Well, we have a decent crowd here, so might as well get started. I mean i t's getting late in the conference, and as you just said, we have a decent crowd. So, first off, of course, thank you to you both, Todd and Vern, for joining us, Fastly. Format I've been running here, you know, last five minutes or so, I'll open up for questions, so feel free. But otherwise, I got a handful of questions for you guys, and might as well get right into it. You know, I had the luxury of hearing from you guys in May and then also at the Analyst Day in June. And the big theme that you and some of the others in the space are starting to talk about is actually vendor consolidation.
We even saw one vendor kind of exit the space on the lower end. You know, how should we think about vendor consolidation benefiting you guys? You know, what can you guys estimate as the traffic share on average of your major customers as well from a expansionary side? Like, are you starting to see going from, say, 20% of traffic at a customer to 40%? Is there a way to measure that?
Yeah. It's a great question. You know, vendor consolidation, of course, is a two-edged blade. But for us in the space, especially with customers who are using Fastly, it tends to be very good for us. Anyone going through vendor consolidation because they want to shrink the size of their team, managing content delivery or even edge security, they're almost always keeping the performance leader as one of their go-forward vendors, and that tends to always be Fastly. So that's been... It's been good for us. And generally speaking, it's hard to tell if there's gonna be a lot more vendor consolidation. I think we might be most of the way through that, at least in this cycle.
But when I look at the way vendor consolidation has shook out at large accounts, it's like going from four or five vendors, maybe with a couple of niche providers, down to two. So we tend to go from maybe, like, 30% share to 50% share. And that takes time, and it ramps, you know, it ramps slowly, and I don't think we're fully done with that. But, you know, I hope to get to 50% share at those accounts when we do see that kind of vendor consolidation. I'd say we're like, I think, most of the way through, inning 7 or something like that.
Okay. And just keeping on more the media side of your install base, you know, obviously, you have exposure to some of the largest media players out there. And, you know, obviously, Disney Charter resolved their issues, so I won't pester you about that in terms of news flow. But, is there any way that the writers' strike down the road could start to impact you guys as, you know, the shows that get put to streaming start to kind of get delayed or whatever, and so maybe we have less streaming going on next year? Or how are you thinking about that?
You know, I think there's puts and takes to it. There's more interest in on-demand video and live events, I think, in part because of that. That tends to be very good for us at Fastly. Video on demand is the area where performance matters the most, where our platform shines the most. But, I mean, to be honest, we haven't seen any significant effect, and I'm not projecting one. Our customers aren't projecting a significant slowdown in streaming on the media side, and so I'm not sure if that really is gonna trickle through. But even if it did, the prevailing trend is still, I think, just overpowering in this space. There's this transition from broadcast media, cable media, over to streaming media, and it's that transition that's driving the growth in this space, and that's happening whether there's a new season of Game of Thrones spin-offs or not.
Well, I'm looking forward to that Game of Thrones. Don't get me started on Ahsoka. But look, you are exposed to some of the, you know, applications all of us here use, right? You know, it's a bit of a double-edged sword in terms of the exposure, right? You guys disclose your top 10 customers and, you know, there are some that worry about that, and some that say that's actually a good thing. As you guys gave at the analyst day, you know, 19% CAGR for the next couple of years, essentially, how should we think about that top 10 concentration over the longer term? And then I got a follow-up to that.
Yeah. We run top 10 concentration in the high 30s right now. I think we hit 37% last quarter. And again, I think that's good and bad for different reasons. We've got a ton of loyalty in people who are leaning in and paying very close attention to their vendor in the edge space. That's good for us, that those customers are so loyal to Fastly. But in the fullness of time, I do see that success at Fastly will come with drawing that number down, certainly to the low thirties. And I think that comes mostly from more vertical differentiation in our customer base. We have huge opportunities to really diversify the verticals in which we operate with real market leadership. We have amazing positions in media and entertainment, e-commerce, high tech, but we've got amazing, I think, opportunity, especially right now in brick-and-mortar retail, travel, leisure, healthcare.
When I look at the new logo flow, you know, to me, it looks really healthy because I see so much representation there. We have a new CMO, and we're driving a real focus on new customer acquisition in those verticals.
Got it. And my follow-up there is actually on the security side, right? Probably one of the faster parts of your business right now is on security. Within those largest accounts, not necessarily anyone specifically, but generally speaking, you know, how should we think about that penetration on the security business there? What's the white space opportunity, and why does actually Signal Sciences have the right to win within that?
Yeah, we like Signal Sciences is now Fastly Security. The Fastly Security engine, I think largely it has the right to win for two primary reasons: Signal Sciences technology, which is next-gen web application firewall, next-gen WAF. That's best-in-class security tech, top to bottom. If we have a real security-driven efficacy, security efficacy-driven deal, we tend to do very, very well in those bakeoffs, and that helps us. That's a real right to win there with acquiring so much real, deep security expertise. But the second one is platform, to be honest. The power of the platform is phenomenal, and we have ported that Signal Sciences intellectual property onto Fastly's Edge.
We're in the process of moving the whole management plane into one unified dashboard, one unified management plane, both on the application developer side with a unified API and on the configuration and management side. And when that is finished, and we're not there yet, we have a real opportunity to run a very low-friction land and expand sales motion. And that's when the real platform advantage comes because all of our content delivery customers who are comfortable with our platform, are experts on our platform, when they are ready to make their next security decision, we are effectively already an incumbent. They're already on the platform, they're already comfortable with it. And that's why, you know, to be honest, we are pushing on platform unification probably harder than in any other end project at the company right now.
Yeah, and I think at the end, unless you said that'd be done by year-end or something like that, or it's one of the goals. But maybe switching to the go-to-market side a little bit. You released the new subscription packaging, but very early. I'm sure it's a small part of the business today, if that. Should we think about that expanding outside of sort of the web delivery business, or is that solely gonna be web delivery? How's the early feedback been?
So we had a version of the packages that came with Signal Sciences, to be honest. Their standard go-to-market model was in part what inspired the way we crafted the packages. They're a SaaS-based model. It's a no overages model. It's a very clean onboarding and simple kind of selection. Three tiers, you know, good, better, best, and for that reason, a much easier sales motion. And what we've done is launch that for both compute and of course, for delivery. So now we have the opportunity to sell and onboard customers at a lower friction motion for delivery and compute, just like security. We also tweaked the security packages, so they're all the same, same platform, same model, in the hopes that that'll be easier for our partner, our new partner program to operate with.
We've had good early success. I look at the new customer acquisition flow, in which we do have a decent percentage on packaging in at least one area, and that, to me, seems like a really good signal. I believe there's enormous opportunity to lower the friction of new customer onboarding using the packages. And then if over time, that customer becomes, you know, it becomes so strategic that they wanna lean in and get into, you know, full utility-based billing, a more sophisticated negotiation, great! That's great for us over time. But it just is an amazing way to ease the new customer acquisition, and that's where we're seeing it have the most impact. So, so far, it's been great. And we do plan to release packages for observability as well.
Gotcha. Before getting to some of the observability and security stuff, just wrapping up on go-to-market, remind us how your reps are actually incentivized here. You know, is it completely net new business, no matter delivery, security, or just a new person in the door, essentially? Or, you know, really the core of my question here is, are they compensated on cross-sell at this point or not?
Yeah. Yeah, it's a great question because the, the Signal Sciences acquisition really drove the question of whether we were going to unify the go-to-market, which we did. So we have a single unified sales team, a very small architecture specialist group for security, same on compute, but, you know, 95%, 99% , let's say 95% or more of those resources are selling the whole portfolio, and they're comped on new business, but that includes the cross-sell, for sure. So if they take a security customer, whether it's Fastly Security or Signal Sciences, and they cross-sell a delivery deal, they get paid on that. If they content delivery to security, security, compute, any cross-sell, they're paid on, and of course, new logos, they're paid on.
But I do run a specific incentive on new logo acquisition, and that is really key to the business, right? 'Cause it's that new logo winning that drives the cross-sell for the years to come. And that's how the direct sellers are comped. I will say that we have pivoted. We've got a new CMO in seat, and we've really pivoted our marketing focus to new logo acquisition, like 90% focus on new logo acquisition. We have to drive that. That helps us drive the vertical differentiation motion. It helps us penetrate more of the commercial accounts where we have strong lighthouse accounts and just need to get into the bigger chunks of the space. And so, you know, we work really hard at incentivizing new logos, but we comp on new business across the board. Every part of the portfolio is weighted equally.
Got it. And you brought up observability, and I wanted to go there anyways. And this is, I believe, actually one of your first kind of announcements as CEO, not too long ago, was getting into this space. But when we start to hear observability, you know, the audience here will probably think, you know, like Datadog, Splunk, so forth, right? So, what is it that Fastly is doing in full stack observability? Because I believe at the Analyst Day, you were talking about a little bit different approach, a different way to do it.
Yeah. I think the key here is very similar strategy to our compute strategy. So in compute, we focus on edge compute. We don't compete with AWS or GCP for, you know, core cloud infrastructure compute. We compete for edge workloads, workloads where the edge matters, where low latency, high responsive, dynamic content matters. That's the use cases around compute, you know, whether it's personalization or any type of direct customer engagement, customer interaction, et cetera. And in that way, we're very clear, like, we partner deeply with the central cloud providers, and we do everything we can not to compete with them, but to have multi-cloud architecture. And that's been very successful for us. Same exact thing in observability. We don't compete against Datadog or Splunk or even like a Dynatrace or New Relic. We partner with those folks.
Our strategy here is edge observability. We have a unique vantage point. We have highly connected POPs as close to the user as possible. We can see connectivity stats and latency stats from that vantage point that, you know, is incredibly important signal for a full-stack observability solution, like a Datadog or even a Splunk, and that's the role we play. We sell products here like Edge Observer, User or Client Observer, that give you these direct signals into a full-stack observability solution, and again, we partner with those folks, we don't compete. By leveraging that vantage point, we also kind of sell to our most important buyer, right? Platform engineers that are trying to drive the most reliable system in the world. You want the most visibility you can get, and the edge gives very unique vantage point. That's what we focus on.
Makes sense. Then you're bringing up edge here, so we're just gonna go right to, you know, this little topic called AI. You know, AI inference workloads, there's a thought process, like you guys could capture some of those workloads essentially and have them run on top of the Fastly network, right? So, I guess, how are you thinking about that AI edge opportunity? Why won't those just run, you know, via the hyperscalers kind of extended infrastructure?
Yeah, that's this is the core strategy discussion in our compute business all day long right now. And again, it's about that partnership. So when we're looking at the inference-based workloads or really any AI workload, it's important to deploy in the correct model. We don't train those AI inference models on the Fastly platform. That reverse propagation, it's incredibly resource intensive and incredibly storage intensive. That's gonna happen in the central cloud, and it should. It's gonna happen in the hyperscalers, and we partner with them to make sure our customers can do that, and then deploy the trained model at the edge, where latency matters.
And when you think about, like, large language modeling for customer support, customer interaction, when latency matters, when you're effectively running a Turing test, edge, the ability to run that forward propagation at the edge is incredibly important. That's where our customers are focused on using edge, and it's why our goal is to be... We're not gonna be the purveyor of AI technology. We're gonna be the best platform on which to run your model, at the edge. And the use cases that seem to matter right now to our customers are direct customer engagement, largely for customer support, that's large language modeling, but also any type of recommendation and personalization, whether it's content recommendation on the media side, product recommendation in e-commerce and retail. That's an amazing use case. Takes very little edge data for the personalization.
It has a relatively contained matrix for the inference model, and they can run that at the edge.
So maybe thinking about LLMs here, both from an internal and external basis. You know, one, how are you trying to use LLMs internally or AI internally for, you know, cost optimizations or development? And additionally, like, how should we think about external products beyond kind of what we've talked about here, leveraging LLMs?
Yeah. We already deploy an AI algorithm called Autopilot to do traffic engineering. That's the biggest leverage point for us internally. By engineering our network and doing real-time traffic engineering, it allows us to draw down our costs and use our network and our infrastructure as efficiently as possible. We've seen enormous improvements there. In fact, I think we published a note on the Super Bowl activity from last year on how the Autopilot algorithms were able to make almost every distribution decision in real time without any human interaction. And the real-time nature matters because milliseconds matter, right? Human beings are not gonna be able to do that in real time. So every egress decision was made by that algorithm in real time.
and I think we have a huge opportunity to do so much more of that. Video on demand is just the kind of tip of the spear. All, network engineering at Fastly has the opportunity to benefit from that. That's not strictly LLM, you know, AI , and even inference AI. But as far as the language side goes, you know, we might start to use that in the customer support side of the house, but, you know, our support tickets are relatively sophisticated. We might be a little ways out there. It's also not a huge point of leverage for me. If I can save 1% on network engineering, you know, that's probably better than 25% in customer support language modeling benefit.
I don't know if if Ron were in the room, he might say, like, "Ooh, bots are saving money." But with Ron not in the room... Look, you're not gonna guide to 2024 here, and I know that. How should we think about the puts and takes for 2024 across, you know, all, all parts of your business: delivery, security, observability, the new packages? Just walk us through kind of the puts and takes.
For 2024?
For 2024.
Yeah, I think we have a real opportunity to drive the cross sell. And since our existing business is largely content delivery, I think there's a lot of upside on security. We're gonna see that with platform unification coming live, early in the year and really being able to drive that. We're already seeing some benefits just in cross-training and, driving a little bit more maturity in our security sales motion in Europe. And that's important because Signal Sciences, at time of acquisition, was almost entirely a U.S.-based company, so it takes a little time to normalize that expertise. In 2024, I think the revenue, the top line, bottom line, is gonna be driven on security and, content delivery. I'm still tracking compute and observability as incubation programs. We'll see, we'll see a lift on compute, revenue. The increases there are great.
As a percentage, they're enormous, but it's really customer acquisition that's gonna matter 'cause it's 2025 and 2026 when we're really gonna be leaning on compute and AI workloads as really the growth engine.
Got it. Just roughly five minutes left, so happy to take a question if there's one at the time. Well, maybe while we're thinking about it, you guys did actually a little bit of an acquisition here in the quarter to Domainr. Just refresh us what that gives you, and where else you feel there might be, you know, a need or a hole in the portfolio at this point, and why you chose not to necessarily go down the, you know, zero trust route that you're seeing some of your competitors do?
Yeah. Domainr is super interesting tech. It gives an entire DNS, like, name management, DNS management portfolio, 100% programmatic. You don't have to-- Like, everything that's slow and painful about managing domains is pure, purely automated through APIs, thanks to that technology. And the moat on that is also super interesting. It's the contractual agreements with the regulatory agencies that Domainr has that makes that possible, and without those, you can't do it, and that's slow to, slow to get. So, I think it's just an amazing tuck-in for us. It's part of our network service product line. It's 100% within our buyer and sales motion. It's got great go-to-market synergy. It's an amazing tech, and the team's come on board super smooth, and the price was right.
It's like picture-perfect tuck-in, man, I love it. As far as M&A strategy going forward, you know, we would look at strategic tuck-ins like that, that bolster our product lines, but I feel pretty good with the four product lines we have. Our CTO organization continues to incubate new, workflow on the AI side, new offerings in storage and regulatory compliance. I don't perceive any real significant need for a kind of more strategic M&A. And to be honest, look, we're still digesting the Signal Sciences acquisition, and I hope to be most of the way through that next year, but, we, we feel pretty good about the tech we're delivering out of our team right now.
Yeah.
So obviously, compute is very early, still needs to talk about some sides of things. But as we think about, you know, you kind of have to make this balance between getting testers and developers on the platform, but then making the decision of starting to monetize that. So can you talk about how you decide, you know, when you start monetizing and how you kind of push more down to the platform and then switch those sides?
Yeah, I mean-
Repeat the question. It's regarding Compute Edge.
Oh, sorry, because it's not being-
Yeah.
Yeah. Yeah.
The question is, when do we start thinking about and pushing on the monetization of the Compute product line and not just customer acquisition? Yeah, it's a good question. You should know, like, one of the reasons that I run a product line like Compute in an incubation motion is to get the monetization right. So it's not that we're not monetizing it, of course, we are. And we've got metering models that, you know, we're experimenting with and we think are correct, but even recently, we've simplified that metering model to lower the sales friction, give more customer confidence, et cetera. And, you know, as you change those models, you change the way the monetization shaped.
But, you know, look, if I wanted to, I couldn't get my product managers to look away from their revenue growth and bookings growth. That's what they want to be measured on. And it, you know, it's been good, but the long-term health, getting Compute into the growth motion, where every sales rep is competent, every partner can close that deal, every part of our go-to-market engine is firing on all cylinders, you need a critical mass of customers, and you need those customers to be really all agreeing that the pricing model works for them, that the value proposition can be simply understood and simply rendered by our go-to-market teams, that the lighthouse accounts are in place, et cetera. And that's what we're focused on. That's why I considered incubation.
We're gonna start, hopefully next year, but really, as soon as I can, we're gonna be starting reporting core business, growth business, and incubation separately, and you'll be able to get a view of, you know, what that incubation business looks like. Yeah, there is revenue there. We are tracking it for sure. Yeah.
Maybe just with the last minute here, you know, how do we keep the acceleration going? You know, keeping that line of sight to 19%, is it, is it gonna be more delivery share gains? It sounds like you're pressing on that and security, you know, 2024 a little bit, and 2025, and then the incubation products, 2025, 2026, and anything to call out on capital usage here.
Yeah, I think in 2024, it's gonna be. Like, we're gonna drive that growth in security and delivery and, you know, not necessarily with the offerings we have, but by bolstering those offerings. Our bot mitigation on the security side is in beta. It's gonna, it's gonna launch very soon for full availability. Huge opportunity to upsell there. Content delivery with the Domainr acquisition, with a lot of the regulatory compliance stuff. We've got a huge opportunity to grow that business through portfolio expansion as well. We've got our DDoS security managed service offering has done extremely well. It's very early days. We're looking forward to that stuff. Security proxy work is an expansion opportunity for us. So we've got tons of opportunity for expansion in those existing areas.
And then, just like you said, as we look at 2025 and 2026, AI, compute, advanced edge storage, there's tons of portfolio expansion.
Got it. Well, we're up on time, so appreciate your time and you guys making the trip out here to Nashville, and hopefully, we have you here again. Thanks, everybody.
Thank you.