Fastly, Inc. (FSLY) Earnings Call Transcript

Okay, everybody. I want to welcome all of you. I'm Vern Essi, Vice President of Investor Relations here at Fastly. I want to thank all of you for joining us today, those of you here at NASDAQ MarketSite and also those of you joining us on the webcast. Obviously, we appreciate your interest and time spending on our story and glad you all made here today. A couple of quick housekeeping notes. Our Investor Day presentation is on our IR website. It was posted there about 15 minutes ago. You can go there and download it and follow along with us today. Also, we engaged corporate advisers to help us with this Investor Day, they're going to be conducting a survey. So you get an e-mail after our conference concludes. Please participate in that. We appreciate it. And we also value your input and of course, your support. So this Investor Day is a long time in the making. Last time we did this was about 3 years ago. And I would say we definitely appreciate your commitment and patience as we've evolved over that time. We've seen a lot of changes in our ranks, and change, of course, is never easy, but I think we've had a lot of positive outcomes as a result. Yes, we've grown in revenue and scale, but I think we've also grown as a company in ways that you can't measure financially. We're much more mature. We've been able to tackle the challenges, solve customer problems, and I think put a good foot forward with a high level of confidence. I think you're going to see that today here at our Investor Day. And I spent the earlier part of my career on the sell side, sitting in the same seat as some of you. And I wanted to say that I really respect the equity research craft. I mean no disrespect to the buy side as well. You're all working hard there as well. But I do want to say, like, I really appreciate you all looking into companies, doing your analysis, challenging management team, certainly having differences of opinions. But I do want to say that whatever you want on your ratings, whether it's bullish or bearish, know that you can trust us to always be neutral on how we behave with you, just as neutral as our cloud strategy. I also want to say that we do also like to say that we move fast and we play safe. And on that note, we also want to play by the rules. So safe harbor here is on the screen. For those of you who know it, we encourage you to read the disclosures and risk statements in our SEC filings. I think you all know how this works. Let me turn to the agenda today. I'm going to start off with Kip Compton, our CEO. He's going to kick off our strategic positioning and growth opportunity and walk through that. He'll then host a fireside chat with our Founder and CTO, Artur Bergman. Our Chief Product Officer, Kelly Shortridge, is going to discuss our product platform and opportunities. We're then going to host a Q&A session and then the first half will conclude with a small break. Our President and go-to-market, Scott Lovett will kick off the second half. He's going to talk about his go-to-market playbook as well as his commercial engine. Our CFO, Rich Wong will then provide a financial overview and target model. And then Kip is going to take the stage again to close things out, and then we're going to open up again to Q&A. We anticipate wrapping all this up around 4:00 p.m. Eastern. For those of you here at NASDAQ, you're welcome to join us for a small reception thereafter. And as the day goes on, and this conference goes on, there'll be plenty of opportunities for you to find reasons to love Fastly. But maybe you should hear from our customers first. [Presentation]

That's great. Okay. So I'd like to welcome Kip Compton to take the stage. Thank you.

Thanks, Vern. And we're proud of a lot of things you do it Fastly, but it's always great to hear straight from customers. So first of all, thank you for being here today or for joining us online. As Vern said, this Investor Day has been a little while in the making. We've been looking forward to today and having an opportunity to share with you where we've been, but perhaps more importantly, where we're going as a company. Certainly, the business has evolved been really important ways over the last 3 years that we'll outline today. I think you'll also have an opportunity to meet our leadership team a bunch of the key members of our leadership team are here today. You'll be able to hear from them directly in presentations as well as Q&A. And that's one of the other big changes of Fastly is the leadership team evolution over the last 3 years. There's a few things that you're going to hear today. One is how we win through performance with a differentiated edge cloud platform. Now when we say performance, often it is about speed. We do believe that speed matters in business. We believe it matters in the experiences that customers and users get. But sometimes performance is in terms of the efficacy of our security products or in the reliability and resilience that our distributed platform delivers. You're going to hear about how we have one platform, one network on a global basis and how that enables us to deliver multiple product suites and enable customer outcomes in ways that are easier for our customers and, frankly, operationally and financially more efficient for Fastly. You're going to hear about how our platform is positioned to where the market is going, about how AI and other trends are evolving in a direction where performance and latency and resiliency will become more important as they're deployed to more mission-critical applications. And then last, but certainly not least, you're going to hear from Rich Wong, our CFO, about how we're going to convert the momentum that we've got with our platform into long-term value creation. Before I dig in, I think I'd just share a few numbers about Fastly. We identified a TAM of $22 billion we're somewhat disciplined in the way that we approach the question of TAM. This is TAM that we have products today in these markets to participate in. We do believe that given where things are going and the quality of our platform that, over time, this TAM will grow as we address additional markets. Obviously, given our revenue, we have relatively small market share, and we like to see this as tremendous runway for us. With the products we have today, we can grow substantially faster than the market for many years. You can see across the top here, some numbers reflecting the momentum that we've gotten in the business. But I thought I'd mention the over 5 trillion requests that we process on an average day at Fastly. That's an enormous amount of volume at scale that's being handled very efficiently and it provides us with a tremendous amount of data as we go forward. To be honest with you, though, the number that a lot of Fastlyans are most proud of on this chart is the 97% average customer sat score. That's obviously an incredibly high number, as an average. And it's something that we take very seriously. And I think it's one of the reasons that some of the largest and most important companies in the world rely on Fastly. So I said I'd talk a little bit about where we've been and where we're going. I want to take a moment to do that. A couple of years ago, we said that we were going to accelerate growth and profitability. We said that we were going to do it through cross-sell and upsell within our existing base. And we said that we thought that one of the absolute key things for our success was multiproduct adoption because as we drive multiple products on our platform with a given customer, a relationship with them deepens and the value that we deliver increases. We took a bunch of actions, one is we retooled our innovation engine. We have new leadership for engineering and product, and we've accelerated the pace of innovation and product releases and are driving a much more iterative model that's much closer to our customers so that we're more assured of delivering customer value with each release. We refocus on platform-led value. We don't sell products that much anymore. Scott will talk about it. It's really how a combination of product suites on our platform come together to the customer's problem and deliver an outcome. Of course, we've talked about our go-to-market transformation that Scott's been leading now for some time, and we've made tremendous strides there. He'll share with you where we're going. But that's been a pretty extensive program. There's been a lot of leadership change structural change, incentive change, change in what segments we focus on, change in geographical focus, all designed to make sure that we're focused where performance matters, and we have our highest win rates. Last but not least, we focused on disciplined execution. I think you've seen that with things like margins and OpEx over the last several quarters. But my entire leadership team is committed to disciplined execution where everything that we're investing in and everything that we're working on moves our strategy forward. And we've gotten results. We've now posted three consecutive quarters of 20% or higher year-over-year growth. We've had 5 straight quarters of improving NRR capped off at 117% last quarter. We've driven the company to profitability. Last year was the company's first year of profitability with operating margin of 15%. And we continue to drive capital efficiency with 6 straight quarters of positive free cash flow and we'll have a chance today to discuss why we think we can continue to do that with the capital efficiency of our platform. I don't say this to say, "Oh, we're done." This is merely where we are today. and we're going to keep driving improvements in the business and keep building on the momentum that we have today. So to understand Fastly's story, you need to understand what we mean when we say edge cloud. So if you follow us, you know we keep talking about how we play at the edge and what happens at the edge. So I thought I'd just take a moment to explain what we mean by it because it's a term that's used in a lot of different contexts. And when you hear about cloud, almost certainly what people are referring to are the hyperscalers. And this is an incredible engine, huge data centers. In fact, it's in the news, the data centers are so large recently. And the play there is to drive an enormous economy of scale, driving an enormous amount of transactions and traffic to one place where you can get enormous efficiency. And there's no doubt that those clouds are the best place to run a lot of applications. It's an incredibly efficient and incredibly flexible environment. That said, there are some applications and some functions that are not a great match for the cloud. And they require things that are only available at the edge of the network in a more distributed system that's closer to users. Being close to the user means a few things. One is lower latency, faster response time. Another is more bandwidth. It's a shorter link to the user. And then lastly, it means that you can process requests right as they're entering the network, which can be important in terms of routing requests to the optimal place to ultimately serve them. Can it be important for security use cases where you want to or need to intercept something like a DDoS attack before it gets back to your cloud because if it makes it back to your cloud, you're already done because your cloud is overloaded by all handling all those requests. So the edge is where we play, and we believe that there's very specific functions and capabilities there that distinguish it from traditional clouds, and we tend to work with the traditional clouds because almost all of our customers have both cloud and edge aspects to their applications and deployments. So let's talk about what we think is important at the edge as Fastly. Now it won't surprise you to hear us say that performance and speed is incredibly important. If you think about it, it makes sense. If you're going out to the edge in a distributed system because you're getting lower latency and more throughput, what's the point in having anything less than a very performant system would blunt the whole reason you're even dealing with the edge. That's why speed and performance is so important to us at the edge. Security is obviously critical. We deal with the tax every day attacks on our customers as well as a tax on our own infrastructure. Resiliency is important. One of the things about the edge being a distributed system is that a lot of our customers are able to use our services increase the reliability and resiliency of their services. We'll talk a lot about unified platform today and how that has tangible benefits for our customers and also for Fastly. We'll also touch on edge compute and AI inferencing. We see this coming with new models just in the last few weeks, released that now respond in less than a second, in hundreds of milliseconds. I'm thinking of Jev, but there are many others. We see AI evolving to a place where performance is going to matter where responses are not measured in the minutes, and the edge has significant value to offer. So I talked about what's important at the edge, what is the edge? Like physically, what are we talking about here? This is the Fastly global network. We talk about POPs. These are points of presence. One thing you'll notice is we have 166 POPs. If you look around, that's actually a small number compared with a lot of our competitors. And that's actually quite deliberate. From day 1, we've spent a lot of time thinking strategically about the balance of how many POPs we have and where we place them to get the best outcome. If you place -- if you put too many POPs, you end up so fragmented that it's very difficult to get good asset utilization, and it's difficult to get high performance on things like cache hit rates because you just -- the POPs are too small. Obviously, if you're too few, you lose the benefits I talked about at the edge. So we've been very deliberate over the years in how we designed this network. I think if you look at any statistics, you'll see the efficiency. Recently, I had an experience that kind of reminded me of this and brought it home for me. We have a major retailer that moved off of one of our competitors to Fastly. They -- by the way, they were a full platform win. So they came across not just with delivery, but with our security, compute, bought DDoS, everything. And they got everything configured, and most enterprises don't move all their global traffic at once. They'll move a piece at a time to make sure that, that transition is going well. So they picked their Icelandic traffic for their initial transition. And Iceland, I assume everyone knows that, that's the little island there in the middle of the ocean south of Greenland, you'll note there is no Fastly POP in Iceland that was not one of the places that we strategically decided to place a POP. And they cut the traffic over. Everything looked good, but the next day they called us and said, we have some things -- we have some questions for you. And we were kind of surprised because it looked good on our side. Team got on the phone with them, and they said, look, what we want to know is how are you delivering better performance to our users in Iceland than the people we moved off of given that you don't have a POP and they do have a POP in Iceland. So I think that lands like what we've done here. I would encourage you not to measure us by the number of POPs and to realize that one of the ways that we're able to drive both the efficiency and the outcomes is with this very strategic balance of POPs around the world. So on top of this edge cloud, we deliver product suites. And these are, if you will, the tools in the toolbox that our customers use in combination to solve business problems and get business outcomes. And this is where a lot of innovation is taking place and we've been accelerating. The company, of course, started with network services or delivery, CDN, in a few minutes, our founder will join me on stage, and we'll have an opportunity to talk about that and how this has evolved. But we've then added quite a bit to the platform over the years, including edge compute capabilities, of course, our Next-Gen WAF, the observability bot in DDoS, which kind of completed the web application and API protection suite and has led to a lot of the growth we've seen in security recently. And then just yesterday, we announced some AI-focused products as part of Fastly for AI, and our Chief Product Officer, will be here to talk about that later today. So as we increase the number of product suites on the platform, we add to the value that we bring to our customers. It's kind of a combinatorial thing. The vast majority of our customers use multiple of these suites. And as we deploy more and more suites on the platform because it's one platform, it actually tends to get more and more efficient as these things are able to share resources. And we've gotten some momentum. 30% of our large customers today use four or more of those product suites. That's roughly doubled each year for the last 2 years. 2 years ago, it was 7%. A year ago, it was 14%. Today, it's 30%. So that cross-sell is working, and that gives us deeper, stickier relationships with our customers. Of course, we're the highest rated by Gartner for Edge Distribution Platforms. But we've also won Customers' Choice for 7 years in a row on web application and API protection. That's a security award for our security products. So we're very proud of the momentum we have both in terms of how our customers view us, how industry analysts view us as well as the business results that we've been able to drive. Now I've talked about the customer value. of this unified platform. But I want to take a moment to talk about why it's so important to us in our actual strategy and our operations. One is, as I said, greater wallet share and the ability to just grow faster by having one platform. Because it's one platform, it's easier for Scott's team to help customers on to additional product suites because they're already on the platform. It gives us, over time, greater pricing power and more strategic relationships as we cross-sell, and we have more customers using more product suites. We believe that that's one of the key predictors and indicators of our business success. Of course, higher retention as customers are using more services and solving more business problems and then more efficient use of our common network. I think this is something that we don't talk about perhaps as much as we should. If you look at the significant improvement in our gross margins over the last year, which I'm sure Rich will be happy to discuss. One of the key things is the fact that on Thursday night, our network might be delivering Thursday Night Football. On Black Friday, the exact same network is handling e-commerce. The next week, it might be handling a video game release. That's all the same resources, and those resources are dynamically allocated in real time to whatever the load is on a global basis. That enables us to deliver a much more efficient network outcome than a competitor, an alternative who has siloed networks for security and for compute and for delivery and so forth. It also makes it easier for us to launch more products more quickly. There's no need when we launch new products to deploy specialized capital or have a physical deployment process, we can just launch them onto our multiproduct platform. So I talked about how we view AI is evolving. I wish that it was going to be a straight and linear line as you see on this chart. Actually, Artur and I will chair a chart of what it really looks like in a little bit. But conceptually, we see AI evolving from things like generative AI, obviously, to agentic today, and we see a lot of agentic traffic on our network. But heading towards new applications, whether it's physical AI, where latency really matters or it's the things I mentioned earlier with Jev, where we're trying to get immediate decisions out of models. We see that as something that will create more opportunities for inference at the edge. We're not believing that running frontier level LLM models with trillions of parameters at the edge is a great idea. It takes a lot of resources. You don't concentrate as much capacity to it. And frankly, if it's going to take a minute for the model to react, the benefit of being at the edge just isn't there and the efficiencies of the cloud win. That said, we're seeing a bunch of use cases on our platform today. We've talked before about the fact that we see AI traffic growing 6.5x faster than human traffic. But here's some additional snapshots of things that we're seeing on our platform. We've seen -- actually, I think this first statistic is just in the last month, a 7x increase in the distribution of AI models across the platform. I think this is interesting in terms of traffic on our network, if you will. It's also interesting because it starts to suggest that these modelers are being run in more environments. And I think that's an opportunity. So far this year, we've seen 11x growth in MCP traffic and services on our network. If you're not familiar, MCP is model context protocol. It's how models integrate with each other with agents as well as with resources like databases. We've seen a 12x growth in essentially content authenticity, so much using AI to authenticate content as, I guess, not generated by AI at some point. And we've seen the 34x code generation is incredible. So 34x growth just in the last couple of months in that use case. I'll emphasize, these are snapshots of what we see today. We don't have a crystal ball for AI. I don't think anyone does. But we're seeing these use cases come on to the network and scale -- they start small and scale rapidly. Now why we win? I mean we have right at the top of the slide there, performance, and there's no doubt that, that's our calling card. As our retailers in Iceland noted, it can be quite a contrast and astounding. But there are a bunch of other reasons why we win. Certainly, foundational to that is resiliency. That's something that I think has been in the news a little bit in the industry. We've architected our network as a distributed system with resiliency in mind. Cloud neutral flexibility is important. We are not offering cloud computing services or competing with hyperscalers in that way. And that means that it's easier and more natural for a store together to get customer outcomes as almost all of our customers use the cloud as well. Strategic engagement with enterprise customers. You heard a little bit of that in the customer video. These large customers view the Fastly team as an extension of their team, and we love that, and we think that's incredibly important to them. And then last but not least, the unified programmable edge platform. Our platform is more configurable and more customizable than our competitors, which means some of the most advanced engineering teams in the world use it to do things that they can't do anywhere else. And any one of these enables us to win, but our strongest deals are usually when multiple of these come into play. And these are some examples of the sophisticated customers that we win with. I don't think anyone who follows Fastly would be surprised to hear that 8 of the 10 largest streaming platforms are customers. I'll leave it for you to guess which two are not. But also 6 of the 7 largest technology companies in the world rely on Fastly. So it's broader than just streaming. And of course, 6 of 10 of the largest video game providers. So we're very proud of the large customers that we serve. They're some of the most sophisticated and demanding customers in the world. And I think one of the things you'll hear about today from the team how we're going to bring the power of the Fastly platform to more customers. We have a few clear long-term strategic priorities that you'll hear about throughout today. First of all, how we lead with the platform. And Art and I are going to have a moment to discuss that in just a few minutes. Second of all, how we're scaling outcome-driven solutions. Our Chief Product Officer, will come up right after me, Artur to discuss that. And then how we expand our market reach and our President of Go to Market will be up to discuss that. So you'll hear about each of these priorities through the day as we go forward. Now after that, Rich is going to get up and he's going to present our plan for growth for the next 3 years, which drives our revenue well over $1 billion and has an 18% CAGR at the midpoint for the next 3 years. So we're excited about that, and Rich will share more details when he takes the stage. Now I mentioned the team, so I wanted to introduce the team here. All of the leaders across the top are presenting here today and will be available for Q&A. Some of the leaders across the bottom are in the room as well, so you may run into them. They don't have a formal presentation role. But this is a team that's evolved quite a bit over the last year. And I'm incredibly proud of the team that we've built. And I think the results that we've begun to deliver illustrate why. So this is how we'd like you to think about Fastly. We're a market leader with a differentiated platform. I think the growth and the gross margins that we're driving at this point are evidence that support that. We've reaccelerated the business. Three quarters in a row of 20% or higher percent growth, we now have momentum. We've improved our product diversification and revenue diversification. I mentioned how we've, in the last 2 years, quadrupled the proportion of our customers who use four or more product suites. We have multiple pathways to drive growth above the market rate, whether that's expanding our product portfolio, upselling, cross-sell within our existing customers, new logos, geographical expansions, we have a number of ways to get there. And then last but not least, we believe the edge is going to become an increasingly important control point in AI over the next several years. So with that, I'd like to invite our founder and CTO, Artur Bergman, up, and we're going to have a fireside chat. Thank you.

Thank you, Kip.

So I met Artur for the first time when I was considering whether or not to join Fastly as the Chief Product Officer. Spoiler alert, I did, obviously. But I met Artur at a sushi restaurant in San Francisco. I did not know -- I didn't have a particular set of expectations for the dinner. But it went on, I think, for 4 hours. In fact, it ended when they kicked us out. And we became somewhat self-conscious of the fact that we were the only people left in the restaurant. And over that discussion, I came to believe that Fastly had a really unique platform. And there are a lot of people who have contributed to that over the years, but probably no single person has contributed to it more than Artur. And I just felt like the company had an incredible opportunity in the marketplace given the quality of the platform. So Artur played a key role in my decision to come to Fastly and has been an incredible partner ever since. But you founded Fastly to solve a problem that there was no solution to in the marketplace. Maybe you can talk a little bit about what was different and distinctive about that approach and how it kind of flows through to today's world?

Absolutely. And it's also long and very good, and I think for Fastly, a very valuable dinner. Ends up with Kip's sitting here. I'm good to see you all. I know some of you from a long time and other new faces, but it's good to be here. We started Fastly in 2011, and 2011 AWS had existed for a whole couple of years and had changed how developers wanted to use infrastructure. And we needed fast performance around the world. We had a very slow low time in Europe, and we knew that performance drove usage. So I cofounded the Velocity conference, which was the first conference on SRE and performance in 2007. And all these metrics on performance, we knew. But we had a very dynamic site. And the edge providers at a time where -- maybe you could describe them as anti-cloud. Everything you want out of our cloud, you got the opposite. And so you had no visibility and no control. And we ended up building our own, we dropped our page load time from 22 seconds in the U.K. to less than 3 seconds, I think. And then we ended up starting Fastly based on that technology, because it turns out that a lot more people wanted access to this. And the only people who kind of had that power were someone who worked for hyperscaler because they already have their own networks. Otherwise, you were kind of out of luck. That was the founding story to provide the real-time instant aspects of a platform and a modern developer-friendly edge environment.

And a lot of that platform still is in place today because you guys took an incredibly software-centric approach that -- I think one of the things that is different.

Yes. The entire founding team, the entire founding team are software engineers, not network engineers or hardware engineers or data center engineers. And so we solve problems with software. I hate spending money on dedicated hardware, and we built it from scratch up with the software on the servers, but also how we architected the networking infrastructure and networking stack, and we ended up with a very hyper-converged architecture where we control everything from the software that runs on the switches all the way back.

Now some would say that CDN is commoditized. And I'm wondering, given your long experience in that market and that technology, what your thoughts are on that?

Yes. No. There is one legacy provider that has a very large market share as we have been taking it over time. They still have it, and it's not typically a commoditized market behavior unless you have a natural monopoly, which would argue that does not fully exist here. The reality is that there is plenty of bandwidth delivered that, I would say, is commodity. If you're downloading something in the middle of the night, does performance matter? Not really. If you're waiting for a game patch so that you can actually play the game with everyone else, the performance does matter, and that is not commoditized the same way. There is certainly plenty of video out there that is commoditized, but World Cup live streams, not really commoditized, right? Like people pay a lot. And if you don't deliver it, you are a broadcaster in trouble. On the API side, from day 1, Fastly was designed to handle API traffic. API traffic is not really a commodity. Some of it is, some of it is not. So CDN is a very, very wide term. And I wouldn't say it's increasingly commoditized. I think there are the buckets and those buckets have just existed. I think some of the noncommoditized buckets have actually grown over the last 10 years.

So a new class of traffic is agentic traffic. They generate a ton of activity. What are you seeing there?

Yes. I mean this is so exciting, like the entire technology space. I feel blessed. I don't -- I didn't think I would get the chance to experience another '95 to 2000. And that's what's happening, if not even more. That was a long time ago, so I don't fully remember it. And it's changing the traffic. We have been trying to answer the question, what does agentic traffic mean? And we went on and looked at -- for data around this, and we found this lovely chart. So PyPI for people who don't know is the Python Package Index. They happen to be on Fastly, but this is a public data set that you can query is on big query. And this is all the software modules that you download, so you don't have to rewrite code. And you can kind of see it grew 4.8% before. It's been kind of slowly, steadily growing. That is like human usage of python going up, right? So it kind of grew as more and more people started using Python and as more and more developers in the world. Then this thing happened, Claude Code Research preview. And traffic took off. I mean if you ever watch what your agents do, the first thing an agent does is to download stuff from PyPI because for some reason, pretty good reasons, python is the preferred language of every LLM on the planet and to avoid writing code, they download things from PyPI. That is not human behavior that is driving the increase. That is agentic behavior. And so we kind of like -- this is a great chart kind of showing what I've been calling talking to customers about the tsunami of traffic that's coming this way -- their way because once this code is written, it then has to do things. a lot of the things it does is API calls and other things on the Internet. And that means that as this continues, and as you can see, the chart -- the curve is actually accelerating still, those agents are going to talk to our customers. And so I've been calling it the tsunami of traffic. And it's super exciting. It's -- we have disconnected infinite usage from human time, right? Like there's only so much we could code python or watch video or serve the web and buy things or do whatever. And now they -- AI doesn't sleep and they work when we do other things.

So how does this influence the importance of the edge? Or what kinds of things do you see -- how do you see the edge participating in some of this?

There's used to be a pretty strong correlation with increased visitors to your website and increased revenue, right? I'm selling things online, more people come to my website. I know my conversion rates. Now we're looking at a situation or -- if it's ad-driven similar -- now we're looking at a situation where agents can go around and shop on a lot of websites. And so you might see a significant increase in the amount of eyeballs or viewers -- eyeballs is wrong word, agent viewership for the same amount of transactions. You can't really opt out because then you won't have any transactions eventually. So you have to drive down the cost of serving these users. And the best way of driving down costs to serving people looking at your things is caching. And doing it from the edge means less bandwidth traversing long distances and serving it from an SSD or memory is cheapest and fastest. But you still need to be able to instantly cache and validate, instantly change everything. And so all these attributes that Fastly had from day 1 and it's built into our architecture becomes even more important. It is also a lot of traffic to load balance to send to the right place. You need to secure it. So you need to run WAF, you need to run bots, you have to have great DDoS. You have to be able to personalize it somewhat -- and all of that just is what customers have been doing today but at a much greater scale, which makes it more important because otherwise, you will not handle these massive spikes.

So where do you see inference at the edge going?

At some point, technology will probably reach a point where running larger models everywhere makes somewhat sense. But today, the large models are not multi-tenant. GPUs are not multi-tenant and they take quite a while to respond. So I think in the immediate term or next couple of years, especially given the shortage of both data center and power and so on, where we are -- the data center power is not an issue, but if we needed to put much more GPUs it would be, is small models. So you can use a large LM to train small models. Historically, training models was very annoying because you had to have a data scientist team and you have to interact with a data scientist team. And now like LLMs are great data scientist teams like you can have them classify data and then build models. Like I just trained the model at home to tell me if the pool cover is open, and any kids are near the pool. And I think I paid less than $100 to OpenAI to train the model. And now you can tell me with very, very high confidence. At 90% confidence at 20 to 30 frames per second in a model that's less than 4 megabytes. So those models, you can run at the edge, right? Tyler, who is my co-founder and distinguished engineer at Fastly. Last week gave a talk where he showed a 3.3 megabyte chess model. And it was kind of upset because it consistently beat him at chess. But a very small model still really powerful that was running in our compute on the edge. So if you can couple large models in a central location with small models at the edge, which kind of fits how caching has been working. If we don't have something, we go back to origin I think that is the answer of what inference at the edge actually is over the next couple of years.

That makes sense, and it's exciting. Now I had an interesting conversation with Scott, our President of Go to Market yesterday. And one of the things that came out is he said that you spend more time with customers than he does. So can you tell us a little bit about that?

Yes. We only exist because of customers, right? Like we don't have a business for some academic reason or because it's on our business. We have a business because customers buy from us and use us, and we deliver value to them. And it's super fun, right, because I get to work with customers and see what they do or what we help them do for their users, whether it is like a massive spike in traffic that they can handle or protecting them from attacks, particularly being hacked or tons of kind of cool use cases out there. Only way I learn about those is by going out and speaking of customers. And then bringing back to Fastly what we need to do to help them do more creative and more innovative things on the edge. And so that is really like a large part of my driver. The drive for me is spending time with our customers. We're also so critical to them that we end up spending sometimes a low time with them, and it's been 15 years now. So some of them are got to know quite a lot, and I've seen them switch companies and immediately bring Fastly with them. I kind of see -- it's kind of a personal KPI like when we help our customers get promoted, it feels really good.

Absolutely. So looking ahead in terms of tech investments and platform capabilities at Fastly. What do you see kind of on the machine for the next couple of years?

I think underlying all of it is efficiency, right, because of where we are with the hardware cycle and so on. We have to make sure that the platform stays highly efficient so that we can then deliver all these amazing things into it. Efficiency also goes very much together with performance. In an inefficient system is rarely a performance system. Like if our WAF can filter for an attack 10x faster than our competitor. It typically means we use significantly less resources to do so. And all of that comes down to this passion and focus on performance and efficiency. On top of that, helping customers deal with this tsunami of traffic and the complexities of this traffic. You can no longer have humans in the loop if the attackers do not have humans in the loop. If traffic is not directly human-driven. Steering it, low balancing it also needs to not have humans in the loop. It used to be pretty easy to -- easy is the wrong word, but you could predict the max usage of -- on the Internet website based on the amount of people in the country if you were launching something in your country. That's no longer true. You see it's a few bots. They're Pokemon bots now because frankly, that's the new cool thing. But just a lot more to do that. And then we have the small model, large model, how to help people reason about that and think that and then give lots more visibility into how all these systems work together from a security delivery acceleration, resiliency, reliability point of view, to tie it all together. And it's just super exciting. It's like these are hard problems. I spent 15 years at Fastly. I kind of had thought that some of these core Internet performance security scaling problems were sold. And it turns out there aren't and there's still a lot more room for improvement, which -- yes, that's pretty fun. And certainly, agents have helped us find and analyze and fix issues much faster and at a scale we haven't been able to do before.

That's amazing. You are an incredibly successful entrepreneur. At this point, could probably be doing anything you want. But I guess it's what you just described that keeps you at Fastly because you're just as engaged as ever.

Yes. It's -- the access to the data we have, the technologies we have would be hard to replicate anywhere else. And the same with the customers and the relationships there and the team. I feel very lucky to have those components to work with on a daily basis. And going out and hearing customer stories, you mentioned Iceland. The Iceland story. I -- there is a long-term customer defer nearly a decade, I've been trying to convince them to move all their small, high-value API traffic to us. And they moved a lot to us, but that part was like this is complicated, it's working like you should move, you should move. And I think they used some agents and then they help the move and they moved. And they're like, it's 9x faster. Our app is showing a 9x performance improvement for when users are like moving around inside the app. We should have moved 9 years ago. I don't know like I told you wish you would believe me. But I'm really happy that you finally moved. And then I'm looking at it like we can probably make it even faster. And there's still opportunity. We could pick up -- I wonder how fast Iceland would be if we put a POP there. But yes, I mean, all of these things are just super exciting and super fun, and I get to work with amazing customers and amazing coworkers.

No, it's remarkable to have a founder and former CEO, so engaged in the business and so engaged with our customers and with the team. So I think Fastly is lucky to have you as well.

Thank you. And thanks, everyone. I'll be around here for -- if anyone was talk to me, I'm not particularly scary. So -- and yes, this is -- I mean, this chart kind of sums it up the challenge, but also the excitement on here. So absolutely. Thank you. Thank you, Kip. Thank you, everyone.

Thank you. So that was fantastic. Thank you, Artur. I'd like to bring Kelly Shortridge, our Chief Product Officer, up to talk about our product strategy. Thank you.

Thanks, Kip. I love that fireside chat illuminating our journey. There's a lot to be proud of. How we're taking it to the next phase? And also how we're going to drive differentiated growth across the whole platform? I'm Kelly Shortridge. I've had the privilege of serving as fastly Chief Product Officer since last November. But I've been at Fastly for 5 years. and worked across both engineering and products, where I led the security business and during that tenure drove growth to 53%, including the expansion of our portfolio. Now I'm looking to do the same thing across the entire platform, in Fastly's whole portfolio. And what I bring to this is really a resilient mindset. I think everybody at Fastly agrees how foundational resilience is to our platform to the company, just like Kip said, also bring hands-on prototyping and a relentless focus on ROI as well. And it's not just my mindset, it's the mindset of our customers. Like Artur said, there's nothing more rewarding than talking to dozens of customers every month. And what we see is that they win by continuously adapting based on evidence based on insights, and they don't win by standing still. And customers choose Fastly because our platform unlocks that for them. It's not because the technology is covered for its own sake, just like Artur said, it's because we help them transform their business. We help make sure that their business keeps improving. So now that you know where I'm coming from. I want to talk to you about how we're evolving the platform. Kip just walks through the platform journey, the company journey. It's my pleasure to show you how we're going to transform to tackle our biggest opportunity yet. So first, by building a broader, more accessible portfolio that brings our industry-leading technology to a larger group of users in industries. And second, by delivering a platform that meets customers where they operate, across a variety of cloud environments, private infrastructure environments and also agnostic of which AI models that they use. And third, by advancing our platform innovation, the innovation engine Kip talked about, by using the knowledge we gain, that data that Artur talked about from across our platform, because we are that centralized service provider that's so trusted to deliver differentiated products and solutions to our clients. Then fourth by extending our reach into the always-on industries that need us most and that we haven't penetrated today. So doing this well is what gives us the right to win in these markets where we are underpenetrated today. So with these four pillars in mind, let's talk about where this journey started. And as you heard from Kip and Artur, this has been a journey. I want to talk about how we're evolving our product portfolio, but it's important to level set on where we were. Where we started, as you heard from Artur, it was a heavy networking focus, right, differentiated architecture but a networking focus. We only have one security product when I joined Fastly. It was sold piecemeal and really on a platform for developers who craved that raw technical power. We were a platform that was also built by developers, for developers. But where we stand today is an enterprise-ready multiproduct portfolio. It's a platform with the breadth and configurability that meet a wide variety of customer needs. We have a security portfolio now with five products. And that includes patented differentiators as well. And it serves as an on-ramp into the platform sale, which Scott is going to talk about later too. And over the last 4 quarters, I think this is a very important point. to keep in mind is that half of our new business deals included security products. That's real proof that security is that new on-ramp. It's that new entry point into the platform sale. And it's also a value driver of our platform. As Kip pointed out, we're increasingly complementing our delivery security solutions with other solutions as well. You saw that yesterday with the Fastly for AI announcement, I'm going to dig into it a little bit. This involves a full suite of solutions that also complements where we're going next. And that's really what excites me the most. And I know it excites everybody at Fastly the most too. And that's we're running that same playbook that we ran for security, again, we're running it again for the whole platform. Now we're continuing to invest in technical differentiation for sure, but we're going beyond that, provide the experience-driven capabilities like reporting and insights that enterprises need. And as we do so, we're attracting the world's most complex companies to our platform, and that's because of our outcome-driven approach. Now these are companies where downtime stops a manufacturing line, it holds to trade. It stops the revenue spike from flowing at the end of the day. Always-on businesses have to be available 24/7 regardless of what the Internet throws with them or what their AI agents do. So we're no longer building a platform for people who love technology for its own sake. We're really building for executives, their teams and their AI agents who need the platform to just work at the end of the day. So that's where we've been. Let's talk about what that means for our customers today. Now our customers want to unify the platform. They don't want disparate point solutions. They want a unified platform. The unified platform that we built for them is really all about enhancing the outcomes they can achieve as a business. What we hear in the market is that every enterprise in every industry, no matter the industry, they face two key questions, which is, is this technology investment paying off? And is it helping me win in my own market? That's especially true in this AI era. What this looks like in the specific experience industry by industry, but it's the same shape, which is that enterprises want their teams to spend more time on what makes them special in their own market, less time fighting the complexity that comes from distributed software systems that they have to build to succeed. And that's where Fastly comes in. It's delivering, again, not just discrete point solutions, but it's unified and integrated platform that's intentionally built to enhance these customer outcomes. It's a platform that's reliable, that's fast, that's safe, secure that really crucially is also resilient from design all the way through the solutions that we offer. And crucially, that it's meeting customers wherever they want to run their wherever they want their workloads to operate as well. And so when we combine that technical mastery that Artur talked about, with a set of integrated configurable solutions, the results are really powerful. Let me give you a very specific example. So take a large retailer client. I love this example. They first added -- they were a platform customer. They first added Bot Management on top of the existing solutions leading into the holiday season. What's fantastic about the story is they literally thought there was a statistical anomaly because they saw double-digit conversion rate increases during the holiday period because they adopted our Bot Management solution. And that's what's special about our platform is we have security solutions that are driving business outcomes. And those are the outcomes that matter to our customers. It's not security for its own sake. It's aligned to their business outcomes. And now we have the opportunity to go even deeper with this customer and all the other customers who have adopted this multi-product platform, bringing them more sophisticated solutions to meet their emerging and very complex needs. And one key takeaway here is that this is all about expanding beyond our historical niche in delivery services to bring a broader, more configurable platform to enterprises with a more compelling value proposition as well. Then we've had this recap of what customers need. Let's talk about our long-term priorities. And Kip did cover these, Scott is going to build on them again after the break. Lead with the platform, scale with outcomes-based solutions and crucially expand market reach. So let me show you how we're bringing it to life through investments into the product portfolio. Let's start with the portfolio itself and how we're making it broader and more configurable as well. So a few years ago, again, one security product. That was it. When I took over that business, I realize we had a lot of work to do. We had to roll up our sleeves. We did get it done. And crucially, we've also sustained it. You saw last quarter that we had 43% growth year-over-year in the security business. And like Rich talked about, both DDoS protection and Bot Management are growing in the triple digits. That's a really powerful outcome of that portfolio expansion in the security business. So now we're going to run that same playbook across the entire platform. It's 6 suites, each a significant product and value driver in their own rights, plus cross-platform add-ons and services that enhance that overall platform value proposition. And each of these suites is an on-ramp into the platform. It's a way that a new customer can come into Fastly, start gaining value from our platform. It's also a way that customers that are already with us can expand the value that they get from our portfolio as well. So the way we'd like you to think about Fastly as this ever-expanding sticky portfolio of on-ramps that customers buy into and never leave. And the configurability of the platform means we have this, I think, Kip referred to it as is combinatorial value. This virtually limited combination of product suites, add-ons and services, all tailored to our customers' needs and the outcomes that matter to their business. And here's the number I want you to take away above all others, which is that 72% of our customers now use at least 2 product suites or more and 30% use 4 or more. And that's that 30% number is up from 7% just 2 years ago. So it means it's doubling 2 years in a row. That's really powerful, and it's a tangible proof point that this portfolio strategy has been working in practice. So that's the clearest signal and it's not just about how much we sell any one product. It's really about how many of our 6 suites each customer adopts. It's about the overall value that each customer is getting from the platform itself. And that is a shift for us. I think many of you have seen. Our technology has always been excellent. But for years, we sold it one product at a time. Our financial results didn't compound the way that our technology really deserved. And that's what a portfolio fixes because today, when a customer is adopting a second suite, a third suite, a fourth suite fifth. That's not just growing wallet share, but it's really becoming an infrastructure that they rely on in their business, just like Artur said, they bring them to their next employer because of how much they know they can depend on us. And that's a fundamentally stickier and more durable revenue relationship with customers than any single product could deliver. And it's also how our technical excellence finally starts compounding into the financial excellence we feel we deserve too. What we've also learned is it's not enough to just have a large portfolio of technically differentiated or even differentiated solutions, the experience we create around this portfolio and around these products also matters to enterprises. So let's talk about usability. We're investing heavily in democratizing and simplifying the platform. Today, our platform is really powerful, but it works best for the deeply [ technical ] user and developers who already know how to drive it. We want it to work well for everyone who touches it, the executive who needs to know the exact insights right now to make some sort of decision. We wanted to work really well for the AI agents building on our platform, we need to harness the full power of what we offer on our customers' behalf. That's core to our transformation. It's from power that works only for experts who understands the technical weeds, to power that works for however our customers actually operate. However, enterprise is actually operating practice. So what does this mean concretely? It means simplifying the user experience, both for human users and coding agents like Claude Code, quite famously, that we see quickly growing in prevalence, also means automating expertise across the portfolio, so customers don't have to be experts. They can push a button and it just works, Fastly handles it for them. And we're building dashboards and reporting that make it easy for busy executives to get those answers right now, translating technical insights into business decisions. And we're leveraging our cross-platform insights deliver actionable business intelligence to help our customers make better decisions. We'll talk about a concrete customer case study in a second on that. So we're investing in making every single user's experience as simple as possible. And as we do so, we're meeting a broader range of customer needs, including supporting the standards that let coding agents seamlessly inter-operate with our whole portfolio and platform. Just the same way a human would, and again, leveraging the full capabilities that we offer. To be democratize and simplify the platform well, we're going to expand this platform into that broader customer base that Scott's going to talk about. This is how we're enhancing usability. Let's ground this in a customer story. Making a concrete with a customer case study. Let's take a national newspaper. This national newspaper brought us an existential threat. AI crawlers that we're scraping their journalism that they monetized with no licensing agreement, and that threatened their entire business. We co-innovated with them linking arms to deliver content guard less than a year later, helping them detect, classify, allow, block, otherwise monetized and manage bots that we're engaged in that unwanted content scraping, and consuming significant bandwidth as well. Our platform really critically gave them the business intelligence that they needed to negotiate watershed licensing agreements with AI companies and gain a lucrative new enduring revenue stream that transformed their business. This is a publisher with hundreds of thousands of digital subscribers. They have hundreds of millions in revenue. This is real scale. And we protected their intellectual property. We help them create these new monetization paths, and we gave them greater control over their own business model. And as their business grows in size and complexity as they adopt their own AI internally, we have integrated the platform capabilities to grow with them. We make sure that with our platform, customers never outgrow us. What we build for one we can now do for tens of thousands of enterprises facing this exact same challenge each in their unique way. This is a customer case study in action. Let's talk about the architecture that makes it possible. Now architecture is cloud and model-neutral by design. Let's talk about what we mean by that. So what we see is that enterprise infrastructure today in enterprise software architecture is not this neat and tidy environment. It's only becoming more complex. One more we see our customers using a multiple kind of mix of clouds the private infrastructure and now they have AI agents, especially coding agents from multiple providers, and they all need to work in harmony. They all need to inter-operate. And with these AI agents that are now shipping and writing code, into this messy multi-model, multi-cloud world that multiplies interconnectivity and interdependency faster than any human team can actually keep track. They need new solutions. And it's why every executive I talk to out there, that's a CIO, CTO, CISO, they describe the same feeling, which is that they feel like they've lost control and they feel like they can't keep up, they feel like they don't know what's even running in their environment anymore. I think of it like a wild west that's forming where it's not just one messy environment. It's a patchwork of clouds, private infrastructure held up by the swarm of coding agents, with no consistent way to see control or make sense of any of it. They need a partner who can help them manage their intertwined needs who can help evolve with them as they need to evolve their business across this increasingly complex environment. And that's where we come in. Our platform is built from day 1 and also especially now to be customers wherever they are, whether it's cloud-based, whether they have private infrastructure. We're cloud neutral, so we work across every major cloud and across every type of private infrastructure that they have. And we're also AI model neutral, which is an increasingly important point here. We work with whatever AI model that customers choose which, from what we see is usually multiple AI model providers as well. It's not locking our customers into any one AI vendor's road map or pricing structure. In other words, we're helping them diverse that Wild West that they see forming by giving them the flexibility for wherever their software is operating for whatever AI models that they're choosing. And that architecture is a tremendous advantage where we see a massive opportunity going forward. This is an important point about our cloud neutral model neutral architecture, and it leads into us talking about how we're governing AI. What we see is that AI is touching every single part of our value chain, every part of our value chain. And we see it the same way that we saw security just a couple of years ago, which is an enormous opportunity and one we know how to capture because we've done it before. Yesterday, we -- I hope you saw the press release, we announced 3 advancements that help strengthen our role in governing AI, which is AI Runtime Control, the AI Firewall and API Enforcement. These reflect our initial steps in the forecasting, the resilience and the steering that we hear loud and clear in our customer base in the enterprise market that they need for this new world. That's the forecasting resilience in steering they need to adopt coding agents at scale with much more to come as we build for a future where enterprises have to confront the wild west of increasingly probabilistic software which is very different than the world they've known to date. But none of this stands alone, plugs directly into our capabilities that they already trust like the Next-Gen WAF, Bot Management, ContentCard, DDoS protection. We're applying that same playbook in security that got us those great products that took us from 1 product to 5 that took us to being present in over half of our new business deals. And just like security, where we see this longer-term massive opportunity and upside. We're just getting started with these Fastly for AI solutions. This is step one. But here's the point to remember. Our opportunity with AI is platform-wide. It's platform-wide. It creates more interactions to accelerate. AI creates more software to secure and govern and more complexity to control. That's what enterprises have to contend with. Now before I close with this discussion about how we're leading into this platform approach, let's talk about one area where we see significant opportunity as well, which is compute. And this relates to how we're strengthening our role in governing AI because compute and AI run time control work is a synergistic payer. We're investing more in that strategy of where products combined create more value than enterprises than both products alone. So compute handles the first efficient, fast a logic pass. So then AI runtime control only routes a workload or a query to the rare case that needs a large generative model. That's what the field calls Cascade inference, and that's core to our approach going forward. That cascading inference, that symbiosis between compute and AI runtime control is what makes AI economically viable to adopt at scale for enterprises. Pairing AI runtime control and compute is really just step one of us marching towards this vision of collective computation and cascade inference. It's early, but the direction is clear, and also our priorities are very focused right now. We're starting with solution enhancements, extending compute into other workload types, extending it to interoperate with our existing solution suites that we talked about and expanding also how we improve our execution. As part of the portfolio play here, which we're playing to compute as well, we want compete to be a critical on-ramp too. We do that by finding repeatable use cases, just like we linked arms on that Bot Management case study. We'll roll out experience improvements to assist this as well with more push button solutions where it just works, faster ramp-up times as well and also recommendations that are automatically tailored to the customers needs, and we're going to deepen our value prop, especially on that front in that tailoring by leveraging that unique data we collect from our platform that Artur touched on as well. We really see data and intelligence as a differentiator here going forward that we infuse across the platform. The point is, we feel we're truly defining this next era of edge compute solutions. We're also excited about what this is going to mean as we continue to evolve towards this portfolio and platform-driven approach. Talked a lot about the platform, how we're expanding it, how we're creating that synergistic value here. Before I hand things back to Vern for some Q&A, I want to quickly touch on the other two pillars of our strategy. Now the second pillar of our strategy is going about scaling outcome-driven solutions for our customers. From a product standpoint, the best example of this is how we have this customer-driven innovation group. What you're going to hear from Rich as well as this customer flywheel that we've been operating. As we've grown and matured as an organization, we've learned to leverage customer-driven innovation to power that flywheel. This doesn't happen by accident or chances, actually a result of a lot of discipline. It's a result of the disciplined approach that we're showing on the slide here, where we observe customer behaviors across the platform based on our unique vantage point, we pattern match the common challenges we see across the market and across every enterprise, no matter the industry. We codevelop solutions with customers to validate real-world demand. We productionize that capability of the platform, then we roll out and scale the new solution across our installed base. Let me give you an example. There's a high-growth payment experience management company that partnered with us for nearly is a year as we built out multiple products, informed API discovery, API inventory and then API enforcement. Now we came out yesterday in that exact sequence. And they even handed us their API schemas to make sure that we build it right. That's real trust from a customer. And that's the magic of a design partnership. It's making sure that a product has market fit before it ships. It's making sure that we can sell it, standardize it across all enterprises to thousands of them. It creates new on-ramps into the platform, right? New entry points into the portfolio and expands the existing set of products that our customers can adopt and get more value from this well. We can do this again and again, focus on scaling these outcome-driven solutions here. So now let's talk about how we're expanding our customer reach. Scott's going to cover this in a lot more detail in his section after the break. But I want to touch on how this really links with this multiproduct play. That's at the heart of what we're talking about here today. First, we're protecting the core. We're deepening our relationships with our largest customers built on years of trust. While we're making our technical depth radically more usable as well. But at the same time, we're expanding into segments we barely touched. Like I said, where we're underpenetrated today, where they're hungry for a portfolio that solves [indiscernible] resilience challenges because they're 24/7 businesses. Think of a way to illustrate this is considering the difference between an e-commerce customer and an industrial one. An e-commerce customer public Internet early. They chase speed and scale. Now if you think about industrials or financial services or pharmaceuticals, where we're underpenetrated today, they've been slower to adopt this. It's not because they're as sophisticated though. I think we all know that, it's because their stakes are different. It's not just independent shopping cart. It's again a halted trade. It's a stopped plant. These are serious consequences to their business. And it's precisely for this reason that customers need more than just a vendor. They need a trusted and unified platform. They need an integrated solutions provider with a configurable multiproduct approach that meets their evolving needs. We feel that these 24/7 industries are exactly where we're poised to win next. And we're going to do that with innovative products that specifically target these underpenetrated industries that up level the insights they need to succeed and evolve in the AI era. And we're bringing it all together. Recapping the Fastly opportunity. A broader, more accessible platform, meeting customers wherever they operate in the messy reality of how they operate, innovating with new outcomes driven products and solutions and extending our reach into always on industries that really need a platform just like us. They're hungry for our solutions. The number to member is the one I gave you earlier, which is more of our customers are adopting more of our portfolio. They're adopting it really fast, quarter-over-quarter. And that's the clearest evidence that this portfolio strategy is working. Again, that security playbook, I talked about that 43% growth year-over-year we saw last quarter. That's a playbook we're going to be running across the whole platform going forward. And this is how we turn technical excellence into durable compounding growth going forward. I speak on behalf of all the Fastly, we're very energized by this platform opportunity in this multiproduct play here. We're ready to deliver it. So with that, Vern, please come back to the stage as we prepare to take your questions. Thank you very much.

Okay. Thank you. Thank you so much. So now is our Q&A portion. For those of you on the webcast, please submit questions. We'll be rolling through there. We also have mics here in the room for those live in the audience. A couple of hands up here. Let's start with Jonathan here. All right.

Thank you, everybody. Really appreciate the presentation so far. Jonathan Ho with William Blair. One thing I wanted to start out with is on the chart that you showed with the exponential growth in AI traffic and connections, can you help us understand how that maybe translates into potential revenue opportunity? And maybe specifically, what types of product add-ons and expansion in capacity utilization that can drive over time?

I can start and then maybe Kelly can talk about the product there. It's early. I think I don't know if Kip or Rich can talk about some of the revenues we have seen around this space, but it is early. And as a percentage of the traffic, it is not that much yet. But it is growing. I think Kip had a slide, the other slide on the -- Kip about how much faster traffic is growing in human traffic. But it's still a fairly small amount.

Yes. I mean I think of it in sort of three legs. First is just volume. I mean as there is more traffic on Fastly's platform, there is more volume, there's more opportunity and that happens sort of automatically, frankly. The second part is things like what Kelly talked about with AI runtime control where, in addition to volume, we provide value-added products that help manage the new types of traffic. And I think that's, for example, bot and DDoS, we think are being driven in part by agentic traffic, and we're seeing the triple-digit growth rate that we've talked there. The third leg is what Artur and I talked about a little bit, for example, edge inference, but where we start actually running some of the AI functions on or in the Fastly network like edge inference for instance. So that's kind of how I think about the opportunity. We're seeing increased volume for sure, and that's what that slide -- the slide that I used was traffic and request volume, not revenue. So we're seeing that volume already. And then you could see us rolling out products that capture that. And then we -- for the reasons ate outline, we see inference and some of the other things coming down the road. I mean it's tricky I feel like I didn't answer your question, but we don't have like a separate AI revenue bucket or we can say, well, these dollars are because of AI. But what we see is the traffic volumes being driven by AI and that in terms drives revenue.

Okay. Next question for Fatima here.

Fatima Boolani from Citi. Art, I want to go back to some of your commentary around this the period that we're in right now for Fastly, both from a technological standpoint and for your customers, is that something that you've seen since 1995 to 2000 era. And then you also characterize CDN as maybe misperceived as a broadly a commoditized arena. But you delineated that there's commoditized buckets and noncommoditized buckets. And I'm wondering if you can give us a crystal ball from your perspective on how you expect that mix to change, especially as I think about some of the revenue targets that you have, which or slightly lower than what you're realizing today, but I just kind of wanted to get some of the ingredients on how you think about the complexion of the traffic over the next 3 years, commoditized versus noncommoditized?

I think the bucket of noncommoditized or like higher value is growing faster. And part of the way you can also see that is that traffic tends to need all the security features. Like the commoditized traffic doesn't need WAF, doesn't -- it probably needs DDoS, but that's about it. And so with the WAF or the security products in general, growing faster, that tends to be highly correlated with the higher value traffic. And that's probably the best way for me to look at that. In the media space, you kind of see something similar. The non -- the higher-value traffic needs on antipiracy protection, content protection and much higher rigor around that. And you saw there was -- a while ago, we had a joint announcement with La Liga on the privacy side. That content is much more valuable than other content. You kind of see that as well. And on the changes, I mean, what is so exciting for me is that it's changing so fast. And I was reflecting on -- I was talking to a customer in there, it feels like Fable came out a year ago, but it was actually only 3 months ago, right, it was a big deal. And so it's very hard to predict. It's kind of -- it feels like on a weekly basis, something new or changing. And in all of it though, the complexity is going up, counterbalances with our ability to use agents to manage their complexity. And part of what we really have to do internally and help our customers is to make sure that we use all agenda abilities to manage the dramatically increased complexity that agents are adding. I look there, but -- thank you.

Yes. Thanks, thanks, team. I have a question online. This is a product question regarding our release yesterday for Kelly. How does AI runtime control and AI firewall work architecturally, specifically, how they differ from injection products that run on endpoints? With runtime control specifically, does it run from an endpoint device or cover the network as traffic is delivered?

The answer is it runs on the platform, runs on the differentiated network architecture that Artur and Kip talked about. I believe the spirit of the question is, is it similar to kind of where some of the detection response players playing. That's really the world of employee laptops. We're playing in the world that like Artur said, is the one becoming especially complex, which is a world of enterprise production software. So we're looking at prompt injection attacks coming in to the software, companies are building different their own markets, not so much the employee laptop space that you might associate with traditional info protection. So all running on the network. It's part of that multiproduct play, where everything interoperates and customers are able to gain leverage by adopting solutions here.

Jackson.

Jackson Ader at KeyBanc Capital Markets. Did you -- does Fastly benefit directly from consumer AI use cases like news, OpenClaw, whatever might be coming? If so, how and like which products, which revenue streams?

Yes. I'm pausing because I -- we have a lot of respect for a couple of customers' confidentiality. But we do benefit from those things. And we're seeing use cases across delivery and security, including privacy capabilities for agents.

It's relatively easy. You go look at the major hyperscalers and consumer public brands, how they are implementing privacy with regard to server side, inference versus client-side inferences consumer agents and they will -- they mention us. We're a part of that. So it is. And I think the privacy aspect of this is also going to increase quite a lot over the next 5, 10 years. Both driven by consumer demand and driven by regulation in Europe and other places.

Take our next question from Peter over here.

Peter Levine with Evercore. Maybe to piggyback off of Billings question, you talked about tsunami traffic that's coming up. What does that look like? When does it happen? Obviously, we're seeing this peak now. And then second, when you think about traffic, as MCP adoption grows, like do you expect it to create a meaningfully new layer of like traffic for you all? Or is that just mostly just in API in that type of bucket. But just curious like what that looks like in the trap.

Yes, I'll comment and then Artur can add on. I think others in the industry have noted that while the AI traffic is growing rapidly, purely on a volumetric basis in terms of bytes, agents aren't watching videos yet. So the volume of bytes is not as large as the volume of requests for instance. What we're seeing is we launched the API schema enforcement yesterday, so what we're seeing is there are side effects of AI, like everyone is creating a lot of code with agentic coding tools, that code is calling network APIs that goes through our platform and we can apply our API products to that. So I think it is more traffic. It's also, as I said before, kind of that second leg, which is, okay, whether it's existing types of traffic like API traffic that we can earn additional revenue based on value that we deliver, or whether it's new kinds of things in terms of AI traffic, CP is a very interesting example of where we may be able to add value. We see opportunities in both those areas. But I don't know if you have -- I can't predict when the tsunami -- I think to some degree, this tsunami is already kind of coming, but Art, I don't know if you have a view on that.

The when is hard. I mean, we just had Muse and Grok Bot show up in the last 4 weeks, I think, right? And that's probably the first real like super consumer-friendly agent environment. I mean OpenClaw is cool, but it is significantly harder for people to set up. So I think the question on when is hard, the privacy traffic that we're seeing is new. There's a new type of traffic that I mean it started a couple of years ago, small, but that is certainly something that 5 years ago or 6 years ago, it wasn't really -- I mean, we didn't have it, it didn't exist. And it's a different type of traffic for sure. On the API side, in MCP, depending on where you sit in the stack MCP, just another API. But it is -- there is some subtle differences that the agents tend to be chatter and they're doing the self-discovery right. They're also using more streaming API interfaces, which is quite different from how a little traditional like REST-based APIs have worked in the past. I think it puts more demand on the functionality of the platform over time. But it is also like so new and it keeps changing weekly. So it's -- the when is the hard part to predict here.

Frank Louthan with Raymond James. So where does the unified platform puts you guys relative to the competition? How much of an advantage is that? And then I had another question on your Fastly for AI platform product. What's kind of new and different about that?

Maybe I'll talk a little bit about the platform question, and then, Kelly, you can take the announcement we made yesterday. We believe it puts us at a pretty significant advantage. If you compare us with some of our competitors who are operating at larger scales than us. We're driving at this point, higher gross margins than they are, and we create that right back to the fact that we're running one network and one platform where they tend to have a platform for sometimes for each product, frankly. The feedback we get from customers is it's a more consistent experience. So they like that. So we think it's a pretty significant advantage. I mean one thing I would note is in a time where hardware is harder to come by and more expensive, the value of that efficiency is even higher. So I think as we're going forward, and as you noticed, I think we started the year with questions about how our gross margins would withstand the more expensive hardware. I think we've basically -- nothing but raise our gross margin since then. So I think it may be a bigger advantage right now given the current economics around hardware as well. By the way, that's another thing about our platform is when Rich talked about CapEx virtually 100% of that is going to expanding the network rather than maintenance just because of the way we built the network. But Kelly, do you want to talk about your state announcement and what's new?

Yes. Very exciting announcement yesterday with [ Fastly AI ]. I'll point back to that state going from 7% to 10% in terms of customers who adopted for more solutions for a reason because it's -- yes, there are some companies do have an AI firewall, to some extent, doing air runtime controls, some doing API enforcement. None are bringing the platform that we have. None have the complete flexibility that we have. None are able to have that elegant again cascade inference where you're able to have that efficiency of logic pass graduate to whatever models you're using in your own cloud along the way. That's really hurt the cloud neutral model metric element that we see driving again the platform adoption. So I wouldn't think of those as individual project products in isolation until they fit into the platform where customers can use them in conjunction with other products there as well.

Rudy Kessinger, D.A. Davidson. I'm curious, [indiscernible] or Artur, if you could expand on compute, it's very clear you guys aren't taking a cloud approach like some of your competitors. But like just what role do you have to play in compute for these edge applications. And what's the monetization opportunity there?

Yes, I can -- I'll talk to that. We believe, as I kind of outlined that there's a bunch of functions that belong at the edge. And as you said, we're not -- some of our competitors are offering essentially cloud compute services. We don't think we can do a better job than say, AWS offering centralized cloud compute services. So we're focused on how do we support our customers with the use cases and functionality that actually benefits from being at the edge and actually running those workloads at the edge in our high-performance network. So that's kind of a high level how we frame that. Today, we essentially have one compute platform -- product. It's an exceptional product. We have customers who tell us that they're able to do things on it from a performance and security and resiliency perspective that they're not able to do any other way. But it's one product that addresses only one for sliver of the use cases, right? And there are a bunch of technical things, but it's essentially event-driven high-performance Wasm at the edge. We have customers who want to run long-running processes at the edge. We actually do that for a large customer on a somewhat bespoke basis because they wanted access to the performance and resiliency and the global footprint that we can offer close to users. So what you're going to see, and I think Kelly alluded to it, is kind of what we did with the security portfolio. If you go back, we had one product. We said we needed to get to critical mass. We need to complete web API and application protection, we did that. Now you can see the growth in the more material business side coming from compute. Should anticipate the same trajectory with our compute where we're starting with one product, and we'll be adding based on what our customers want and need more compute models so that they can support all of their edge workloads on Fastly.

I think Param is there on the back.