation is actually just give developer tools, developers are scientists, they're going to do tests to see what works best, let's give them this incredibly rich environment to test. In order to test modern application design, you need the content at the edge, you need the compute at the edge, you need the security at the edge and you need the control. And so it's really a reframing of how organizations, from the most sophisticated to those that never thought that they would hire engineers. It's across that entire spectrum. Obviously, the vanguard of this is our core customer base, which has always been the most sophisticated customers in the world. So when you look at the Spotifys and the Shopifys and the Pinterests and the Etsys, those are the leaders. But what we've seen historically in technology for generations is that as those leaders progress, that starts to get adopted. And what you've seen in COVID is an adoption of that technology because everyone looks at who the leaders are in the verticals they're in, and they see that a lot of them are using Fastly. So we have actually benefited from that as organizations have been forced to transform.

And again, to sort of contrast with some of the traditional approaches to a lot of these product markets, when we think about the programmable web, a lot of these product offerings are walled gardens, right, who have no access to customized or program for your specific business needs because you were just hoping that your vendors sort of figured out the right way, right? And so as we -- as you're on this journey, this multiyear, probably multi-decade journey, with the Compute@Edge platform, talk to me a little bit about the new classes of applications and capabilities that are going to be enabled. And if you can highlight any -- some of the early customer examples and use cases that highlight, what Compute@Edge is -- what's that enabling for customers.

Sure. And we're certainly seeing clusters of that. We're very excited to have our Compute@Edge product out in production and being used by customers. So one cluster I'd point to is actually in e-commerce sector. If you think about a traditional e-commerce company, for many years, they have understood the value of speed. And interestingly, most of them understand the value of speed because most of them have actually intentionally slow down their websites. So you really want to see the value of speed, do an AB test, take one of your groups and slow them down by a second. Do they buy as much as there as much in their cart? Do they transact as often? Do they come back? So e-commerce is one of the first sort of vertical or particularly native e-commerce that realize the value of milliseconds. Now on the other side of that spectrum, you also have personalization. So when I give you a personalized experience, inherently, at least in the old days, I can go back to the central server, that had to be computed. And so developers have always been playing off speed, which they know is valuable. Personalization, which they know is valuable. And they've always felt they've had to trade one off for the other. The advent of Compute@Edge actually is that, that trade-off doesn't have to exist anymore. You can do that compute at the edge so that it's lightning fast. So that we certainly are seeing a whole collection of use cases in personalization. Now personalization goes beyond e-commerce, of course. If you look at ads, video ads, that's another market that's ripe for modernization in the sense of knowing who you are and getting an ad to you. So there's a cluster in what I call personalization. Data and compliance is a growing and very important field, where if you're a customer in Germany, you may not want your customers to interact with data centers back in the United States, for example, or you may want to understand that only users can come from a certain location. So we're seeing a lot of data and compliance clustered as well. We talked in our earnings about gaming, and low latency games are a perfect example of where I may want to meet people locally in my geo, have that all done at the edge. Machine learning algorithms, we're seeing some amazing ones coming to the platform. So it sort of stems across all of these. An IoT example that was fascinating, it was around a vendor who was trying to get people to listen to more on their Alexa devices. Well, unlike my phone, where I may be able to use my app to sort of fade one song into another, I can't do that on an Alexa device today. So what do I do? I do that fading that's very personalized for the user at the edge. So all of these examples are leading to just a complete reframing of how applications are being built and monetized and delivered, and we think that's an exciting opportunity.

Yes. And maybe to work in Adriel into conversations. This is definitely a product heavy discussion, and it was certainly -- I wrote it that way. But Adriel, when we came to Compute@Edge, you definitely been cautious to not let people get ahead of their skis. And then sort of we look into sort of 2021 guidance, what are some of the assumptions around key Compute@Edge that you're making as you build that guidance? And then over a longer period of time, the big question is, like how much can -- in terms of revenue contribution could Compute@Edge account for? What are your thoughts there on both of those dimensions?

Well, there's definitely a lot of excitement on the product side, which I would absolutely encourage, but there's an equal amount also on the financial side. In particular, if you think about our business today, sort of just to think 2020 more recently, we've had delivery as sort of the primary driver. And then we've begun to blend in Secure@Edge. And in particular, 2021 is going to be a year for us for Secure@Edge. Signal Sciences clearly is a big accelerant to that. And in particular, when you think about security and you also think about compute, as Joshua mentioned earlier, it makes most sense to do it at the edge. And from a cost standpoint, historically, cost of revenue for us, bandwidth was about 30% of that cost. So interestingly, both for Secure@Edge and Compute@Edge, there is not much meaningful bandwidth impact there. That's why that gross margin is more likely to be a lot higher. In particular, Signal Sciences has 80% north of gross margins in that regard. And you can imagine, Compute@Edge is also going to have a similar like incremental add-in gross margin as well. Now particularly for 2021, this is going to be a big learning for us. We're learning how customers buy, as Joshua mentioned, which use cases are the ones that are most applicable that we can go after. So when I think about 2021, it's going to be very much a learning. There'll be a relatively a de minimis impact from a revenue standpoint. So I wouldn't assume much for 2021. Secure@Edge is going to be, for us, our focus. And now in '22 and '23, certainly, that's going to be an area where they're going to be continuing to ramp up. But we now know in terms of certain use cases that, at least for Compute@Edge, that could be upwards of a 100%, 200%, 300% of their delivery bill, just that alone in Compute@Edge. So we're very excited about the possibility of it. But that's going to be, again, something for more '22 and beyond.

Got it. And then just in terms of some of the questions that are coming online is around the initial thoughts on pricing, both for Compute@Edge and security at edge (sic) [ Secure@Edge ], what are sort of the -- does that look fundamentally different than how the core business is priced? What's sort of the initial views on how some of these use cases might be priced?

Yes. I would think of Compute@Edge also consumption-based or usage-based, similar to our delivery side. On the security side, as you know, with Signal Sciences, they were subscription-based. And I think that there's a lot of great possibilities there for us in the sense that we know there are some customers who would like to have that predictability of subscription. And I think that's something that in the future, we're going to want to blend into our current businesses because I think our choices -- or excuse me, our customers want choice. And so I think we want to be a company that offers them a choice, and so there will be some situations where it makes more sense to be more subscription-based. And in other situations such as compute and delivery, it's more likely to continue to remain usage-based.

Understood. And maybe, Joshua, it's a good time just to talk about Signal Sciences and why the team is so excited from, obviously, the margin standpoint. But in terms from a capabilities and in terms of building that security muscle, what does Signal Sciences bring the team? And where are you going to -- there's an integration effort. So maybe give us the latest update on the integration side.

Sure. So as you talked about in earnings, it was very encouraging. I think that in order to look at these, you've got to sort of look across the important elements, so the people, which is when you go get a business like this, obviously, the founders and the directors have a say, many of the people don't have the same say. And so you have to look over the first few months, are they staying? Are they engaged? Are they encouraged? So those Metrics look great. So that is important because we bought not only, obviously, the technology and the revenue stream, but as importantly, the future of what this group can build. And there's a lot of avenues where we can continue to innovate. So I feel really good about the people side. We talked about over achieving on the targets that they had set. I think that's incredibly encouraging. What it shows is the ability of this cross-sell mechanism to really take root within customers. I was in a demo with a top 10 e-commerce customer last week, and they reluctantly took a demo to look at security. And now we're in, they had agreed to a POC and are running one a week later. So what you see in this product is 2 things that are really unique. One is that it's all integrated into one -- usability matters, and it's all integrated. So in many of these solutions, you'll have a bot solution, a WAF solution and everything is separated and you're logging into multiple screens. So that's one. Two, this is an organization that, from its early days, looked at intent as opposed to a threat signature. And although that sounds obvious, the entire industry has been built around looking at things that look exactly a certain way. These 9 elements are all placed in this order. You swap them around, they don't look like a threat to me anymore because I'm signature-based. So if you look at threat, if you look at intent and to learn, really get this machine to learn what's good and bad separates them from everyone else. 95% of their WAF products are in blocking mode. So if you think of a house and you think about alarm, most WAFs today, you open up the door, the alarm sounds and the next day, you get a list of logs of when the alarm sounded. Like that's actually how most WAFs are implemented across the world on most websites. 95% of the Signal Sciences WAF are in blocking mode. And you come -- you try to come into the house, they keep you out. And what that means is you don't have to have an entire team of IT people scanning logs to try to determine whether this was good or bad. Machines can do this better than humans. So a couple of elements are really important around threat and understanding and learning and the fact that this product is working with the largest and most sophisticated companies in the world, and therefore, is proven and continues to get better. There's a beautiful flywheel in security. So they were able to capture those things, and we're excited to partner with that and continue to expand it.

Great. And then back to Adriel on Signal Sciences. One of the topics sort of debates coming out of last quarter is, how much is Signal Sciences going to bring into 2021? Can you just talk about how you sort of are approaching that equation? What are sort of you're doing on the product side, the joint go-to-market sales that make it tougher or easier to assign what avenue could have -- to be in 2021 for Signal Sciences?

Sure. I'd say one thing to sort of keep in mind is in the usage-based model, of which in 2021, we're still going to be primarily usage-based. And at this point, I'm expecting Signal Sciences to be somewhere about 10% of 2021 today. Now if you sort of take that number, you should exclude it out, you might say, "Oh, what does that mean for Fastly heritage?" And that's probably sort of mid- to high 20s percent. Like, see, what does that imply? And I think what that's really 2 things. One, if you remember where we were last year, I also approached the year 2020 with a 30% growth. And it's really more of visibility acknowledgment more than anything else. And so clearly, 2020 ended up being very different than I think what we all expected. So -- but I'm still taking the same methodology as I come back to 2021, which is here's the visibility that I can see today. I do believe that over time, as we kind of have more successful Signal Sciences, their subscription-based model will give us actually a little bit more visibility so that actually will help us in our future guides from a [ M&A ] revenue basis. But at least today, I'm acknowledging the visibility I have today. And then the other factor is just that 2020 itself is going to be a compare challenge as we go into it. So I think that the low point of that compare will be in Q2, and I think we'll get a little bit better in Q3 and Q4. And historically, I've talked also about the fact that 2020, from a seasonality perspective, also was unusual. Normally, Q2 over Q1 is a bit flat. Q3, a little bit better. And then Q4 is where we sort of generate growth, not only in revenue, but also in gross margin leverage. 2020 was a very different situation than that. And in fact, I'd say Q2 and Q3 were sort of unseasonably different than years past. And Q4 for 2020 was -- we were beginning to get back to something that we normally see. So I was pleased to see that we were getting back on sort of a normal, familiar track, and I'm sort of looking forward to that again in 2021.

Understood. That's great context. And so in the last 10 minutes, Joshua, let's get back to the growth opportunity and the equation for growth, and let's also talk a little bit about the public cloud guys. And I think it's important to look at Fastly and kind of the bigger, bigger secular trends. And the big ones are cloud computing, edge computing and sort of micro services. And when we think about sort of applications or maybe be accessing it from the web or from a mobile app, those different data components and business logic are sitting in different places. Is that the right way to think about it? And then any way to sort of frame out what you think in terms of modern applications that are built in 2021, what percentage of the application logic sort of sits at the edge versus in a central data center in the cloud?

Yes. So if we step back and we think of this concept of being the on-ramp to the major cloud vendors, I think what's central to that thesis is the fact that we are moving into the multi-cloud world. So I think one of the things that we've seen, we've seen this playbook played out historically as well. If you go look at a data center, I was talking to a great friend of mine. So listen, this is the way I frame the data. For every $100 I used to spend on my web servers, I absolutely would spend $5 to $10 on security, and I'd spend $5 to $10 on delivery, and I would definitely centralize that so that I could play those $100 commoditized out against my vendors. We're seeing that exact same process occur in the cloud business. So if you are completely locked into Amazon, you're completely locked into Google and Microsoft, that's really limiting your leverage. And so if you live in a world where you're going to use multiple clouds, then it's really natural to play back the playbook that we saw in the past, where you have these neutral third parties that are actually the ones where you centralize logging, you centralize security, you centralize some logic, although I think there'll be significantly more in the modern era, and you centralize sort of the transport layer. And really what Fastly is, is all of those appliances that were lined up in your data center that you have no place to put anymore. You don't have a place for any of those edge devices. If you then continue to expand that circle and say -- and actually in its ultimate form, Fastly is also part the web server, or in some cases, with some of our customers being higher web server, then you really encapsulate a very broad spectrum of what traditionally was done in multiple, very specific market by -- in mostly multiple specific devices. I think it's that framework which allows someone to understand this incredible transformation that's happening in front of us. I say that to sort of put that into context. And then I think if you go deeper into that context and you start looking at the types of applications that are moving, we already, for many of our customers, are the point of interaction for the majority of their customers. So when you go to New York Times, you're not hitting a New York Times-owned or a cloud data center. That personalization, that pay wall, that entitlement, if you have a waiting room, all of that is actually happening at the edge. So we already have examples where 60%, 70%, 80%, and in the case of some of these media customers, 99.99% of all of the interactions with their consumers are being done at the edge. So I think the answer as to what percentage is very much a question of what vertical you're in, the type of workload and the paradigm upon which you work. We think over time, that's going to dramatically shift. And as Adriel said, when you actually look at compute as a percentage of delivery, in some of these customers, it's 2, 3, 4, 5x is what they're spending in order to do the compute. It is very often the case that we will go in front of our customers' central clouds and save them 30%, 40% on their central cloud bill, and that's because the work is being done at a different location in a significantly more efficient and performant and secure way. So it's hard to get a sort of narrow and fast rule on it. What I would say is in some verticals, a lot of compute is already happening at the edge. In others, it's less, and that we see a very strong sort of steady march towards that over the next 3 to 5 years.

And I think it's such an insightful answer, Josh, because it sounds like in one way, there's just an attach rate to all of this compute that's happening, whether it's in the cloud or in the data center. And so think of it as even in attach rate -- from an attach rate perspective, it's the useful device in terms of understanding the opportunity. What do you think is going to be the competitive reaction function from the public cloud providers as they see more of that business logic happen at the edge? What are you seeing today?

Yes. So I think you've got different camps. I think you have the incumbent camp, which is really in, North American context, one cloud in Europe who is just has a knack for market share, if I can use that term. And they absolutely are very happy to have a world, which is not multi-cloud and where you're embedded in the ecosystem. I think what we are seeing with our deep relationships with some of the other challenger clouds, and it's sort of surprising to even put that term around such large companies, but those who don't have that in natural market share, they understand this is a multi-cloud world, and they value the ability for organizations to go from one cloud to the other because it's part of how they're going to gain market share. So I would say there are different views in this, but one of the things that I'm really proud of is our relationship with Microsoft, for example, and Google and how they are embracing the same view, which is that the edge and this on-ramp to the cloud is a necessary and different component of the cloud that is not sort of a space in time. This will be here as part of the core architecture of the cloud. This is not something that tomorrow gets replaced by Google. Because if you're going to use Google, Microsoft and Amazon, you're not going to use Google to aggregate or Microsoft or Amazon in the long term. So I fundamentally see that with some of our partners that they deeply understand the value of this.

Great. And maybe with a couple of minutes left. I want to talk about the concept of good traffic versus bad traffic. And I think as a management team, you guys have been pretty explicit about what you're going to go after and what you're not. And so a couple of questions on the front. Maybe just identify what are sort of the more commoditized use cases or traffic and how has that mix evolved until last year. And the other flip side of sort of the higher value use cases that you're targeting, where can that sort of mix go down the road?

Yes. It's a great question. I think it's definitely a tale of 2 cities in the public companies. You see those who have made their living in content that is perhaps not as valuable. I mean you see that impact on gross margin, you see that impact on the business. As you said, we are exceptionally disciplined about this. We could grow our top line revenue, but we've always said -- more than we have, but we've always said, we're going to grow our top line revenue in conjunction with gross margin leverage. That's important to us, and it's been important from the day we started. Because, to your point, there is content that is lower value. The way to think about this is, what do you pay for, right? I mean if you -- and where are these moments that are valuable. When Tiger Woods walks down the 18th green at The Masters, and God willing he is able to do that again, you -- that matters, that moment matters. Everyone's going to tweet about it, everyone wants to see it, the company that is in charge of that [ pairs ] desperately about that content. When you send a gift across the Internet, like, it's not monetizing the way it [ sits on its value ]. So where we have seen the value is in content that is highly valuable to customers and their end consumer, where they're willing to pay for it and the quality, and the security and the scale of that deeply matters. And we have always focused our work on that type of content. There's plenty of content that just scrapes by and is low commodity. And every month, they want a lower price. That's a different market, from our perspective, and we're not in it.

And the tougher question, Josh, was how do you think about traffic that is high value today, but then sort of commoditizes over time? Is that -- does that just require you to continually find new use cases, new sources of data to deliver to customers? Or how do you combat that question?

Yes. I mean, I think the beauty of compute, at least from my perspective, is it allows us to connect ourselves to what matters to a customer. So as our customers, for example, see that traffic not monetizing as well, how can we, with compute, help that be the case? So sometimes that's not monetizing because their content is being scraped. So bots are a great option to help support them in that. In some cases, it's because it's not personalized. How do we help with the personalization? So I think the answer is always we need to give value to our builders so they can build value for their customers. We stay -- stand on our shoulders, and we are humbled to be in very much a servant position to that. What you see from Fastly is our customers shine. We stand behind that. And I think philosophically, compute is an avenue to that, and we really think that's important.

Well, that's great. There's so many things to talk about with the Fastly story, but unfortunately, we only get 30 minutes this year. Hopefully next year, we do it in person so we get a little bit more time. But as always, thank you, Joshua and Adriel, for joining us at the TMT Conference, and thank you to everyone who's listening online on the webcast. Thank you very much.

I'm honored. Take care.