Welcome back, everyone. I'm Rishi Jaluria, a software analyst here at RBC. I'm delighted to have with me from Fastly Ron Kisling, who's the CFO, and Vern Essi, who leads IR. Guys, thank you so much for being here.
Thank you.
Thanks, Rishi.
Maybe let's just start with a little bit of a brief overview of Fastly, what it is for the generalists in the room. And maybe, Ron, can you talk about the evolution of the business since you took over as CFO?
Yeah, so Fastly is, you know, first and foremost, a CDN. We built a network that provides really high performance in terms of delivering cached content at the edge to users on a compute network that allows us to also run applications at the edge, and then shortly after that, the company bought Signal Sciences and added security with a web application firewall to the portfolio, and since then, has continued to build out the security portfolio to what really today is a complete web application portfolio of WAF, bot detection, and DDoS, with a compute framework and observability where you can write applications to run those at the edge to provide really fast, responsive interaction with users.
We just launched, and we'll talk more about this later, our AI Accelerator, which really leverages our compute platform at the edge to provide really responsive interaction with users as well as reduced costs. The focus, since, you know, I've been with Fastly about three and a half years, Todd Nightingale, our CEO, joined about two years ago. We've really been focused on, I guess, two pieces, which is really making sure we have an aligned strategy across the organization to execute on go-to-market, driving, accelerating our revenue growth, and acquiring new customers, as well as building sort of expense rigor internally in terms of running a really well-run business to enable us to make those investments in the business.
Great. And then maybe just briefly, can you talk a little bit about Q3? So it's a little bit of a nice bounce back, obviously reflecting the stock price. What maybe changed over the past couple of quarters? And let's go through some highlights.
Yeah. I think two highlights I'd cover on Q3 was. I think first and foremost, you know, despite some headwinds that we're seeing from a handful of our largest customers in the quarter, we continued to drive diversified revenue. So we saw our top 10 customer revenue decline 11%, but customers outside of the top 10 grew 20% year- over- year, demonstrating market share gains in that space. The other big highlight from Q2 was, or excuse me, Q3, second highlight on Q3, really was we had record EBITDA and record non-GAAP net income in the quarter, really coming from the restructuring that we did in the quarter and the ongoing efforts around really expense rigor that we've had in place over the last couple of years.
Yep, yep. So I definitely want to get into the top 10 versus non-top 10 dynamics, but I also want to maybe separate the different pieces of the business. Can you talk a little bit about the different revenue segments, right, between kind of the traditional CDN, security, some of the new edge offerings, and what are some of the efforts that you're putting behind both security and edge to drive better growth there?
Yeah. So we really look at the business in kind of three segments. We have our network services business, which is primarily delivery. It was about 78% of revenue in the third quarter. So that includes the delivery network. It includes load balancing. It includes Origin Inspector. We have the security piece, which includes our complete security portfolio of web application firewall, bot detection, DDoS, our Private Relay, which is primarily used around websites. That was about 19% of revenue in the third quarter. And then lastly, we have other, which includes some of our more emerging businesses, newer businesses, our Compute business, the Observability. And we recently announced AI Accelerator. That's in beta release. We're going to release before the end of the year into general. And so that'll be a revenue contributor in 2025.
While that's only 3% of our revenue, it did grow 85% on a year-over-year basis.
Yeah. Maybe let's dig a little bit into AI Accelerator. So really exciting to see that. Can you maybe talk about, number one, what sort of applications do you see customers potentially building with that? And number two, if we really think about what puts you in a better position to succeed from that, right? The center of gravity is around Fastly.
I think first and foremost, when we designed the AI Accelerator, we focused on a need that exists today in the market because there's a lot of future coming with AI, and what we're seeing is a lot of traction in companies that have customer service with automated response, you know, bot detection on help desks. What this does is we're able to cache the responses from the central cloud locally, which allows us to serve those responses back to the users much more quickly, and because we can reuse those for similar inquiries, we also reduce the cost of the egress of these queries to the central cloud. We're seeing cost savings as well as, you know, a better than six-fold increase in just response time back to the users.
I think one of the things that makes us particularly good at this is we have the highest performing network. So being able to deliver that highly performant response, being able to cache the content that gets used most frequently. We have a semantic engine at the edge that allows us to look at, you know, differently worded queries that are asking the same thing to be able to reuse the cached content.
Yeah, so maybe let's drill into the last one. There's a lot of conversation around edge inferencing. And you know, you've talked publicly about some of the use cases there. How should we be thinking about what that edge inferencing opportunity looks like, especially as a balance of power in GenAI shifts more from training to fine-tuning and inferencing of these models?
Yeah. I mean, I think with AI, similar to what the CDN network did with content early on, it's really about being able to provide, you know, the content or the AI content very quickly with rapid response, reduce the amount of egress, reduce the cost, so make it a lot more efficient to timely deliver content at the edge, and things like the semantic engine that we have or other things that allow you to do, you know, fairly simple allocations to be able to determine what content to serve up from the cached content. I think there's a number of applications even beyond help desk bots where this can serve up that cached content efficiently to users.
Yep. Okay. So maybe getting past kind of Q3, if you think about Q4 now, what needs to go right within the company, within all these different growth drivers to sustain the kind of growth profile you want? And I'm sure there's kind of a longer-term target that you obviously strive towards.
Yeah. I mean, I think, you know, in Q3, we saw a lot of the growth by better than expected traffic from live events, you know, in gaming. And we also saw, you know, better than expected kind of growth across, you know, the non-top 10, where we've seen, you know, 20% year-over-year growth. I think as we look to Q4 and the headwinds that we've seen from the largest customers, I think we're going to continue to see those. But we do see us getting to the end of those headwinds on those largest customers as a lot of the transformation that we've seen there, as they've addressed, you know, lower traffic growth rates, bringing in different vendors has now been reflected in the dynamics and our traffic share. Those are now kind of reflected.
And so as we get into 2025, we're going to be focused on new customer acquisition, accelerating our security growth. And those are a number of things that our new CRO, Scott Lovett, will be focused on.
Yeah. Maybe let's drill down into some of the security side. So you have this full portfolio kind of starting with the Signal Sciences acquisition. What have you seen so far with bot mitigation, and then some of the adaptive DDoS stuff within the security portfolio? And when can that maybe start to help on the new customer acquisition side?
Great question. On the bot detection, we've seen really good response. I'd say the response has really exceeded our expectation, and what we found previously is, well, we had, I think, the most performant delivery network in the space and probably the best WAF. You know, once you get outside of, you know, the largest multi-vendor streamers, people do want to buy a platform. You've seen that migration. You've seen where the companies that have a full platform are continuing to be successful, and the standalone vendors are struggling, and so not having a complete portfolio did create friction and issues in terms of, you know, some of the customers that we approached. Now we have the complete portfolio, and so we're able to compete for all of those deals.
And as we rolled out bot detection, the adaptive DDoS, and focused on improving the integration with the platform, we entered 2025 with a full integrated security portfolio into our platform. And so that enables us to build on these. And then what allows Scott to do is really focus on incentive plans that incent new customer acquisition, as well as expanding the security expertise within our sales organization to drive an acceleration in our security revenue.
Yep, yep. Okay. That makes a lot of sense. Maybe let's stick on kind of some of the transformation in sales and go-to-market, right? So as you've talked about, what kind of, how should we expect some of these changes Scott's bringing to, like, pay out? What metrics are you tracking internally? What should we be paying attention to on the outside?
You know, the focus is really about, and we've talked about this, the key driver for our growth is new customer acquisition. And so he's looking at our sales incentive plan so that it's highly focused on new customer acquisition, looking at the sales team in terms of we deploying them where we see the opportunities. And then particularly, we've talked a lot about accelerating our security growth. We've now built the portfolio, building, you know, increased expertise. You know, Scott brings a lot of experience from the security space in terms of building a sales team to sell that and driving that. So I think if you look at the key metrics next year, we're going to focus on new logo acquisition. I mean, revenue growth is in there, but also what are we seeing in terms of security growth?
Our goal is to accelerate our security growth on a year-over-year basis, and those would be the two metrics I would pay attention to.
Yep. Okay. So we start on the revenue side. Now let's go down the line. Maybe let's first start on the gross margin side. You know, there's been progress over time. I'm sure that there's still a lot of room to grow. Maybe help us understand how should we be thinking about the drivers of future gross margin expansion from here: network optimization, revenue mix shift, et cetera?
Yeah. I would say there's sort of the base piece where we continue to drive as we scale, you know, more efficient use of bandwidth, continuing to deploy engineering fixes to drive efficiency in our hardware. And the other dynamic is just as we expand internationally, you know, as we enter those markets, you know, our economics aren't as good. And as we build up capacity, we start to bring those costs in line with what we see in our more mature markets. So that would be the core gross margin trajectory that we have. But as we see more adoption of the platform, more adoption of security, deployment of compute, you know, those are all additive to gross margins because in many instances, you're running those on top of the same traffic. You know, the incremental costs of adding in security are relatively low.
The next big material adders is just increasing the attach rate of security, compute, and observability to the platform.
Yeah, yeah. Okay. Then maybe let's drop down to the bottom line. How should we be thinking about the potential for Fastly's future, if it's not really free cash flow margin expansion over time?
Yeah. I mean, at a macro level, we've taken a really strong view in terms of looking at kind of what our model is in terms of sort of how do we want to spend incremental revenue. And so we've set a target of, you know, spending at like 35% of incremental revenue on an LTM basis. And with the restructuring that we did in Q3, we're set up for an exit rate going into 2025 that we believe will enable us to make those additional investments and go-to-market around the security expertise, putting in salespeople in the regions where we need it and drive new customer acquisition, and at the same time achieve free cash flow break even in 2025.
Yeah.
And if I can pick up here for a second, Rishi, I just want to, for those that are new to the Fastly story and online, we have talked about our target model in the past in terms of how the business works. And Ron kind of got into the incremental revenue point a little bit here, but for every incremental dollar of revenue that we generate, our gross margin should drop to 80 cents of that dollar. And then we're spending about 35-40 cents on that in operating expenses. So incrementally, we should be driving a 40% margin and an 80% incremental margin. And that's on a trailing basis. If you look at our numbers over the last eight quarters, you'll see us trending up into the high 70s on that ratio for our gross margin.
That's basically just how the business should work as we scale.
Yeah. No, that's helpful color. Really appreciate that. And then maybe kind of putting it together as we think about kind of balancing growth and profitability over the next several years, you know, what should we be looking at? Are there any targets you have in terms of Rule of 40, or should we be looking at like free cash flow per share growth over the next several years? What's the right way for that?
You know, I think, you know, I will come back a little bit to Vern's model in terms of, you know, how we're looking at growing that, which will drive, you know, incremental, you know, operating profit of, you know, 35%-40% on incremental revenue. And that's going to drive us to, you know, profitability, you know, on an LTM basis over time. And I think that's, I think for the foreseeable future, that model can drive our, you know, transition from kind of our operating losses to break even and to profitability while investing in the business.
Yeah, yeah. All right. Then maybe let's turn back to the fundamentals of the business. Let's start with competition, right? That's a big one that gets brought up. How should we be thinking about the competitive landscape and any changes there? And, you know, primarily it's Akamai and Cloudflare. Yeah, let's start with that.
Yeah. I mean, I think as you commented, you know, the competition that we see, you know, largely is Akamai in the enterprise space. We do see Cloudflare coming into that space. I think the other dynamic, you know, Edgio, you know, they did have a presence in a lot of the multi-vendor accounts. I think as they exit the space, it's going to become, you know, more of an Akamai, Fastly, Cloudflare space. I think that will play favorably in terms of some of the pricing. I think what we saw is across the general space. I'd say pricing this year has been kind of in line with the longer-term trends. Across some of our larger multi-vendor accounts, we did see some acceleration in price erosion this year.
We think with the competitive dynamics, the way they are, that that will realign back to what those longer-term trends have been in terms of pricing. And so I think that will help make the business healthier. And I think the way we look at it continued is when we look at the performance differentiation now with a complete security portfolio, we think that there's an opportunity to continue the share gains that we've seen, you know, across the market as a key driver of revenue growth, you know, in the future.
Yep, so now let's think about, yeah, kind of your customer base. You've got the, you talked about the top 10 cohort and the non-top 10 cohort. What steps are you taking deliberately to diversify, right, and because customer concentration in the history of CDN and edge has always been a bad thing, right, so what steps are you taking to diversify that? And then are there tools you have in your arsenal over time? You mentioned maybe pricing pressure gets slightly less bad with Edgio out of the market, but things that you can do to get maybe some level of stabilization in the top 10.
Yeah, I think there's a couple of things. Certainly with the top 10, when we started to see these sort of new dynamics as they were looking at, you know, much slower traffic growth than they've seen in the past, and they brought in vendors. You know, we've changed our engagement where we now engage our senior team with their senior team on a regular basis. And we've seen our visibility in terms of what their plans and strategies are, how they allocate traffic, improve dramatically. And in some instances, we've been able to actually influence that. So that's helped our predictability that, you know, clearly in the first half was not where it needed to be. Secondly, in terms of just, you know, overall, it's really about, you know, continuing to add new customer acquisition. You know, that's where we see, you know, growth drivers across the industry.
It's where we can add customers across the portfolio, which again drives favorable margins.
Yeah, yeah. Okay. That makes a lot of sense. Then I want to think maybe within your enterprise customers, any trends that you're seeing that have changed in terms of some of the KPIs, churn, retention, how they spend over the past, call it two to four quarters?
Yeah. I mean, in terms of the trends, we didn't really see much in the way of new trends in Q3. Our churn was flat with Q2 at a very low level. What we did see was a much larger number of customers that fell just under our $100,000 annualized revenue, and so they fell out of the, you know, enterprise space into the other customer space, and we do see some variability there, you know, because we do measure it just on a quarterly basis, so I say it's somewhat variability, but we didn't see any increased churn. We did see overall customers increase pretty dramatically in the third quarter.
Okay, then I want to maybe think about the edge business, so we talked a little bit about edge inferencing and some of these AI initiatives, but maybe let's take a step back and think about edge computing as an opportunity, you know, it's something that Fastly has been talking about since IPO, and maybe that moment is here where we're actually starting to see real traction in there. Number one, can you walk us through how we should be thinking about edge computing versus central computing and, you know, what are kind of the use cases that are better suited to edge, and then number two, what about kind of your infrastructure again gives you the right to win in that space?
Yeah. I think the applications that are most suited to the edge are those where latency matters, where, you know, you have, you know, reused content. I think the AI Accelerator example is a great example, but there's lots of instances where you have data that you're interacting with the users at the edge and you don't have to go up to the cloud to do it. Again, driving that efficient response. I think we're still in the early days of identifying applications for companies to deploy. I think there's a lot of experimentation going on. I think AI will be an accelerant to identifying applications at the edge using compute. So I do see it still early days. And AI, I think, is a catalyst to build out a lot of those applications.
Let me, if I follow on this too, I think the technology that we have, our platform itself is, to use a phrase here, our right to win. I think one of the big key differentiators that we have relative to our competitors, especially our largest competitor, Akamai, is that our entire network is a compute framework. So all of our servers, everything that we do runs on compute. And we have programmable switches that are attached to that. And it's entirely unified. So within minutes, we can upgrade all of our servers at once with the push of a button and push a software update through. Other platforms are put together in a calico fashion. They have centralized compute in one spot. They might have a stack of routers in another spot.
It's very hard for that to run efficiently and at the same sort of unified platform approach that we have. Those are advantages that we have. Of course, we lack scale and we need some other things on the go-to-market side that we're fixing to compete and be very effective on that. But that would be our right to win. And I think in terms of running compute workloads, our platform was built that way from day one. So for us, it's actually very easy to do. The issue is the market really hasn't come into us yet. We're starting to see that happen probably in the last six months. And certainly with AI and this inferencing, we expect to see a lot more of that next year.
It gets back to the growth rates that we saw, which is, it's, you know, 3% of revenue today across Compute and Observability. We did, the growth rate was 85% on a year-over-year basis. I think we're going to see that accelerate, particularly when we drop our AI Accelerator into that product group for 2025.
Yeah, yeah, so maybe sticking on AI for a second, how should we be thinking about now the very early days, but what would monetization for this look like over time? And does it require maybe different types of pricing models or can it be consistent with the rest of your business?
Yeah. I mean, I think, you know, the pricing models, I mean, you know, there are going to be some iteration of what we have today, but I think, you know, looking at, you know, requests served up is a good metric, particularly with the AI Accelerator. I think, again, opportunities where you can reduce the cost for users, because right now the cost of, you know, these queries is very high. And so being able to reuse those, store them at the edge for interaction, because one of the big uses of AI is, you know, interacting with individuals and giving them information at their fingertips. And so delivering this through, you know, an edge is the perfect way to deliver it. Because absent that, you are looking at, you know, higher latency and higher costs.
I think there's going to be a lot of applications where the actual delivery to the user is going to come through an edge platform, just like content today, you know, is almost always served through an edge network rather than always served from the central cloud. I think you're going to see the same thing with AI.
Yep, yep. Okay. That's helpful. I do have a few more questions, but I just want to make sure if there's any questions from the audience that you have for Ron or Vern. Okay. I'll get back into things. Maybe I want to go then talk about international. You talk, you know, that can be a little bit of a drag on gross margins initially, but as you scale up, you kind of get it at parity. Number one, can you walk us through architecturally what can you do to, you know, bring those gross margins to the right place or is it just scale? And number two, how do you think about identifying the right geos where it makes sense and maybe where it even makes sense to walk away from because the margin opportunity just isn't there?
Yeah. You know, I think, you know, largely a lot of building it, we have a good plan in terms of identifying what's the pipeline there, what are some of the key customers. A lot of times there may be an anchor customer with a level of traffic if it's a brand new market that we're going into. I think once we're in a market, it's a lot easier to develop it because we can respond to growing the capacity very, very quickly, and we've seen instances in some markets where, you know, we've been able to go back to our bandwidth providers, you know, several times a year with, you know, higher traffic and reduce the rates, and, you know, that affects all the customers, not just the new customers in those regions.
The deployment into new markets is we've got a good process in determining, you know, what is the, you know, anchor customer. And then as we expand, it's very easy for us to grow with that capacity almost real time as that traffic grows.
Yeah. And just kind of sticking on traffic, you know, you have seen some upside from live events and the like. How do you think about, you know, what impact that's going to have on your model going forward? And what ways can you maybe get a little bit more predictability around that on both the upside and downside?
Yeah. You know, I think broadly, some of the work we've done with our largest customers who are doing a lot of the streaming activity is going to help us a lot. Revenue diversification is the other big piece. You know, we're starting to see that with the growth outside of the top 10. You know, and while some of the decline in the top 10 concentration came from declines in the top 10, we're also growing the other piece. So as that number comes down, that also reduces some of that volatility and some of the challenges in predicting it. But I think one of the things that's a great opportunity is we are seeing more and more broadcasters or content trying to do live events. And it's one area that we do exceptionally well. I'd say we do it better than anybody else out there.
There's only a handful of, you know, two, maybe three, that can even do this reliably. And I think that represents a great opportunity as well as, you know, companies are moving more and more to doing live events versus static events. And so, you know, the growth of streaming around sports has been a great opportunity as well.
Public service announcement.
We were not involved in last week's major live sports event for the record.
Good to know. So.
Sports event is a broad term for that.
Yeah. Sorry.
So then I want to maybe think about on the security business. You know, you've been rounding out your portfolio, you know, primarily through organic development. Where do you see maybe adjacent opportunities within security? Are there net new areas that you think you can go into? And how do you weigh the build versus buy decision there?
Yeah. You know, I think there's, if you look at security, first and foremost, you know, it's always evolving because you have, you know, the people who are trying to get past security are always innovating as well. So you're going to continue to need to develop your products and address those issues. You know, I think API security is next. You know, we do a lot of that today through our web application firewall. But I think over time, what we've seen is, you know, we've addressed these issues in our existing products and then moved to maybe productize those, creating more of a dashboard, more productized, easier to deploy with more sort of reporting and observability about what's happening.
That's kind of what we've seen across some of these new areas of security where we can address it and then we continue to enhance how easy it is to deploy.
Yep. All right. Maybe in the minute we got left, how should we be thinking about the impact of AI on security? And maybe we need to think about it from two angles. Number one, where can you inject AI in the stack to make your security products more effective or even come out net new? But then number two, as you talked about, attackers are getting more sophisticated, right? Getting AI in the hands of them means that there's a lot more attack vectors. How do you think about that?
Yeah. I mean, I think there's lots of opportunities both in terms of if you're looking from, you know, firewall perspectives, you know, I think AI could be very good at looking at the types of traffic. I think it can be particularly good in terms of managing DDoS, in terms of, you know, routing and traffic and determining, you know, we use some of that today in terms of, you know, how we route traffic. And I think there's a lot of opportunities to sort of take that same, you know, AI technology that we're using in terms of routing and build that into the security products.
Yeah. For one example, just hypothetically, like we can take an origin and mimic it on the edge and we can rip it down and recreate it every 30 seconds. So if you're someone that's trying to hack, you're going to have a tough time chasing that origin around. We can play games like that on our end, just as a hypothetical example. We haven't really started to roll out these features, but these are the sort of things that we could do in security that our competitors cannot, especially those that ship appliances. We're a software solution.
Great. I think it's a great place to jump off. Ron, Vern, thanks so much for your time. We really appreciate it. Thank you, everyone.