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

Good morning, everybody. This is Tim Horan, the cloud and communications analyst here at Oppenheimer. It is my pleasure to have the Fastly CEO and new CFO, Joshua and Ron, and thank you all for attending. Joshua and I've done this a couple of times before. Ron, and -- it will be his first go around here. We'll try not to beat them up too bad. I think you all know that drill. You can ask questions by chat. I will turn mine on in there in a minute, which I haven't done yet.

But Joshua, thanks so much for attending. I know we got a lot to talk about. There's a huge amount going on. But a lot of change in the last 6, 9 months and the last couple of years. Can you just give us an update on what the company's strategy is and position in the market? And where you're focused from infrastructure and R&D perspective at this point?

Yes, absolutely. It's a pleasure to be here and nice to be here with Ron. So as you know, the world is changing very quickly. And over the last 2 years, that's kind of an understatement and almost getting a little bit tainted saying those things. I hear these executives talk about digital transformation, and I even sometimes roll my eyes. But the reality of what we're seeing, and I was just in a call yesterday with an executive, with a very large customer, is that they are in the process of rethinking so many things. We showed in the pandemic that we can act quickly. And instead of the 6-month time line, we can challenge everyone to do a 1-month time line or a 2-week time line. And I've seen that mentality carry over. And so there's a few fundamental things that are shifting within the organization. The first, we had already started shifting pre-pandemic, but it's almost on steroids at this point is developers are the new decision-makers within businesses, and they are driving change. So that's one thing. The change that they're driving is that they want to be able to write code because code differentiates them. So the fear of all these executives has been and always has been -- has most accelerated them in this notion that somebody is going to innovate in their basement faster than I can innovate. With 3 developers, you can go faster than with my thousand, which is counterintuitive, but with a new set of tools that these developers have, they can go quickly. That's not just innovation on the good side, right? That's just not somebody trying to build a better product. That's also innovation in terms of bad actors. And so we see innovation happening in both of these things almost at the same pace, this breakneck pace. Somebody is going to come and undermine my business with a better business model or better technology, somebody is going to come and undermine my business and steal my stuff, hack into my website. So you've got these 2 passive innovations. Both are being driven by this DevOps-Sec mentality. And what I mean by that is people who, together, all work together, who build tools, so the technology, the software can go through the funnel as fast as it possibly can, the fastest velocity and out to users. Now in the old days, you might remember that we used to do this, let's set a launch of our website in 6 months, and everyone works diligently on that. It's not the case anymore. These actors, good and bad, are releasing software every hour. So that pipeline is a continuous pipeline. And if you look at what they're driving for, they're driving for, at least on the good side, security, stop the bad actors. They're looking for performance because we know that every millisecond adds more to our cart, you click more ads, you view more pages, and they're looking for scale because you don't know in what country at what moment you're going to suddenly become popular, maybe you're subject of a meme or something else and your products you're selling or you're selling. So we need great performance with scale, we've got security. So that's really where Fastly is focused. Fastly is focused on those 3 areas, and we do that by delivering to our customers an edge cloud. And what we mean by that is we've got servers all over the world, and those servers deliver, for the customer, the experience that our customers have in mind. So on our servers, you're not just getting the content. You're not just getting an image, you're not just getting a video. You're actually getting the whole experience, all the personalization, all of the lodging. And that's what we call compute. So we've got compute and Compute@Edge that sits across our servers. So in terms of innovation, for Fastly, it's really in those areas. We continue to do significant innovation on the delivery side. We are investing heavily in the network, we're investing in all of the network technology that makes the experience fast because that's what our customers want. On the security front, we continue to invest heavily to really counteract this really quick cycle that the hackers have. And that means that we have to not patch our systems or put updates to our systems every 3 months, like a lot of competitors do. We have to do this every second every day all the time. And then on the performance side and on the code side, we have this initiative that we launched recently around Compute@Edge a year ago, and we're starting to get good traction. Where -- that's where organizations look at our applications and rethink how they're being developed and that's a completely new way of thinking. In the old days, all you could do is sort of put code on servers. Now you can put code on servers in the central data center, and you also can put code at the edge in order to make sure that, that application is delivered. So those are the key areas: security, performance and scale, and that's Fastly's business. That's what we do.

And on -- so the Compute@Edge, exciting concept. What infrastructure -- why is that coming about now? Why haven't we done that before?

There are a whole bunch of drivers for that. I'd point to a few. One is 5G. The reality is that on our cell phones, we can do more, and the latency and the packet loss between us and the cell tower and actually the telco network has shrunk to such a point that we actually are in a position now across all devices, we were with gigabit into homes, but we now are where that experience can be so fast and interactive that it actually makes way more sense to do it as close to the user as possible in these central -- in these locations that are -- where the Internet connects. So one is the advent of faster Internet, lower packet loss or just a more stable experience. So that's one driver. The other driver is our drive as humans, what we know and what all the tech show is that the more personalized, more interactive these websites are, the more we'll engage. So there's a drive based on all the neuroscience that what we know is that if we can get those to be more -- get those dopamine hits in our body faster, we actually can drive better engagement. And so it's -- I'd say it's a combination of those 2 things plus the technology has never been at a point where we could do this. And now in -- and that requires enough memory at the edge, fast-enough CPUs at the Edge. So there's also this underlying technology drive which has finally allowed us put enough compute, enough storage and enough application logic, all out there in these locations in order to have this function. So it's not one thing, it's multiple things that make this a new era that we're walking into.

And you're close to the customers that want this, really these 3 products and extremely low latency. Could you talk about some of the new use cases you're thinking about? And how important it is to them and new business models you're seeing?

Yes, so I mean it starts with old business models being rethought, I would say. If you think about an e-commerce store, what we know is that the most successful ones personalize that experience, like I was saying. So why do you have to trade-off? If you're an engineer, you ask yourself this question. I want to make it personal. I want to make it fast. Both offer business value. And today, that's a trade-off. That trade-off doesn't have to exist. You don't have to say for every personalized page, I've got to go back all the way to a data center in Washington D.C. from Los Angeles. That's a trade-off people are making today. If you want it to be fast, like you'd put it on -- I can make it dumb, I can dumb it down, I can take all the personalization out of it and I can serve it to you from LA, if you live in LA. But if I want to make it interactive, I'm going to take you all the way across the country. So our first use case that we're seeing come out in production are actually about personalization. So how can I make it unique to you? We then have the advent of these protocols of these real-time interactive experiences which aren't even related to personalization necessarily, but are reactive. They're reacting to what you do. I have an input, I give you an output. Traditionally, that input, output system has also had to drive you all the way back across the country or to another continent. Because the servers, there's only sort of one brain that's able to do this. As we can take that brain and put it on locations all around the world, you now can do more. So we see that in its basic form, just input, output. We're also seeing more sophisticated models that you traditionally associate with things like machine learning, where I can take an input, I can do a calculation or infer from a model that's created and I can send that out. And there are some fascinating use cases also that we're seeing as the world becomes -- as data and privacy become a higher priority for our customers. We're also seeing this idea of starting to do that at the edge. So I know where you are, I know where you're coming from, I know who you are, I know what data you can access, and I can actually route it the right way, and I can make sure that I comply with all of the many, many laws that are out there in order to protect privacy based on jurisdiction. Those are 3 examples. There are many more, but those 3 definitely stand out. And there's a few examples, one that we talked about in the earnings letter, the shareholder letter around GraphCDN. We have another one, LaunchDarkly. And these are all examples of that. Lastly, one area that is growing for me is the use of compute and security together. So we have a partner called HUMAN. And they are -- they developed some of the most sophisticated bot detection solutions for ad fraud. What they know is that, in the same way, they have this trade-off. I need to make it fast to make sure I detect it, so I don't slow down the ads. But if half -- if you're coming in from a huge bot net and that button is just scamming and clicking on the ads, like that's a bad thing. So how do I trade that off? Again, these trade-offs, and what I think compute brings is to really fundamentally reexamine the foundational trade-offs that we have made over the last 20 years in the web industry. I need it to be fast, but I can't. I need it to be personalized, but I can't get it fast, those trade-offs. So that HUMAN example is a brilliant one. We're allowing them to do it faster at world scale and make sure it's secure, getting the value proposition of the business.

So I think virtually no applications now have this edge compute or very few, a couple percent maybe. But if we're sitting here 5 years from now, what percent of use cases will have some Compute@Edge capability?

Yes. There's a continuum here. I would say, if you look at -- Fastly started 10 years ago with this idea that we wanted to empower developers and engineers. And so if you look at Fastly's customer base, and admittedly, it's not the whole Internet. We only have 400 enterprise customers. It's very early days. But almost all of our customers are using edge compute. And as I say, it's a continuum. Because the sort of use cases where you're putting your machine learning algorithm on and you're inferring, that's one extreme of sort of what we haven't seen today in production levels at scale across the Internet. But we are seeing many customers as a run of business, as a normal course of business, using edge and computing at the edge in order to deliver the largest websites of the world. So if you go look at the largest websites in the world, they are using some form of this today. And you're absolutely right that the sophistication of it and the complexity of it is changing dramatically. I think in 5 years, because of the concentration of the Internet in a few hands, if you just look at it by traffic, my guess would be we'd be upwards of 90%.

So the question is, how do you dominate this Compute@Edge? And I'm not -- I use the word dominate loosely, but maybe you do. I mean it seems to me like you need a neutral infrastructure provider, frankly, for a whole bunch of reasons. And you need someone who's going to create a platform with network effects, right, where the developers feel comfortable. You're not competing with your customers and they're not competing with one another. Is your goal to create a platform or are there -- how do you do so? Are there any platform companies that have been created that you try to emulate?

Yes. I mean there's a lot there. I think you absolutely -- I think you hit on the heart of the business. We benefit by being a neutral third-party. And I was talking to a mentor of mine and a CTO, one of the companies that probably spends the most or the second most on data centers over the last 20 years. And they told a really interesting story about how they thought about the economics of a data center, right, a traditional data center. So you would throw in -- the mentality was I have one commoditized thing, and that's the servers that I bring into my data center. And the more commoditized I can make that, the better for me. For every $1,000 I spend, I spend $950 on these commoditized servers, and that's what I need to commoditize. And then what I need to do is I need to build, I'm going to invest in technology that allows me to do that. And that means I need neutrality. I need that technology to not be aligned with the server business. And I need that technology. I know I'm going to spend more on that technology, but it's critical for me in order to accomplish my goal, having the lowest possible price on that $950. And so what's really interesting about that mentality, I think, if you're going to spend $1,000 and $950 goes there, is that that's the way we always thought of building out our data centers and our compute infrastructure by getting compute, like the core compute to commoditize. And I think the edge is in many ways that middle, neutral third-party that allows us to do that with the central compute vendors. And so I agree with you. I think our customers are coming to us and saying, "We need a neutral third-party." You cannot be my central cloud to do this job because one of the core parts of your job is to distribute traffic across the central clouds, is to centralize some services. So one, I agree that there's a need for a Switzerland here. And there's not going to be one. There'll be multiple. So that's number one. I also think that player is going to have to be a platform, i.e., exactly what you said, people are going to have to be able to build on it. They're going to have to feel like that building is not leading to their own demise in terms of their competitive stance. And they want to know that there is this flywheel effect, right? I mean what our customers get when they come on Fastly is they get these products out of the box, they can come -- they can create their own, they can extend them, and the benefit that comes from that from other customers circles around. So the more people we have on the platform, the more features and modules we have, the more value they get. And that's really that platform effect that is such a beautiful thing. So I think I look and say, that's the characteristic that we are going to look for. Now Fastly is in the early days. I mean, I talked about it earlier, we've got 400 enterprise customers. If you go look at the large network vendors, they've got 30,000, 50,000, some of them 100,000 customers. So we believe that this is a very early story with a ton of future potential, and that's how we're looking at it right now. So we're investing.

So when you look at like -- I'm not sure where you think the mindsets are right now of the big 3 cloud guys, but AWS has a joint venture with Verizon for the wavelength building out right to the edge. Microsoft talks about building an intelligent edge. But do you think that they will embrace more of a neutral provider to help them build out an intelligent edge? Or are they still trying to do with more vertically integrated?

I think part of this is this term edge, and we try to demystify it a lot because ultimately, there are a lot of edges, right? If you think of your cell phone, that's an edge. I've got an edge in my house. I've got a data -- I've got a tower that, that cell phone connects to when I'm on a cellular network. That tower connects on to the Internet at an important juncture. So there are multiple edges to the Internet. And I think in order to understand what the large cloud vendors are doing, one has to appreciate that there isn't one edge, that this is a continuum of points. What's important about Fastly's edge is we are what we think of as the developer's edge. We're the last place in the line that a developer can truly trust that their code is going into a compliant environment with the compute and memory capabilities and storage capabilities that they need. And that's different than a phone. I mean I can't trust what's on the phone. I've got really limited compute and storage capabilities, and so I think all along that continuum. So if you look at some of the initiatives that you talked about, they are very different initiatives. A lot of those initiatives are actually about bringing core compute closer or improving the network connectivity between devices. All of those are important initiatives, but they're not the edge cloud. They're something different. And so I think the answer is, we continue to believe, and I think this is validated by the partnerships we've just announced, one -- we've announced a number of them with the large clouds and enhanced one recently with Amazon, is that, that relationship necessitates the 2 parties. We need the core compute. We need the edge. Both are important, and I think they play different roles. So I see a lot of synergy there because we're all trying to drive to the same thing. We all want to lower the cost of -- for our customers. We all want to make sure that, that experience is fast, secure, scalable. But our mission is not very different. We just do things that are different.

And so Joshua, is there any platform companies that have been developed that you kind of admire and you kind of see what they've done and you try to emulate a bit? Or is this kind of greenfield, we got to figure it out on our own?

No. I mean I think there's some great examples out there. I look at the developer community that Microsoft's been able to build and how that has withstood the test of time, right? People build on that platform. They share on that platform. I think the underlying strengths of that Microsoft business have really shown in the last few years. So I think that's a really nice example of a business that has as in some ways is a platform that has embraced the cloud, that has a strong developer following and that has really transformed this developer view. If you think about their view 10 or 15 years ago, I wouldn't have been the one who came up here and said, "Listen, this is a group that really understands non-Microsoft developers." Like that was a [Walt Gartner] approach, right, which is not actually going to work the same in the modern era. But they completely transformed that business. I look what they're with GitHub. I'm very impressed. And so I think that's a really good example of a company that has executed in the enterprise, because I think that's important, right? Fastly is an enterprise company. We serve the largest customers in the world. You look at our $700,000 average price tag for enterprise customers. We're not talking about the small fry here. We're talking about the largest in the world and those that aspire to be. So I really admire what [that scheme] is doing, I must say.

So you just reported earning, and I know you and Ron have been working basically around the clock since you report earnings, talking to investors. Do you want to hit us? And obviously, you've been saying some of the right things that your stock is stabilized and bounced back up here. Can you talk a little bit about what you're telling investors on the quarter what you felt was important? Any other color around the primary questions you're getting asked?

Yes. I mean, as a business that's driving towards $1 billion of revenue and 30% year-over-year growth or greater, I think one of the challenges is when you have a disappointing quarter, like the one we came out of, it hurts. So I think I'd start with the fact that it was a disappointing quarter from my perspective. It was disappointing for our customers. We had a large outage in early June. That outage had a knock-on effect, both in the quarter and in subsequent quarters. If you look at what we saw out of that quarter, we continue to see that we're a business that is not so large that we have the ability to just sub in and sub out large customers. So we had -- we talked about one customer in our top 10 who paused because they had some remediations that they needed us to take care of post outage. Thankfully, we've taken care of those remediations and it's now in the customer's hands to decide when they come back. We are very confident that's coming back. We also had several customers who were not -- didn't have the best fidelity on the timing that they thought their projects would launch on. We thought that they had locked into some timing in Q3. It turns out, unrelated to the outage, unrelated to Fastly, we had a few initiatives that pushed to Q4. I think the side of the story, and you can look as the glass half full, for me the glass half full is we had an outage, almost all of our customers are back 99% plus, and we've got one top 10 right now that we're really focused on bringing back into the fold. We had situations like this in the past. And if you remember, we had a previously disclosed largest customer. They never left the platform. They remained on the platform, but they did taper their usage given some of the geopolitical issues that were going on at the time. So Fastly has a history of having these customers and having them stay with us, and I believe that will be the case. For the delays that we saw, which were unrelated to Fastly, we're still built into those particular applications, I think that the silver lining is we see that hitting in Q4. We think all of these things are an impact only in '21 right now based on what we're seeing. And that sets us up to get back to that growth path that I think we all believe we should be on. I mean I said on the call that I see 10 years of 30% growth, and I believe that. I mean, fundamentally, we are so early in this story arc, 400 of 50,000 or 100,000 enterprise customers, there's no reason that, that shouldn't be the case. So I think the ultimate story is a hiccup, a disappointment. We don't like disappointing our customers by taking them down. We don't like disappointing investors by not having a forecast that we think is as accurate as we'd like. But things are changing in our business. It's some early days. And I think the trends continue to point in the right direction for Fastly. This is the right market at the right time, and we have the right solution for it.

So I know you discussed it in the blog and on the conference call a little bit but can you talk a little bit about the most important steps you've taken to ensure that customer networks do not go down again and other options that your customers might have?

Sure. So that's a very important question. I think one is to say Fastly -- we've been in business for 10 years. The last issue we had of this magnitude was 2015. And so I think that sort of speaks to, one, how rare these are, but also acknowledge that they can happen, right? So that's sort of -- that's one thing. I think if you look at where we have put the focus, the focus historically, if you look at our historical issues, the challenge was we found them really fast, but because of the nature of the software, it actually takes us some time to get back up and running. So it's sort of like the machine that has a little bit of a start-up time. And we've really focused over the last couple years on making that -- taking that start-up time from literally what could be half an hour down to like 90 seconds. Now what we found in this incident was actually the opposite. We actually -- because of those innovations on the start-up time, we were able to start up really quickly. We just had a delay in actually identifying this. We have quite a complex system. We have -- if we talked about the 400 enterprise customers, but thousands, couple of thousand customers, and some of our customers have hundreds or millions of their own customers. And so when you have something that goes wrong in the system, to identify and to pinpoint really, really quickly why it went wrong and who -- and the rationale for it can be quite a complex undertaking. Now as I say, we've been exceptionally good at that. In this case, I'm very proud that we're able to do this in 30, 40 minutes, but I want that to be faster. And so a lot of the remediations are in that area. How can we make sure that we get faster pinpointing the challenge? And how can we automate some of the things that we have been doing manually in the case that something is not going right? So we have a number of remediations that are all pinpointing around when something looks like it's breaking, how can we jump in automatically and isolate that and triage that for our responders. And there are a lot of innovations that are going in that [revenue]. I mean like the companies we talked about earlier, Fastly is also in this DevOps-Sec world, where we are constantly pushing code as well, and we also need to make sure that our own code is safe. I mean in this case, this was a bug that we introduced to the software. It was a very edge case. We didn't find it in any of our testing. We actually didn't find it when we rolled it through all of the historical configurations we've had. We didn't find it when it was in production, but we also need to acknowledge those are there. So I think that's really the area of focus. Now in the medium and long term, one of the reasons we've built Compute@Edge was for our own purposes. And one of the values of Compute@Edge is its ability to isolate one customer's request from another. And in that case, that would have -- if we had run this type of thing through the compute platform, we actually would have been able to isolate it. In fact, the compute platform was unaffected directly by this issue. Now the compute platform sometimes ties back to the current non-Compute@Edge platform. So there are sometimes ties together, but for customers that were just purely running on compute, they were isolated from this. So that's a good sign because our vision has always been to migrate as quickly as we can all of our customers to that platform. And that's something that is part of the medium-term undertaking here.

So do you think you've made a lot of headway since the outage that this won't happen again? Are customers pretty comfortable at this point?

Yes. I mean I think that, again, glass half full. If you think of our 400 enterprise customers and think that within a couple of days, they're almost all back, every single one. Now we're focusing on this one that is a top 10 customer. But I kind of take solace in the fact that 390 and many have decided that they do trust the mitigations and they trust our path. I think I'm not happy until it's 100%. I mean let's be clear, that's the goal. But I like where we're at in terms of achieving that goal. And I think what we're going to see, given the relationships and most importantly, the importance of what we bring. I mean, the customers are back not out of charity. The customers are back because we help their business. And I think that's the key, and I think we'll absolutely see that in this top 10 customer as well. My hope is in the -- as soon as we can.