Akamai Technologies, Inc. (AKAM) Earnings Call Transcript & Summary

Excellent. Thank you for joining us. My name is Keith Weiss. I run the software equity research franchise here at Morgan Stanley. And very pleased to have with us from Akamai, Executive Vice President and Chief Financial Officer, Ed McGowan. So Ed, thanks so much for joining us.

Thank you for having me. Great conference.

Excellent. I wanted to start and -- with a high-level picture about how Akamai has been evolving over the past couple of years, both in terms of kind of solution portfolio. We've gone significantly from being a CDN company. Now it's a security company, right? And it's going into becoming a cloud company, but also kind of how you guys are thinking about distribution.

Sure. Yes. Good question. I think we've seen, if you don't evolve in the CDN market, it's very difficult to survive. We had 4 competitors go out of business. I give Tom a ton of credit. He got us into security about 12, 13 years ago now, and it's been a great -- we started off in web security leveraging the platform. So that was a huge benefit, whether you're doing application layer DDoS protection or you're doing web app firewall, leveraging the platform is such a key differentiator for us because, number one, very cost effective; number two, it's great scale and performance because you're delivering that web application firewall while you're delivering the content from the same server at the same time. So the user has a great experience rather than having to be backhauled somewhere and slow down the experience. So it was a very natural thing for us to do. Plus we see every Internet user multiple times a day, tons of data and information. And we've grown that business to about $1.8 billion -- the web security business. We broke that out this last quarter. So that was a great pivot for us. We're taking that knowledge now. We're getting more into enterprise security with Guardicore, with our Access product, now with API security, which is kind of both, right, because we brought an application which uses APIs. So very natural from a distribution standpoint, sell through to our web app firewall customers, but also think of how big the market is from a non-web application perspective. So it dovetails with our cloud business as that grows as well. And then the last piece over the last 3 years, we've journeyed into the compute market. And we got there really by -- for 2 reasons. One, we had a cloud spend that was growing out of control because when you're a security company, you process a ton of data. We're doing a lot of compute on that data. So we were seeing our cloud bill grow 60%, 70% a year, and that's just unsustainable. So we said, all right, we got to bring this in-house. How do we do it? We also had customers, 1 very big customer who we've talked about, Apple, that is running private really on us and want to run your own code, and we did have that offering. So we've built something custom and say, you know what, we're getting a lot more requests for that now, and that's not sustainable. So Linode came along; very, very great deal for us. It was pretty accretive right away. It had the capabilities, the broad capabilities. We had to obviously put a lot of effort into it. We saved an enormous amount of cloud spend. If we hadn't done that, our margins -- gross margins will be a couple of points lower. And we have now have a repeatable, scalable offering for customers that want to do anything that they choose to on the platform from a compute perspective. So we also think we're very differentiated again with the platform. We connected the platform to all of our data centers where we have compute -- big compute facilities. We just launched a managed container service now that you can run pretty much at any location. So it's, again, a very nice synergy for us, but there was real clear signs of demand from our customers as well as our own needs to get there. So very logical for us to go down that path. And so the transition has been CDN is in a bit of a decline. The traffic growth on the Internet is not nearly as robust as it's been. And our web security business is slowing a little bit, but our enterprise and API business is growing really, really fast. So we continue to innovate and add new services. And from a distribution perspective, we are leveraging a lot more channels now with the enterprise security. So some of the -- we talked about Deloitte & Touche on one of our calls. I think was on the last call. But we're working with a lot of the big system integrators and the value-added resellers to help us bring those products to market. We'll be doing the same thing with API as well. We've got some partnerships there as well. So hopefully get a lot more scale out of that as we go.

Okay. I want to unpack that answer. It's kind of in a couple of dimensions. And starting with the evolution from kind of CDN to DDoS protection and web application firewall. It seems super logical in terms of like you have this network that can deliver all this traffic...

Absorb. Exactly.

Basically, you're kind of running in reverse. By and large, very similar customer, like, the first one who's buying CDN is going to be also the person who's buying web application firewall. The question in there is, in that -- like understanding sort of CDN business has its pressures on it, when you think about web application firewall, when you think about DDoS where you guys are industry leaders, how much more room for growth is there in those markets?

Yes, it's a good question. I think there's still room for growth in the sense that our existing customers generally will start off with their mission-critical apps. So we see more and more customers standardize and have all of their applications protected by a web app firewall. So there's existing growth inside the existing customer base. There's also new innovation. So for example, 2 years ago, we came up with Page Integrity Manager, which protects the third-party content on a site. So WAF protects the base page, Page Integrity protects the links and all the third-party content, which is an attack vector that can be exploited. And so we'll continue to innovate. There's other products like bot management that we came up with that was -- effectively what's happening is most of the traffic online today is actually machine and not human. So giving our customers the ability to understand who's interacting with their site and then giving them tools to deal with it, block and attack or if somebody is doing nefarious things, you can give them, say, price scraping bot, you can give them a different price list, right, trying to fake them out and whatever. And then we're now invented a human front product to know that this is Keith Weiss' credentials, but it might not be Keith because I just saw him log in over in Australia and now he's trying to log in San Francisco an hour later. That's not possible. Something doesn't make sense or the way you're typing is different. So we continue to innovate, and I think there will be continued innovation in that suite of products. So we do expect that product line to continue to grow. And there's more websites coming up online every day and that sort of thing. So there'll be -- but we just don't expect it to be growing as fast as it has in the past, very high penetration in the customer base.

And it seems like the commonality of most of these solutions is it leverages a lot of the data that you guys have in the network. And jumping to Guardicore, Guardicore has been a very successful acquisition for you guys. You've been able to grow that business well. It was a great technology, even before you acquired it. I had run across Guardicore, and it was very well regarded. I'll be honest with you, like I don't -- you guys have done well with it, but I don't understand how it fits in that industrial logic in some way, right? It's like it does -- it's not necessarily the same buyer that necessarily leverages the network. Like how did you guys make that leap of like Guardicore makes sense?

Yes. So we kind of went down the zero trust path, and it makes sense. Like you think about like our remote access product, for example, a much safer way rather than getting in behind the firewall, you have the application dial out. Any time you move security to the cloud, you run it through performance challenges, right? So if you're doing secure web gateway, for example, if you're proxying traffic back to handful of data centers, the performance for your users is going to be terrible. So we got into the zero trust market saying, we can leverage the network and the data and all that stuff in the same way. Guardicore, we're looking at micro segmentation as sort of rounding out a portfolio. And it just was an opportune time to pick up the technology. And what's interesting with Guardicore is no matter how good your perimeter security is or you might have endpoint solutions and have a mill phishing and all that sort of stuff, something always gets in. And with -- especially with AI now, the attacks are getting more sophisticated. The deep fakes are so much better that if you have a user that inadvertently clicks on something, the malware gets in and spreads. And that's where the damage happens. So what Guardicore allows you to do is segment your network using agent -- basically agents, so it's not deploying lots of hardware. And you can basically look at -- discover what's going on in your network. So if something gets in, you can discover it. You can set up rules to basically manage traffic flows as it goes across your network. And it was part of an overall vision of having one pane of glass to do all the different things that you would do for zero trust. It happens that that's a market-leading product. And that's what we found was we had much better success with that than we did with say Swig. We were a little bit later to market with that product. We still sell it. It's not as big of a demand for us, and we're focusing on micro segmentation. I think the market really is pretty much anybody, right? It's really a good hygiene. We're seeing financial institutions, pretty much all of them are adopting that. Governments starting to adopt it now. But it's something that really gives you incredible visibility of what's going on, on your network and limit -- just limits the damage of an attack.

Got it. Got it. Guardicore, I mean -- and zero trust architecture, somewhat different buyer from your traditional buyer who is kind of managing the website. And this necessity is kind of building on a little bit different type of a channel.

Yes, absolutely.

How far -- like that effort is mostly complete, though, right, in terms of getting into those additional channels and building up that...

Yes, for the most part, there's still a few. We always want to continue to leverage the channel model as best you can and work with new channels, come up with different ways of going to market, but we've done a great job there. Most of our business, with the exception of our installed base, really, comes from the channel. So -- and you're seeing a lot of nontraditional CDN logos that come in through the channel for Guardicore, which is great because now we have solid API security, solid compute. And yes, you do have a different buyer. One of the trends that we're seeing, and we're leveraging this a bit with our CDN customers, WAF customers, is -- you hear a lot of talk about platformization or really buying from fewer vendors. It's a trend. I run IT at Akamai so I focus on that, and we're trying to cut down the number of vendors we use. You have different point solutions. It's expensive. You have -- a lot of companies will get acquired and all that stuff that gets kind of messy. So using fewer vendors is definitely something that we're doing and we're hearing from a lot of customers. So we have that relationship with the CISO, with our web application firewall business, bot business, et cetera, and that can be leveraged to get to the right buyer inside the company. So we're doing some of that as well.

Got it. Got it. And then how should we be thinking about the sort of the durability of growth within that security segment? I think you targeted like 10% growth both in terms of like the current year, but also you guys see that being durable on a go-forward basis. How should we think about that equation of durability? Like what's going to enable that?

Yes. I mean we've had a great run so far. And we've been largely producing a couple hundred million of security revenue every year. So we have the capacity to do that. I think the mix will shift a bit. So we're not expecting as much growth from the web portfolio. We think it still will grow. So that will be driving some decent growth in terms of absolute dollars. The percentage will be a little bit lower. And then the Guardicore and API business is very large opportunities. API, if you believe the analysts -- no, not you guys, but the product analyst folks. They think $3 billion market potentially by the end of the decade, right? So that's a big market. We have a market-leading product there. I can tell you that most customers are asking about that because there's really -- it's a huge attack factor that's being exploited very often. And we're getting a tremendous pipeline, growing that very, very fast. That could be a $1 billion product for us. I think Guardicore could as well. So I think over time, what will happen is you'll see -- as the web business slows down a bit, the Guardicore and API business, which is growing much, much faster will become larger and obviously maintain that growth. But we think those are $1 billion product potentials. So maybe even more.

So if you think about it, sort of holistically, you're trying to put together kind of waves of growth, if you will. We -- Guardicore and API management, sort of the ascending wave of growth that can sustain growth as the more mature products start to slow down a little bit. And the -- I guess the extension of that is should we be expecting additional acquisitions as like new opportunities that present themselves, you will continue to bolster out that security portfolio to garner the next wave.

Yes. No, it's a really good question. And I'd say the answer to that is yes. I -- there will be a combination of innovation, but also some acquisitions. And if you think about like one of the big challenges of being this big as a company of 25 years old, investors want growth. If you go and invest in building, let's say, we want to build our own version of Noname. It takes years to get the product out and growing to the size that it is now, 570 or something like that. So by doing -- I think we do acquisitions pretty well, not everyone is a success, but we've had a lot of success with them. And getting a company that's around the Guardicore size, the 20 -- I think it was 20-something million when we bought them, and Noname was around 40. Getting them at that size, where they're at the point where they're now looking at major investments in go-to-market to really get to scale has been very successful for us. And so we'll do things more likely in the enterprise security side. I think the web security side will probably be a lot more innovation. So as new threats emerge and we make the product more robust to handle whatever new threats come along, I think you'll see a lot more internal innovation there. But if we have the opportunity to get some market-leading products that are adjacent or make sense from who our buyers are, we'll certainly take a look at it. I don't think you'll see anything massive from us. I think it's same type of size, a couple of tech tuck-ins potentially and then kind of that acquisition of a couple of hundred-million-dollar size acquisitions, but not transformative. I don't see us doing anything like that.

It sounds like to me like on the web side of the equation, you guys have a ton of data that's flowing through your network. You get a lot of visibility on that. The -- like within security, it's account masking, right? Like there's new threats that come out. You get visibility into those threats and you have the capability to build solutions effectively against that. And that's what's going the web security. On the enterprise side of the equation, it's going to be more opportunistic to leverage your distribution channel leverage the customer base to be able to feed into that consolidation theme of being that...

Being a trusted vendor and leveraging that. Absolutely.

Got it. So when you're thinking about those acquisitions, and this is me taking an opportunity to test my theories, which -- it's -- I think about it almost in terms of like an area under the curve, right? Like if you're going to develop this solution, it would take you 2 years to get somebody in the market and another year to get traction. If you acquire it, you get the technology upfront. Yes, you have to pay $200 million, $400 million for it, but you have those 3 years in the market with much more momentum to begin, and if the area under the triangle is less than the $400 million, it makes sense.

Right. Absolutely right.

Got it. So I want to shift gears and talk about the compute side of the equation. And the interesting thing about compute is that there was a capability in Akamai already. You guys had programmability on the edge. You had the -- your customers were taking advantage of that. There was functionality that they were running on the edge. How does sort of the compute at the edge capability, how did that compare to what Linode brought into the equation?

Yes, sure. Good question. So the compute at the edge would be -- I'll use an example of, say there's a company that wants to start up a video service, right? So they need to obviously procure a lot of hardware or potentially use a cloud. And so what they do is they say, okay, I need to build a content management system, I need storage for all my content, I need encoding, transcoding. So there's a whole bunch of compute needs for the customer. Now let's say -- you could build that on Linode. Now we have obviously a lot more robust now and it has a lot more features and that sort of stuff. You could build that out on Linode. You can't build that on the old Akamai system. But what you could do, let's say, for example, they said, hey, I want to have on my website, I want to have a waiting room application because I'm launching a new video or whatever and it's going to be wildly popular. And I don't want to overrun my log-in server. So can you build me an application that someone shows up and it says, you're next in queue or whatever? Sure. You can build that using JavaScript and put that out on the edge. Our edge servers can do that computation and do all that work. Or so you want to do A/B testing. You're doing -- you're showing ads to a different audience, you want to understand what's happening with -- who interacts with what -- whatever. Use JavaScript for that as well or web assembly, whatever. You can do that, but that's sort of more of a function as a service, but that full stack of building and running the actual -- the website itself, the commerce engine, the coding and transcoding, we didn't have that functionality. Linode brings us that functionality today. So you can -- so if you think about from a spend perspective, let's say, I'm spending $100 million a month on a big property, probably 80% of that or so is going to the compute vendor. Maybe 10% of that is going to the CDN. And of that, maybe a couple of million would be for that function as a service. So think about the opportunity size is so much bigger. And now one of the things that we're bringing to market that's a little bit different is the distributed nature of our computing is more distributed than any other platform. And you might say, well, why does that matter? Well, latency is a big issue. And like we -- Tom talked about this on one of our calls, we had a financial services customer who used us to comply with Apple Pay. They were using a hyperscale, they couldn't meet the latency requirement, but they could with us, right? So that's just one example, but there's a lot of different examples there. But it just opens up a much bigger market for us. And the other business is still growing. We do have customers that come to us and ask us for things to basically augment their website and do certain functions that you can run at the edge, that makes sense to run at the edge, but it's not a replacement for your compute spend.

Got it. Got it. And if you just think about sort of the scale in terms of your plant, 3,500 points of presence?

4,300.

4,300 points of presence. Right, general.

Sorry, I work with mathematicians, so I have to be accurate.

So the 4,300 points of presence, you could run those functions in all...

Pretty much. Yes, yes, you could, yes.

Today. How do we think about that scale versus Linode? Like does Linode have to be like -- will it ever be 4,300?

Probably not. I mean I don't think it makes economic sense to do that. Technically, there's no reason why we couldn't do that. We just launched -- last week, we announced our managed container service. So that's -- we're doing some testing with the customer right now in over 100 locations that you can run your code in those managed containers, right? So if you have a node functionality, we're leveraging the platform now, we can roll that out, people can spin that up as they want. So we will be taking advantage of the scale, but it's really going to be dictated by the customer demand. I think technically, you can do it. It just doesn't make sense to. Some of the sites we have are pretty small. You may have a rack of 5 servers that's serving to a college campus or something like that. So it wouldn't make sense to run something there. But we do have some pretty big deployments where it does make sense to be very widely distributed.

Got it. So it's -- Linode is an augmentation sort of that compute capability on a go-forward basis?

Yes. Absolutely.

It's a cost-saving initiative for you guys. You bring your in-sourcing and some additional capability. In terms of the end customer, to what extent is the Linode customer the same customer that you guys were serving before versus sort of an extension that you're going out to a new customer?

Yes. So the acquisition we did, there was, I'd say, very few -- well, there was a few customers of ours, but it was mainly developed for the developer, the hobbyist, the guy that -- the CEO who started the company said I designed this so I never had to talk to a customer. So it's super simple, easy to use, right? Swipe your credit card, you're in a virtual machine in 5 minutes, right, up and running. And we did -- we do have a bunch of customers that have told us, I use that for myself. I use it for my demos or something. I'm doing an all-hands meeting or something. And I never thought of using it as -- for the company because it was just a small company. But now that you guys are doing it and you're adding all the capabilities, you connected it to your backbone, you've got PCI compliance, you've got all this other capability, that's interesting. I also like the platform. I'm interested in using it. They had 150,000-ish customers at the time. So a very different business than what we do. We have 8,000 really big customers, biggest web properties and stuff like that. We're adding a lot more customers now with compute in different verticals that we don't typically sell to, but don't really think about their website all that often. So it's -- there's some overlap there, but it was a lot of new customers. And some of them have blossomed into pretty good sized customers as we add more functionality. And all those customers get all the benefits that we're adding. There's a lot more locations now. There's better performance. There's more security. There's -- starting to build out an ecosystem of ISV partners. So it's been going great. So to me, that's the -- probably it will be our biggest business at some point in the near future.

So you can almost think about Akamai in terms of, when we're talking about security, there's the ways of growth in security that's going to enable you guys to sustain that 10% overall growth in security. You were talking about -- was it 20% growth in terms of compute?

Compute, yes.

As a similar type of kind of ways of growth, you have your core kind of capability in the Akamai network now you have one...

Exactly.

A much bigger part of the equation. And then we think about CDN, kind of security, compute the -- as we shift more and more to the compute side of the equation, just as that's growing faster, it becomes a bigger part of the business. The mix shift should support better growth for optimizing [indiscernible].

It should. That's exactly right. Exactly right.

Got it. I'm sorry, double clicking into the -- just one more on the compute side of the equation. In terms of distribution, so it's no longer just the, like the self-service model.

Correct. It's still available.

It's still available. But you guys are looking to add more of a kind of outbound motion?

Exactly.

Where are we in kind of building that up with second [indiscernible]?

Yes. So we are -- we hired a gentleman from AWS who runs our sales overlay team. We're building out that a bit more, probably have another 50 or so people added to that. All of our reps sell compute. So this year, we changed our comp plans pretty dramatically. You're not going to make your numbers if you don't sell compute and security. It's a -- we're going to be hiring a bunch of hunters as well because my ROI on hunting is so much greater. If I were to sell to, say, a big oil company. If I'm selling them CDN and web security, probably not going to come out with a big order, right? But if I'm selling compute and enterprise security, it's happy hunting. There's a lot of opportunity there. And we have a differentiated offering, et cetera. So the ROI on our hunting model will be much greater. So we're going to be investing in some more hunting as part of the distribution channel. We also do have a -- stable channel partners that will bring us business over time. There's an opportunity for SIs, actually. We think about how the process we went through of moving all of our compute off of a hyperscaler. We're about 50% cheaper that list. So you're spending a couple of hundred million. If you want to move to us, we can save you a bundle. So if you want to put a process in place and go through the change management and the whole process of moving and doing all the inspection and all that stuff, an SI could sell that as a package, right? And there's a lot of consulting dollars there for them with a huge ROI. So that's another possibility.

Got it. And when you think about your competitive dynamic, I remember when you guys first acquired Linode, the position was like, listen, we don't have to sell against AWS, right? There's a lot of business being done in compute. We could take some of those small parts of the business that AWS hasn't been focusing on. As you guys have been able to improve Linode, if you've been able to upmarket as you have more distribution, does the purview on like competing with the big guys change at all? Like do you now look at like -- listen, we could come in and present a value proposition for 40%, 50% cheaper against an AWS or an [indiscernible]

Yes, we are getting -- seeing a lot more wins now from the hyperscalers, not massive wholesale changes where somebody like, say, an airline moving their reservation system or anything like that. But it doesn't mean we can't sell to them, so that you could get in there and get additional applications, there's use cases where it makes perfect sense for us. But we are seeing -- we go head-to-head with some -- especially with the performance-based items like, say, someone who's doing ad decisioning, using an ad decisioning engine, much better use case with us, much better performance. So we are seeing some competitive takeaways. And also customers that have data that -- in a cloud that's accessed very frequently so you store your data and you're accessing it frequently. There's very big charges, or large charges, expensive to take the data out. If we actually built some products for our customers, which we call the origin services, where we might be getting a 99% cash hit rate, which is pretty good. But that 1% cash miss, if they're origin's in say, AWS or Google, it might cost them 3x more than the CDN bill because it's so expensive to pull it out. So we offer that for free because we've got one of the largest backbones in the world, from what I understand. And we've connected that so I deliver hundreds of terabits per second of traffic. So the small amount of traffic that's coming off from the platform doesn't cost me anything. It's a big value add for customers that have an egress problem. And some people just use storage. It's just literally just moving storage to us, and that's very easy to do.

Great. And the -- we think about a lot of these Infrastructure as a Service products, the -- it is a sort of commodity underlying infrastructure. Like the compute is relatively commodity. You're talking about selling storage. You're able to sell it for 50% cheaper, but you guys still have 30% operating margins, so you're seeing a degradation in operating margin.

No, we're not.

How are you triple off that trick? And it's not like we're seeing Microsoft or anybody else run at like 60% operating. How are you able to come in 50% cheaper and still make a decent [indiscernible]

Great question. A lot of synergy, operating synergy in the business. So we've moved about 1,000 people out of CDN into compute. So they're doing anything from qualifying hardware, different hardware platforms. They're doing the distribution. So they're going and buying and building the network effectively. We have engineers who are doing it that are experts in things like compliance requirements and they're leveraging that expertise. We have developers who are developing some of the new functionality. We had -- for example, we had a bunch of storage engineers building a legacy storage product that we moved into build the Linode storage, right? So there's a lot of synergy there. We have huge economic advantage in a lot of our locations where we don't pay for bandwidth. Sometimes we don't pay for colo and electricity. And we can leverage those in someone like in a managed container environment where we can leverage some of that. Now obviously, we can't get too big there because at some point, they'll start charging us. But there's some big advantages there. And then also just the margin in that business is massive. Now these guys are building massive plants and like sometimes their own power plant and spending a ton of CapEx. So there's obviously a timing issue there. But our CapEx is very reasonable now for the first several billion dollars of revenue. Maybe at some point, we go get a little bit bigger and maybe there might be a hit to margin at some point if we're investing way ahead of growth at that point. But right now, we are seeing very, very healthy margins because there's just so much profit in that business.

Right. Got it. I would be kicked out of the software analyst club if I didn't ask you about generative AI. So where is the -- sort of how do you guys think about the opportunity, perhaps within Linode in terms of servicing some of that demand from generative AI. And what parts of the equation do you think that it fits best for?

So training, no. We're not going to go after the training market. You need a ton of CPU and you see companies investing billions and billions of dollars on that exactly.

Big, giant clusters. Right.

Exactly. Huge clusters. But the infra side is something that is we're starting to get a lot of wins in actually, and you're seeing some pretty interesting use cases. We have like a search engine that's doing like on-the-fly image manipulation and things like that. We have a text to -- an application that's doing voice text to image. So like you could say, build me a whatever, picture of an elephant or something and it draws it for you. And that inference is running on Linode in a distributed fashion, which makes a lot of sense. Customers are using GPU with us. You can use -- you can do some of these inference models you're capable of doing with CPU and really smart algorithms, you could do that, but a lot of customers are opting for GPU. So we are starting to see a lot of use cases where we're winning, and advertising is a good example of seeing more and more inference use cases coming to us. And the nice thing is, some of these customers I've never heard of. So it's like companies that just tells you how enormous the opportunity is. When you think about our web business, it's really the biggest banks, the biggest commerce, the biggest media. But for compute, it's everything. And even start-ups are big customers, right? Because you think about you raise around of $20 million, you're not going to go spend it on buying your own servers. You're going to go to a cloud and eventually it becomes a problem. That's usually when we buy them, it's a problem. You look at the spend, it's usually the biggest line item next to people. And it's usually growing much, much faster than revenue. So there's a lot of opportunity there as well.

Got it. I always give you a hard time about margins. Can AI help on that side of the equation? Are there kind of internal use cases? Are there ways to sort of drive -- squeeze out additional efficiencies out of the business using AI on a go-forward basis?

There's an interesting case we're running now where we've built a large language model for Guardicore. So Guardicore is a pretty heavy lift when you go to install it. And we use it ourselves. It's great. But it's a heavy lift. So going through the discovery phase of understanding what's going on in your environment and going through the installation process can be difficult. But -- and we usually have a lot of services revenue that goes along with it. We can potentially get some good scale on that in terms of, hey, your service margins are decent, but you don't want that business to become too big, right? So that's a possibility to potentially scale Guardicore by leveraging that, just to have better customer satisfaction but also just get some leverage on services potentially. I always say the scale and engineering is funny. When you talk to those guys and say, no, we actually need more engineers, and you want to -- why is that? Because there's more QA and all sort of stuff. But I do think there will be, at some point, some efficiency in engineering, a little bit in sales with some of the tools that we buy from vendors like a Salesforce where you can get a little bit of a productivity gap, hard to measure, but there's potential there as well.

Got it. So maybe to wrap up in the last 4 seconds that we have. I'm gonna ask you a big question for 4 seconds. We talked about sort of the kind of the wave of growth and sort of the mix shift on your revenue base being a positive kind of driver over time of that. We have a lot of confidence in the durability of compute growth and that 20%, and security is going to be a durable at 10%. And the negative impacts of CDN start to wane, partly because it's not going to be as bad on a go-forward basis, pricing getting a little bit more moderate, but mostly because it is getting it to be smaller. So there's almost a mathematical equation for improving kind of top line growth over time. The -- I guess the question is here is like what's the time frame that investors have to think about in terms of where we're going to be talking about Akamai as a double-digit top line grower but also double-digit like earnings grower on a go-forward basis?

Yes, I think you can get to the earnings grower faster because you're going to get margin expansion. And also we -- I said on the call over the last 10 years, we bought back 16% of our stock. So there's -- the share count has gone down by [ 60%. ] So there's opportunities there as well. So between buybacks, margin expansion and top line growth, you can get to double-digit earnings much faster. Top line growth it could be a few years out, and the ingredients are very simple. The -- probably the fastest thing will be getting the CDN business back to sort of a low double digit -- low single digit, excuse me, to flattish type. So that's going to just stop being a detractor of growth. Security, we think, as we talked about 10%-ish, we think that's durable. Compute could be even bigger. So if we -- traction and compute, you have the opportunity to potentially do it sooner. It could be a couple of years before the top line gets there, but I think you can get the bottom line there a bit faster.

Ed, thank you for joining us.

Thank you, Keith. I appreciate it.