It is. It's a big table.
Hello, everyone, and thank you for joining us for our Growth Stock Conference and today's session with Fastly. My name is Jonathan Ho, and I'm the cybersecurity analyst for William Blair and Company. Our speakers today are Todd Nightingale, who's the CEO of Fastly, and Ron Kisling, who's the CFO. Before we begin, I'm required to inform you that a complete list of research disclosures is available at our website at www.williamblair.com. With that, I'll hand it over to Todd to do a brief overview of the company, and from there we'll go into fireside chat. So, Todd.
Great. Thank you so much. I'm not meant to stand up there and do this.
No, you can sit.
Thank you. Great. Appreciate everybody's time here, and thanks for the focus on Fastly. Fastly was born as a content delivery company, but over the past few years really has evolved into an edge delivery platform. That edge platform is capable of really delivering best-in-class user experience. We partner with our customer base, who is building apps and websites and streaming services to deliver the fastest, safest, and most engaging web experiences possible. It is that partnership with our customer base that delivers that outcome of fast, safe, and engaging experiences on the web. In order to do so, we focus on delivering best-in-class edge technology. That includes content delivery, the highest-performing content delivery service in the world, edge security, including DDoS, bot mitigation, and web application firewall services.
Edge compute is part of our emerging platform, allowing bespoke serverless compute at the edge, which allows for the best personalization on the web and, of course, observability. Increasingly, our customer base is focused on building the most reliable sites in the world and building an observability feature set capable of giving them the visibility they need to do that is really part and parcel to this. This transition from a point product in the CDN space to a full edge platform has had a host of really significant improvements to the Fastly business model. It's helped us diversify our customer base. We are now engaged at far more accounts and a far more diversified set of enterprise verticals, not just media and publishing, but increasingly high tech and retail, e-commerce, hospitality, health care, et cetera.
It’s also helped us diversify our revenue into these other spaces so that we’re not as dependent on the CDN space. It’s helped us drive profitability across the board. We’ll talk more about how those numbers are shaping up here as we see it through the rest of this year and beyond. Overall, this platform strategy has let us really double down on the value proposition that made Fastly compelling and important from the beginning, which was performance and programmability, which is really the value proposition for the end user and the customer, and really expand the platform value proposition, which allows us to drive larger wallet share at our customers and reach new customers in new verticals.
Excellent. Thanks for that overview. Maybe just to kind of kick things off a little bit, I think you talked a little bit about diversification of your customer base. The company has done a fantastic job of being able to move away from some customer concentration challenges in a relatively short amount of time. Can you help us maybe understand some of the actions that you've taken and how you were able to achieve this?
Yeah. We had 40% revenue concentration in our top 10 accounts at the end of 2023. We saw that drop all the way to 32% with really a change in the buying motion in large media streaming accounts at the beginning of last year. That whole transition, we were able to post positive growth every single quarter. Obviously, that was a very rapid shift in that customer concentration. We were able to post healthy growth outside that top 10 pretty reliably. In fact, last quarter, we posted 17% growth outside the top 10. That has really been based on this platform expansion being, I think, a really stronger and stronger solution for enterprise application development team, platform engineering teams that are looking for one complete edge solution for both, most importantly, for security and delivery.
I think it's also built a strong foundation upon which, as we start to see the recovery in those large media accounts, we can really build some very strong growth results for next year.
Excellent. Excellent. Ron, with maybe less customer concentration, what could this mean for Fastly's margin potential and potential for retention improvement over time?
Yeah. I mean, I think one of the things you see straight outside by having a robust set of customers in that space is more predictability. It's seen as a more reliable space. There's a little less volatility, particularly than any one customer can have. We also see really good ramp from customers when we sign them up over time. As we accelerate the number of new customers into that category, it can be a big contributor to growth. I think when you look across margin, I think what really drives our margins is a healthy mix of customers, a broad set of enterprise customers. We will talk about this, I think, in a minute, but also more attached in terms of customers who use more of our products are all things that can drive improvements in our gross margin and increase stickiness of those customers.
Excellent. Excellent. When we look at the potential for the company to return to double-digit growth after maybe a challenging period over the past few years, we've seen the exit of Edgio, StackPath, and Lumen. These are big competitors or small competitors that have exited the market. What underpins sort of your confidence that you'll be able to see these trends improve or maybe reverse over time?
Yeah. StackPath and Lumen might have been relatively small, but Edgio did carry quite a bit of traffic, quite a decent amount of revenue. I think the biggest difference with those exits in the space is the pricing environment. I think what drove their exit was that they just had very high customer concentration, and they had really almost all their revenue coming in through that content delivery space. For us, with the diversification of our business on the portfolio side and the customer side, we have a stronger foundation. Now, with their exits, a pricing environment that I think is advantageous and will be for a few years, which will help us drive gross margins. It'll help us drive customer retention and loyalty and give us that opportunity to expand in every single account to a much broader feature set and wallet share.
Excellent. I mean, if we dig into this notion of the portfolio concept a little bit more, one of the more exciting things is the developments in cybersecurity around your business. Can you talk a little bit about how this has been a critical pivot point and maybe what excites you about the company's opportunity to build out that portfolio and to see that mature over the past few quarters?
Yeah. In 2024, we invested a ton in R&D on the security side. I think the portfolio expansion shows that. We launched a bot mitigation product. We launched a DDoS solution that works out of the box, turnkey DDoS solution. We have expanded the feature set within our web application firewall and in bot and DDoS as well. That investment has all been driven on customer engagement and co-development with our customers who are looking for their edge platform to deliver these solutions. We know that there is customer demand and traction there. Having made that investment, we are really shifting to the go-to-market. We have brought on go-to-market leadership, both on the sales and the marketing side with strong security background. They have brought in good leadership under them. We have sharpened up our comp plans to be focused on this side of cross-sell and portfolio expansion motion.
We can clearly see the demand in the market, and right now are really focused on rising to that challenge. We feel really comfortable with the portfolio. We've had great success on the early data on the newly launched products, both Bot Mitigation and DDoS . Obviously, we're investing and feel really confident about how the back half of the year in 2026 will.
Yeah. Maybe if we can dig in a little bit into the differentiation of these products and how they sort of stand out in the market a little bit.
On the WAF side, this core intellectual property came from an acquisition of Signal Sciences a few years ago and really delivers best-in-class efficacy on web application. This tends to be the kind of flagship product in the web application API protection space. Because of that high efficacy, the vast majority of our customers run in a fully reactive mode. The system automatically turns on the protections and blocks traffic. It does not just raise alerts and thereby becomes wildly more valuable because the total cost of ownership drops and the actual security protection increases without requiring human verification and human intervention or to actually put the product in blocking mode. On the DDoS side, for years, Fastly has provided best-in-class, highly sophisticated DDoS protection, but we have not always had a turnkey product. Our DDoS protection required highly sophisticated users who were highly engaged on our platform.
Now we're able to deliver that efficacy that we've been known for at large sophisticated accounts to a far, far larger customer base. To be honest, DDoS is a solution that should be delivered on the same platform as CDN. There is enormous attach value here. If you're offering an edge DDoS protection service away in a different cloud from your CDN, you're going to be ping-ponging your traffic between clouds, which is highly inefficient. The attach is just incredibly valuable. On the bot mitigation side, I really believe this is going to be a question of speed of innovation. We launched the bot solution. We enhanced that solution significantly. In Q1, we added protection for AI scraping bot detection. It's a cat-and-mouse game.
This is a market where there is an adversary who's constantly innovating in terms of how bots try to beat bot management and bot detection. With our kind of team and our focus in this space, we're building more and more sophisticated solutions. It's going to be a game of speed of innovation, for sure.
Excellent. Ron, how do we think about the business model for cybersecurity? Are these revenue streams potentially higher margin, more defensible over time? Talk to us about some of the advantages here.
I think it's really important in a couple of ways. One, as we talked about earlier, more and more companies are buying a platform. They want to buy delivery that has the security they need. I think that's why you've seen some of the sole delivery providers exit the market. Entering this year with a more complete security portfolio gives us an opportunity to increase our security revenue. Importantly, a couple of things. I think, one, we find that customers who have more of our products are much stickier in terms of higher retention. As we sell more of the security portfolio, we create longer-term customers. From a margin perspective, our security runs on the same platform that our delivery runs on and our compute.
The more products we're able to sell to our customers, the more we raise our gross margin. Today, we currently have just under half of our customers running two of our products. With the expansion of the security portfolio, if we can bring that number up and increase it to two or three, that is a significant contributor to improving our gross margins.
Excellent. I don't think we can have any of these fireside chats without a discussion on AI. And so.
AI is shocking.
There's been a lot of talk about AI in the marketplace. You touched on this a little bit, but there's been all this controversy over the AI and whether they should have access to data, privacy, monetization. What does that opportunity look like for Fastly to participate in? How does the AI revolution, maybe on the agentic side, start to play a role as well?
Yeah. I mean, I think there's a lot of momentum here on thinking about how the training sets are built. And that's something that's near and dear to our customers on both sides of the house. The AI bot detection is about detecting bots that are scraping data in order to train your competitors' models and whatnot, competitive pricing or competitive product messaging or whatever it might be. We are dedicated to building a safe internet where people can present their data without having to worry about or without the ability to protect against all of their data being scraped away by their competition. On the other side, there's a real opportunity here for us to serve those training data sets appropriately from people who want to sell them. In fact, we launched object storage functionality into our compute portfolio in Q4.
We have already seen a large strategic customer come in and start distributing their training sets that they are selling for AI training through that platform and leveraging the performance of Fastly, which is incredibly important for this use case. They need to deploy these training sets to the models with extremely high bandwidth in order to keep their customers' models as they continue to upgrade from one version of the model to the next to retrain each step of the way. They also need the control to be sure that folks are only using the data set that they have paid for and licensed, et cetera. That use case is a super powerful use case. It is an important place that we can play on the training side of the house.
On the forward propagation side, on the actual use of these models, in Q4, we also brought our AI Accelerator to market, fully generally available across our platform. That allows folks who are using LLMs to both lower their costs of their AI costs in their central clouds and GPU costs and also provide far faster response times to folks who are running LLMs like for customer service or customer engagement on B2C sites. That is certainly an interesting use case. It has the benefit that I think Fastly can deliver in terms of delivering better performance and better TCO.
Excellent. Excellent. How do you think about making more incremental investments here in terms of compute or infrastructure to prepare for things like inference-based AI? What does that look like? What does the opportunity set look like as this becomes more and more of a reality?
Yeah. As far as the CapEx spend stuff, I'll let Ron add color there. We are always evolving the infrastructure that we're deploying and tracking the use cases that our customers want to deploy. Importantly, we have a fully software-defined infrastructure. We don't have a myriad of appliances and custom hardware in our clouds. All of our POPs are built as simply as possible from a hardware perspective so that our software can be as flexible as possible. Because of that, we've found the ability to deploy this feature set and functionality within our regular spend envelope. I do imagine that that will continue.
Yeah. I think if you look at that spend envelope, what we've said is generally we're very efficient on CapEx. We set somewhere between 8%-10% of revenue, including capitalized internal software.
Our external CapEx assumed maybe is half of that. We do think in the medium to long- term, maybe that number comes down to 6%-8%. Within that 9%-10%, I think that captures the ability to bring in any of the hardware that we would need to support AI.
Excellent. Excellent. Yeah. In terms of the edge compute market, when are we going to see sort of broader adoption of edge compute? And particularly with your investments in WebAssembly, what does that market opportunity look like?
Yeah. Serverless compute, we use WASM, we call it WebAssembly technology to deliver that. The more that people start to move really dynamic content and dynamic functionality onto the edge platform, the more that tends to accelerate. We saw pretty good growth this past quarter in that segment. I think it was 67%, but I don't want to get that number. I think we have an opportunity to drive that kind of very high growth rate for quite a while because we're seeing adoption on the AI use cases. We're seeing more and more dynamic services being built on the platform. As more folks are looking at Fastly beyond CDN, as they're looking at Fastly as an edge platform, it becomes more and more attractive for their development team to leverage that type of service because they're already looking beyond CDN.
It is really part of this platform story. There is a buying motion here that once they are adopted onto the platform beyond a single point product, that it is easier to expand from one to the next.
Makes a ton of sense. I mean, we'll open up the questions for the audience in a little bit. Go ahead and get prepared. What maybe has to happen to accelerate the adoption of that edge compute? What are sort of the next steps or what sort of killer applications maybe have to take place for that?
Yeah. I would say from us, I think continued vigilance in being the highest performing solution in the market. We pride ourselves on that. This is kind of the most programmable way. Customers can actually deploy their own code into the Fastly platform and run it on every server around the world. It is a very, very powerful solution. Performance tends to be the reason people do this. They traditionally have served static content from the edge, and everything dynamic has had to go back to their core, their origin servers. By pushing it to the edge, we can radically increase the performance and finally let them provide highly personalized service to all of us without having to compromise how performant that is.
I think from the customer base, we're really seeing this trend, which is more and more verticals focusing on the digital experience that they are providing their customers. I was just talking about this earlier today. Every airline has a loyalty app that probably everyone in this room uses every day. How personalized that app is to you and how performant it is, how quickly you get the results to your searches and your ticket and pulling up your boarding pass, et cetera, it matters deeply. We're seeing more and more traditional enterprises like that really investing in this space. We've seen a lot of success in the airline space, for example. There are a lot of verticals like that that are focusing deeply on building a sophisticated and differentiated digital experience. It is that transition that is driving growth into our compute.
Yeah. Maybe one last one on my side on that last point. When we think about potentially the applications becoming more sophisticated, you have AI coming on board. Does the world sort of start to move more and more into Fastly's favor over more traditional delivery methods and edge compute? Can you talk to maybe what has to happen there?
As people start to focus more deeply on how good the user experience is, I think that shifts to Fastly's favor. We say we want to make the internet a better place where every experience is fast, safe, and engaging. Those websites and apps, streaming services, whatever is being built, the outcome that matters is the user experience, how all of us experience that app, whether it's video streaming or airline or e-commerce. The more that teams realize, the more that these buyers are focused on the fact that the user experience they deliver, how responsive it is, how safe it is, how reliable it is, that that determines their bottom line, that it's more directly tied to their profits and their outcomes, I think the more likely they are to choose Fastly and invest in the Fastly platform.
Excellent. Go ahead.
[audio distortion]
Can you repeat the question? Yeah.
Sure. We launched the AI Accelerator in Q4 last year. Who are we replacing in that space? Who is the incumbent? That's the question. This is an emerging space. The central cloud, the AI clouds, the large incumbents in the space, some of them even in their technical spec have a proxy architecture that allows for this or supports this kind of motion. They all support it, but some of them even are specifically articulated. There hasn't been someone in this space. I mean, this space has only really existed for a couple of years. Our AI Accelerator would compete against other solutions referring to themselves as like an AI proxy, of which there's just a couple starting to emerge. There's no real incumbent in that space. We have focused our functionality on acceleration for the end user and TCO for the customer.
There are proxies that are focused on tracking of developer adoption or platform migration from one backend to the next, which is all feature set we're considering adding. But it's an.
Speaking of AI, is there an opportunity around MCP or multi-context protocols so the AIs will leverage services from each other? I can imagine there's a connectivity component there as well. Or too early?
It's a good question. The large backend, like shared clouds like the OpenAI or the Gemini infrastructure, is so well connected already, I think probably not an opportunity for us. As people are running their own models on their own infrastructure and their codes, then I think we do have an opportunity, especially as they're going to need to distribute training sets around the world, and they're going to want to control how much egress they're really buying.
Other questions? All right. I guess we'll give you a few minutes back.
Amazing. Thank you so much. I really appreciate everybody's interest in Fastly. Thank you.
Thank you.