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

We've seen a lot of inbound interest from our customers on what we call co-innovating on our compute platform, which we put in our other revenue category. What's happened is I think there's just now in trying to work with LLM to run agenetic workloads through an orchestration layer. We've been involved with conversations along those lines with some customers. We're also seeing a lot of one-off instances of where we're working in conjunction with someone running an LLM to, in some instances, use our storage product to sort out. I think even last year, we talked about the images with Shutterstock, things along those lines, where we've had interesting programs that have come into play. Where we're seeing the biggest vector of near-term growth in AI right now has been in what we would call like bot management. We have a product called ContentGuard wrapped around that. On our prepared comments on the call, we talked about 1 of our wins we can talk about publicly, which was with a French media organization, Lamonde. They're running basically -- this product called Content Guard, which basically can detect when agents come in to scrape their network or it's perhaps an agent working on behalf of a benevolent organization around where they already have a relationship. They can use this product to basically filter whether or not they want to reject the agent or take it in and also have motion around that. So it's sort of a new era of, if you will, with agents coming in and instead of them being mostly harmful, you can start to get a monetization pathway around that. We're excited about that program. We're seeing a lot more interest around those sort of developments on the bot side with a genetic traffic in our customer base. So Security is probably the most near-term vector. Of course, we see a little bit of that happening in just planal delivery, where we're seeing Agentic traffic come into our network. And then lastly, on the compute side, as I said, there are a lot of discussions around running compute workloads that are agentic. We are seeing some business there, but it's a little early to tell where there's real momentum and like where we'll be in, say, a year from now, it's still very early innings.

In the long run, do you think that there is -- because I totally appreciate the fact that a lot of what we would call, I guess, like a genetic use of the Internet, right, where an agent visits 4,000 sites instead of 4, right, for a human. That traffic though, is not necessarily what customers would consider high-quality traffic that they are willing to pay for. So I'm just curious whether there's been any customer feedback on that only on the content side, bot management and ContentGuard, I completely understand. But anything on the delivery side where customers are seeing an influx in traffic and saying, this isn't exactly what I'd like to be paying for.

Like so it's interesting because the way our delivery products work, we charge a lot of customers based on gigabytes delivered plus requests comes in, just has a lot of requests because it's saying quite often. It's usually tends to be lower, like gigabytes of traffic delivered -- our customers, I think they're working on their strategies because I think even the AI traffic, it may not be like the ones that convert right away, but it is still like visibility traffic for them. And so they haven't pushed back too much on like the AI trait's happening. I think what they're trying to really spend time working through is some of the products that we have that helped them figure out like what is the good AI traffic versus the bad AI topic. And that's where both management comes into play. Content Guard, trying to really work with them to like say, okay, let's allow all the good AI topic because they do know that air traffic is not all bad traffic. And so they want exposure. AI is now like the new SEO, right? -- to get that visibility and like that brand recognition, I think they know that they have to like have an AI strategy that allows them to really get at top of like top of funnel for with AI tropic.

AI is the new SEO that is something that other people in my coverage know all too well. Okay. So a little technical that's right. Well, Rich, we'll ask you a question while we get this figured out with Vern. I think do the different segments of the revenue structure are contributing more to the incremental margin. And as 1 segment might outgrow the other in the next year or 2, what should we expect for incremental margin?

Sure. So what Jackson is referring to is the way we've guided around how we think about like the additional operating leverage we should get as a company. And so what we've talked about is a 65% to 80% target for gross margin flow-through which means for every incremental dollar of revenue we get, we should be flowing through 65% to 80% on the gross profit side. And then on the operating leverage side, we talk about 25% to 40% on -- on a last 12-month basis, like we've far surpassed that, right? So the last 12 months for gross profit flow-through, we've had 96%, which is far ahead of the $65 million to $80 million and then on the operating leverage side, I think we were in 79% operating leverage flow through versus the 25% to 40%. I would say the product mix and the synergies that we see having complementary products is very helpful. But I also say that a lot of the flow-through is also on the upside that we see in the business, right? We build our plans with Q2 results, we had a $10.4 million beat relative to the midpoint of our guide and having that big of a beat allowed us to really flow through a lot. If you look at the last few quarters, we just had pretty good beat around like how we've done. And so those beats actually give us a lot of additional operating leverage that surpasses the guide that we've kind of given from 25% to 40% but I would say that like having synergistic products, having 1 sales organization that go to customers and bring the full suite of products really makes us much more efficient and effective as a company, we operate 1 network. And so when we operate 1 network, like a few -- our peers may have multiple networks they have to support. They may have a network for delivery services, and they may have a different network for security and in some cases, they may have like 3 networks for delivery services. With Fastly, we have one network, and that 1 network supports all of our products. I eSignal Sciences acquisition -- they are running on our network. And like they are on our infrastructure like the same servers that are providing delivery services are also providing the web application firewall from Signal Sciences acquisition. That just makes us very efficient and very capital -- less capital intensive than our peers, right? If you look at our capital guide from an infrastructure perspective, our capital guide is below what our peers are. And we're supporting a lot of traffic relative to -- so I would say that it's a combination of like the upside, kind of the synergistic products, but also having a very like single network that kind of provides everything, all the products.

One more question, and then I will come to the audience, so be thinking about it, Aaron. You guys are decidedly not in the GPU game. At this point, do you want to tell us why?

Yes. So I think for Jackson's question, some of our competitors have built out big slugs of GPU capacity and have contracts in place with some of the on the behalf of doing that. We are not in the business of what we would describe as building sort of a neo cloud. We still believe our best positioning and our best value prop to our customers just to be at the very edge of the cloud, having a very open standard, open cloud platform that can cater to all different levels of innovation. We continue to drive on that philosophy. And I think just not to pick on our competitors, but I think if you peel the onion a little bit on the investment that's required to do that and the return on that investment, it's a tougher business model, I think, than what we at fastly would want to tackle. I'll speak for Richard but I don't know if you would want to sign off on these either. But we do see, in our view, enough business on what we're doing right now to have a nice foothold in strategic in sort of this new AI landscape without having to go out and deploy a lot of capital to build like a GPO -- excuse me, GPU Neo cloud to do.

And then if you think about like our compute product and we talked about compute being the majority of our other, our other was about $8 million last quarter, and the compute resources that are needed, we actually can service those with CPUs. So if you think about our network and what we have, I think previously, we were much more IO bound and not CPU bound. And so when we work with our customers on their compute needs, a lot of times, we've come with our whole platform infrastructure viewpoint and say like, hey, what are you guys trying to do? What can -- how we can help. And what we find is that our existing CPUs in the edge cloud can do what they need. And so for us, it's like much more efficient, cost-effective, ROI positive from a capital perspective to leverage the existing network we have, where we might have stranded CPU capacity. And we can leverage that then go into like the GPU side. Aidan, go ahead.

Yes. Just specifically on the CapEx refreshes, what are going to be the impacts of the [indiscernible] coming years.

Operating one network, actually helps us quite a bit [indiscernible] for the new AI and genetic workloads in terms of [Audio Gap] network router, storage, memory, hardware upgrades, what's going necessary, and you've got thousands of servers out there. What's the average life and when you see the and how will that affect the CapEx? I'll just summarize the question. We're talking about CapEx refreshes, what are going to be the impacts of those refreshes maybe in the coming years? And then any impact from pricing that we've seen on components -- good question, Aaron. Thank you for that question. I would say that from a -- I talked about operating in network. I think operating in network actually helps us quite a bit. And our 1 network that we have is a very software-configured network, right? Like it's -- we give -- we're using a lot of like powerful off-the-shelf products that tend to like really be powered by the software we have on hand that makes us much more efficient from a network perspective. I think just earlier this year, we went from a kind of a 5-year CapEx depreciation cycle to 6 years because we do see that the equipment that we have in place actually have been lasting longer and actually gets us closer to our peers around how they depreciate servers on that front. I would say that our -- because of our capital efficiency, like we've been able to use our servers much more efficiently and effectively around a utilization perspective, which is why [Audio Gap] you certainly, our infrastructure and making sure that it can support our customer needs and we build enough capacity in there to really make sure that we can support upside on the revenue. And so most of our infrastructure CapEx that we're spending this year, we're launching new PoPs. I think we talked about in Latin America. We've talked about some in kind of Southeast Asia around getting more international support in regions that we haven't been as strong from a POP perspective. But I think for us, we're much more efficient from a maintenance CapEx perspective, right, because we're using more off-the-shelf servers that are powerful. And so they've been lasting longer -- in terms of like Jackson, you added on to the question around like the component shortage seeing and what does that mean? I think the beauty here is that we watch it pretty constantly. I think for all of 2026, we placed our order in the end 2025. We were able to get a lot of those component purchases in the first half of the year. So what you'll see with our capital spend is that we are front-end loaded because we wanted to -- we saw that component shortage happening early, and we just basically placed the full order in place. And so we're generally pretty good. We've already started placing orders for 2027. But we feel good with the guide that we provided, which is 10% to 12% infrastructure CapEx as a percentage of revenue. And I think that's below what our peers are spending on infrastructure CapEx.

I have a quick follow-up. You mentioned right off the bat, this is your third time as a CFO, your prior CFO experience immediately. I think before Fastly very SaaS-like seat based, right? Like very predictable -- this is your first foray in IA consumption model. Your year in -- how is it going? How should we be thinking about sources of upside in the quarters or in the years in a very different model as CFO is where you came from?

I mean it's completely right. My last 2 -- actually, my last company, we were very SaaS-based. I had revenue predictability that was so easy because before I started the quarter, I knew where it was going to be. I would say that [Audio Gap] we changed a lot over the last 12 to 18 months. I think 1 of the beauties with having Scott here. Scott was also really big and instrumental on getting commits up and so if you look at like our RPO growth, our RPO growth is up 38% year-on-year. Our current RPO is up 44% year-on-year. That additional commit gives us a lot more visibility and shifting from a consumption-based business like excited about the momentum that we're building as a company. I would say the momentum around the product velocity like seeing like the number of launches that we've had and being able to build out that product portfolio on the security side, some of the stuff that we're doing on the compute side. We have a lot of other kind of projects going on in the company. I thought I'm pretty excited about that. I don't want to -- we can't talk about yet. But I would say that we have a number of initiatives we're really excited about. I would say that like bringing in seasoned leaders been really good. We talked about the top 10 growing pretty fast. Like we're very excited about that. We brought in Joan Jenkins, who's our Chief Marketing Officer. She started about 4-5 months ago. And I think being able to build out that brand and being able to like attract new logos beyond like that top 50 or top 100 customers that require performance is really key, too.

All right. Great. We are out of time. Rich Bern, thank you very much. Great start to the conference. Thank you. Thanks, everybody. Thank you.