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

Excellent. Thank you, everyone, for joining us this afternoon. My name is Keith Weiss. I run the software -- U.S. software research practice here at Morgan Stanley and very pleased to have with us from Akamai, CEO and Co-Founder, Tom Leighton. Tom, thank you so much for joining us.

Nice to be here.

Before we get started, for important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. Excellent.

So with that out of the way, again, Tom, welcome to the TMT conference this year. And it's, I think, a really exciting year to have you at the conference. You guys have been great at coming every year. 25-year history but it seems like the rate of change, if anything, is accelerating at Akamai with the investments that you've made, particularly in Linode, particularly like in security and expanding the purview of the company. So can you talk us a little bit through that journey over the last couple of years of how the -- really the purview of Akamai has really expanded.

Yes. Well, obviously, we started as a content delivery company. We started that marketplace about 25 years ago. Last year was significant because it was the first year that security became our largest product line, so a big shift for what was known as the content delivery company, the lead products there, web app firewall as a cloud service, also a market that we started a little over 10 years ago. And today, as we look to the future, very excited about security, a lot of new capabilities there, but now also with the investments in cloud computing that, that we're hopeful is a big source of revenue and growth for us in the future. We have the 25 cities with core compute data centers. And we talked about recently with Project Gecko, adding 75 more cities with our edge PoPs to support VMs and containers and ultimately, fully serverless and automated cloud computing, which is, I think, a very exciting concept.

Right. In that evolution -- and one of the things that I think always been really interesting about Akamai is how natural kind of like the expansions have been, right? So you started as a CDN service, you were pushing content down. And then DDoS is basically just everything going the other way, right, like utilizing the existing network. You have all the data on what's going on, on these networks. Performance becomes a really obvious -- maybe obvious is the wrong word, but natural area to extend into a lot of the security products the same way. Can you talk to us about that natural extension? What made Linode, what makes this compute side of the equation that similar kind of natural evolution of the story?

Yes. Well, we've always been interested in compute. We started edge computing under that name over 20 years ago and have supported function as a service on our edge platform and 4,000 PoPs for a long time now. The big recent shift is being able to support VMs, containers, Kubernetes, the full stack compute and storage. And that was, of course, aided by the acquisition of Linode. And now what we're doing is taking that base and applying all our know-how with delivery and security to make it be really scalable, really local and fully automated. Today, in the cloud, you've got to pre-provision how many instances you want in each data center. And that doesn't scale so well, especially if you're doing a commerce or media event where you get a spike crowd. And if you haven't provisioned it just right, you got an outage. You got a situation. And with Akamai, of course, we do all that in a serverless way historically with delivery and security. It's what made the Internet scale, and now we're applying that to compute, which I think is a really exciting place of next-generation cloud computing.

Right. So given your expertise in security, your expertise in performance, being able to understand loads and balancing loads, you could take an asset like Linode and push it further up market into your more sort of higher-end customer base.

Yes. No, the Linode service, small and medium enterprise, developers, very popular, easy to use, very inexpensive, but wasn't really suitable for mission-critical apps or big enterprise. And that's what Akamai is all about. So we've gotten that to the point now where it really is much more reliable, a lot better capabilities. And now we're scaling it out into our edge regions. So we'll be in 100 cities by the end of the year. And then next step is fully automated. So we'll spin up your containers and VMs on -- based on user demand on demand. We do that for function as a service today, have for a long time but never been done before for real compute, which is where we're going.

Right. So I got ahead of myself a little bit. We're going to dig into the cloud business. But I wanted to start with kind of a state of the union, like where we are in terms of demand trends across the various businesses. And maybe start with the sort of the core delivery business. Traffic has been volatile, to say the least, over the past couple of years. You had a traffic spike during COVID, and then there was the sort of overhang from that. How are you feeling about the current demand environment when it comes to the delivery side of the equation?

Yes. The traffic growth rates are up this year compared to last year a modest amount, not a lot. They are not back yet to the growth rates we experienced pre-COVID, and we're not assuming that they're going to return to that level. And the other component to revenue besides traffic growth is unit pricing. And so we are being more judicious in unit pricing. And as we've talked about, not taking on all the spiky traffic, which is not as economical for us and areas that cost more or the really low-priced stuff, we're not taking as we might have before. And I think that's the benefit of the company as a whole, especially as we put more investment into compute and security, which have much better ROI. And we'll see how the market shakes out over the course of the year. We have seen 2 of the major players give up. You've got another one out there today, more or less in free fall. And so we'll see. Maybe at some point, we'll get some rationalization in the pricing in that market, which would be a great thing to see. Long term, of course, we'd love to see delivery stabilize.

Got it. Historically, we'd always look to specific catalysts like elections and Olympics, which are both coming up in 2024, as a good sort of driver for the delivery business. Any reason that wouldn't be the case this year?

Well, Olympics is better than the normal event because it's over a few weeks, and so you get a little bit more revenue. But think about it as million, so it's not a huge swing. I think the election season really depends on what happens. Hopefully, we're not in some crazy drawn-out thing. We're all glued to the news every day for months, but you can see traffic increases associated with an election season depending what happens.

Got it. And then you recently acquired customers from StackPath and Lumen, not the companies but just the customers and sort of the contracts that you wanted. How should investors think about that as a sort of addition to the growth paradigm and delivery as we go into FY'24?

Yes, purely opportunistic financial benefit to the company. We integrated the customers we wanted within a couple of months. We didn't buy all the contracts, just the ones that we wanted. And we did add on a bunch of new customers that we're looking now to cross sell our other capability, security and compute, too. And we get a very rapid return on the investment there.

Got it. Shifting gears, I want to talk a little bit about the security demand environment. We were hearing from Nikesh Arora from Palo Alto earlier, and he was talking about a little bit of security fatigue of that there's been a lot of security buying over time. Security obviously remains a high priority because the threat environment is out there, but perhaps we've acquired a lot of solutions than that, giving -- getting us as safe as we had hoped. What are you seeing out there? Are you seeing that type of dissolution, if you will, in terms of the security solutions? Or is this still a pretty robust environment for you?

No, it seems like a strong market. We've guided 14% to 16% growth in constant currency. Last year, we did 15%, so pretty stable guidance on a bigger number. What we are seeing is that the penetrations are rampant and increasing and no matter what you buy. And it seems like it's getting worse because of GenAI where you can more easily train up malware, bots to do really bad things and get around the defenses. So I think the ability of the attacker is a lot stronger than it was before, and in a way, that really does help us with our Guardicore sales where we're the market leader because attacks are going to get in. It's just the way it is now. And it really is important that you identify when you've been penetrated quickly and that, proactively, you block it from spreading. Ransomware is really only dangerous when it spreads to everywhere and locks everything down. And data exfiltration malware, if it only got to your HVAC system, it's not a problem. It's when it gets all your data and exfiltrates it, you've got a big problem then. And that's what Guardicore is great at doing, identifying when you got a problem and proactively keeping it from spreading. And yes, you're right because I think enterprises have bought all this stuff and they're still getting penetrated. But that's, I think, going to keep happening.

Right. So the -- you guys have been, actually, really good at sort of -- your acquisitions are -- seem to be very well timed to kind of what's going on in the threat environment. You talked about ransomware as it become a really big problem. And really, the prophylactic for that is micro segmentation. That's where Guardicore comes in. I was reading over the weekend, APIs are becoming like the new threat vector for stuff coming on. And you guys have been doing more in API security as well. Can you talk to us a little bit about the offering there and sort of what you're seeing in terms of that uptake?

Yes. API security is really a new area. It's, the last few years, a lot of attacks coming through APIs that are exposed and have vulnerabilities. And I think the state of any major enterprise today is they've got a lot of APIs that are exposed, some of which that have vulnerabilities and the CISO and the CIO has no idea they exist. I think we're at the point now -- in most of my conversations with CISOs is they now realize they've got a lot of exposed attack landscape. And they need something to help them with it. And so the first thing that our solution does is tell them all the APIs they've got. And then it tells them known vulnerabilities in some of those APIs and ultimately can be used to block it. We've integrated it with our web app firewall, so easy on integration. Literally, that's flip a switch. And we're having great proofs f concept and early traction, a product that we just announced in Q3. Just last week, we did a proof of concept with one of the largest commerce companies in North America. And they knew that they had exposed APIs. They wanted to know which APIs they got to go lock down. So not only did we do that, but we discovered that the most important API they had, which was involved with the user reward dollars for using the site and buying stuff and they build up cash balance, they can buy stuff for free, was being exploited actively. So that was a very successful proof of concept. Not only is it exposed, not only does it have a vulnerability, but it's -- they're being ripped off currently.

You can see the dollars flowing out.

You can see the dollars going out the door. But that's -- so it's a great time for API security. And the competitors are start-ups. And they actually were a little bit larger in their revenue than the company we bought, but we really like the team and the company we bought, and we're investing a lot around it so that now we're in a great position with competitive bake-offs. I think, over the longer term, API security should be a market like web app firewall has become, so a large market over time, and we want to be the market leader there.

Right. In terms of web application firewall and DDoS, those are the original security markets that you guys had gotten into. What's the -- like how much remaining market opportunity is there on that side of the equation? Because I think that's one of the investor concerns, is that web application firewall and particularly DDoS is -- it's pretty well penetrated.

Yes. So DDoS, if you think about it like Prolexic and protecting DNS, a relatively small fraction of our total security revenue and modest growth. And it's -- you're right. It's been around 15 years. So not a lot of future growth projected there. But web app firewall and all the products we've built on top to create our app and API security segment, that's the majority of our revenue growing at a good clip. And the WAF itself growing, but you're right. That's been -- we created that market over 10 years ago. WAF itself not growing so much. But all the things we built on top, doing very well. Bot management now over $0.25 billion a year, growing at a good clip, market leader; Account Protector, new product, but it tells that if a person coming into a bank account with the right credentials, is it the right person or a crook that stole the credentials? Maybe it's a human but not the right human. Client-side protection so that you protect -- we protect our customers' end users from malware that got into their digital supply chain and the big examples of where that's happened and now a requirement for PCI compliance starting next year and had the leading solution there. API security, we talked about, I think, sits parallel to WAF, but it's in that pillar. Just starting but I think it has potential for a lot of growth. So that's, I think, a very strong segment for Akamai. And then the third is inside the firewall, Zero Trust led by Guardicore. And this year, now we're putting that on the same pane of glass and control with Enterprise Application Access. And those are the north-south and east-west as they're called of protecting an enterprise and will be unique in having market-leading products in both and now in the same pane of glass.

Right. So I'd love your view on an industry debate that we've been having within securities out of best of breed versus platforms. You have pulled together a lot of best-of-breed solutions to put together an east-west platform and a north-south platform, if you will, right? Do you think the industry purchasing behavior is starting to move more towards buying a platform? Or are your customers still looking for like we're just going to buy the best-of-breed solution to protect this channel, we have a problem with APIs, we're going to buy API security channel?

Well, it's both. They'd like to have 2 or 3 vendors, not 100 that they have today for a major enterprise. And that trend to wanting to have a few trusted vendors helps us because we are the leading platform play for app and API security, the leading web app firewall by far. Early days in API security, but we have the potential to be the leader there. All the things we talked about, we built on top of the WAF. We're the market leader, and that is viewed as the platform, the go-to platform. And so that's, I think, very helpful for us.

Is there anything you're doing from a go-to-market perspective to incent that platform buying behavior?

Yes, sure. Packaging together, cross-sell, yes, absolutely.

Got it. And then finally, on the -- like security is always incorporated -- the indirect channel partners have been a big part of the security buying for a while. Can you talk a little bit about your current relationships and investments that you're making to sort of broaden out that channel community?

Yes, channel is really important, especially in security. For example, within the enterprise and segmentation is channel only. API security channel first be very heavily channel. It's more on content delivery that was more direct, but security has always been heavy channel and increasing, I would say.

Got it. Got it. Shifting gears to digging in on the cloud side of the equation. Last quarter on the Q4 earnings conference call, you announced a new initiative called Gecko. Can you talk to us about what Gecko is all about? I think you phrased it as one of the most exciting announcements that you've made at the company in a decade. What makes this so exciting?

Well, we're going to take the compute stack, VMs, containers, Kubernetes, and we're moving a lot closer to the end user. It's not just going to be in 2 dozen cities. And by the end of this year, the goal was to be in 100 cities. We're going to be in a lot of cities where there's no hyperscaler presence. In fact, our early adopters of our cloud platform had a big say in a lot of those cities because they want their VMs and containers in those cities, so they can get closer to their user base. And there's no hyperscaler there. They can't get it with a hyperscaler. And they'll be able to have that on Akamai.

Got it. So last year, we were talking about Connected Cloud. Can you talk to us about the evolution? Like how has the solution evolved from Connected Cloud to Gecko?

Well, Gecko is part of the evolution of Connected Cloud. So the vision for Connected Cloud has always been this, and of course, there'll be more and more cities that gets more and more distributed. And then the next phase, starting, it won't be this year, but starting in next year, is to make it truly automated. And one of the barriers, I think, for the hyperscalers to be in a lot of cities is because you've got to provision every city separately. And it's one thing to think how many do you want in U.S. East, how many instances. It's another thing, oh, my goodness, there's 100 different cities. How many instances will I plan? And do I have the right number? That's a nightmare. And our goal, it's just automated. Like we do already for JavaScript apps. We spin those up in few milliseconds based on user demand, so instant scalability. You don't have to think about it. You just give us the JavaScript app, and wherever you need it, it pops up and does the job for you. And compute should be that way. And that's what we're excited about doing.

Got it. So just to dig on that a little bit because we hear a lot about edge compute from a lot of different vendors, right? When we're talking to hyperscalers, like an AWS or an Azure, AWS is talking about local zones or Azure is talking about edge zones or stack edge. What's the differentiation between sort of Gecko and that Connected Cloud versus what the hyperscalers are trying to do with those offerings?

Yes. Akamai is different that we come at it from edge first. I mean we've been architected really at the edge since the beginning using the term edge. We coined the term 24, 25 years ago. And the hyperscalers have started at the core. And so even as they build out their local instances, it's in a hierarchical fashion, and it's controlled by the parent in the local core. It's not something that you think about as just there's all these instances in a distributed fashion. So the whole architecture and ease of use is different. And of course, once it's serverless, that totally changes the game, the way it is today with our JavaScript application.

Got it. So the hyperscalers are more -- almost like a hub-and-spoke models, where you're truly distributed network.

Right, right.

And then what does that enable for your customers? Like what are the customer benefits? Or what are the use cases that would be a lot better served by the truly distributed network versus that hub and spoke?

You get a lot closer to the end user, to where the data is, where the devices are, and that translates to better performance, better scalability and also lower cost.

Got it. We made it 25 minutes without talking about generative AI but...

That must be a record.

It is. I made it 27 minutes into the presentation yesterday. But it does seem like the Gecko initiative and edge AI could have some -- or the edge compute could have some interesting generative AI use cases. Is that true? And sort of what's the work being done to help to enable that?

Yes. So most of the cycles with AI and generative AI are the actual application of a trained model. And many of those, you want to do locally. You get better performance. No need to truck everything back to a core data center. The training of a big model that you want to do at a core data center, massive data, big compute. But it's the use cases where you deliver the value. And so already today, we are running that on our compute platform and to do all sorts of things with inferencing. So content -- the content you get when you go to some sites is based on an inference engine based on there is a model that's trained but then what you're -- how you're interacting with it, ad targeting. With our Bot Manager solution, there, we want to know are you a bot or a human. With that Account Protector, are you the right human? AI, very important for that. And the implementation of those things runs on our edge platform today. And we do it by -- CPU-based, much more cost effective. GPU is very important for the core, for training and the heavy use there. We do support GPUs, are buying more -- most of that actually today is for graphics for the gaming customers. But for us, the focus is other people's AI engines running on our compute platform.

Got it. Got it. I want to talk a little bit about competition across all kind of 3 buckets. Maybe we can start on the delivery side of the equation. You mentioned earlier some of the competitors dropping out of the marketplace and perhaps, another one kind of rationalizing in the near term. The question that I've always gotten from investors is why the hyperscalers aren't going to be kind of natural competitors to you guys on the delivery side of the equation.

Well, they are.

And they are. So how has that competitive dynamic kind of evolved? They're also the customers, right?

Yes. Yes. So 2 of them are huge customers, use us for delivery and security. One of them even uses us for compute. And that's -- we differentiate ourselves with better performance at a better price. But they are big competitors. And in particular, if you give them your compute business, often they will give you delivery at a much lower price or free in some cases. And that's because the compute usually 10x the delivery value at a higher margin. And we are in discussions now with some of the big media players and doing the same thing. We haven't closed any deals that way. So our compute revenue today doesn't have anything to do with delivery. But that might happen this year. So a big media renewal, they want to save money because they all got to save money. We say, well, we got a great way to save a lot of money. Give us a bunch of your compute, yes, and we'll give you discounts and delivery then because it's much better for us. We grow revenue. Better for them. They save money and better overall margin for us.

Got it. Got it. And then on the compute side of the equation, what I've heard you talk about what you guys are trying to do with Gecko and connected computing, it's not necessarily going straight after the hyperscalers, right? But it's more so there's a lot of use cases for which the distributed network or a smaller vendor or something more cost effective could be much better. So like, one, is that a correct characterization of how you think about competitive dynamics versus the hyperscalers? And two, who do you compete with for more of those like niche use cases if you will?

Yes. It's a pretty big niche that we're going after. It helps to care about performance, and it helps to care about price. And we can give the biggest price reductions for apps that are data intensive, so data moving around or a lot of hits because that's where you get huge bills from the hyperscalers, and we have much bigger price reductions in those models. So our focus has been our big media customers, big gaming, big commerce, and those guys are spending a fortune on the hyperscalers. It's easier to migrate to our platform, especially big media because we already have a good media ecosystem on our connected cloud so that you may think you're locked into that workflow on a hyperscaler, but we have replacements for it. So it's not flip the switch and you've migrated or lifted and shifted. It takes effort. But it's pretty well set up there. So we would aspire for big media commerce, gaming to go grab a lot of that cloud spend, which is a significant amount of revenue today. Big media, the hundreds of millions of dollars, in many cases, $1 billion a year with their archrival.

Got it. So prior to the acquisition, you looked at Linode as a potential competitor for like a digital ocean, but it seems like you're gearing towards a different kind of competitive space.

Totally different. So we like the Linode business. It's growing modestly. We most care about it over the long term because we want developers to know it and like it and feel it's easy to use, but it's not because of the revenue, which today, we don't want $150 million to go away. We'd like to see it grow. But all the investment is about big enterprise mission-critical apps, which is not what Linode could do before. But now with what we've added to it, can and that's our focus.

Got it. And then that becomes a more interesting market opportunity. And then finally, in terms of the competitive environment in security, who do you see most often in sort of that security competitive environment?

Yes. We talked about sort of 3 pillars and there are different competitors in the pillars. You've got the sort of the oldest and most mature is the scrubbing. That's the Prolexic service and DNS. There we'd see carriers that bought Arbor or Radware. Occasionally, we'll see Cloudflare there. The big pillar for us is WAF and all the things on top. There, we're the market leader by a wide margin. Imperva, probably closest #2, but they don't have a lot of the capabilities built on top. F5, a little bit with a legacy approach. And at one point, they bought Shape. We see them less now in the marketplace. And then on the -- inside the firewall with segmentation, Alumio would be an example, probably the #2. Enterprise Application Access, that product probably is ZPA or Zscaler is who we'd see there.

Got it. I want to take a moment to see if there's any questions from the audience before I keep going. All right. I want to turn to capital allocation strategy on a go-forward basis. You haven't been shy about doing M&A in the past. You've built out the security portfolio that way. Linode came through M&A. How should we think about the M&A appetite for Akamai on a go-forward basis? And maybe more broadly, how do you think about that buy versus build decision?

Yes. I think the strategy has been pretty consistent and will remain as it's been. We don't do many large acquisitions. We're really careful about it. We did do 2 in pretty rapid succession with Guardicore and Linode. And both are very strategic reasons. It's not impossible you'd see another security acquisition sort of like that, but we're very careful with what we spend. So we got to be really convinced that it's going to be -- get the return on the investment. And today, still, I think a lot of those companies are still in a big bubble. They've come down in valuation, but they got a long way to go before it would make sense for us to do an acquisition at a larger scale. Compute, we have the big pieces we need by and large, and tech tuck-ins are possible, technology hires, that kind of thing. Security might see other adjacencies, but we're careful about the larger spends there.

Got it. So it sounds like we'd be more likely to see tech tuck-ins to like put in functionality...

Tech tuck-ins, product adjacencies that make a lot of sense, synergistic with the platform that we have.

Got it. Got it. So you guys repurchased $650 million in stock in FY '23. You're going to be generating more cash flow in FY '24 versus FY '23. How should we think about the allocation between increasing shareholder return, either vis-a-vis shareholder stock repurchases? You saw Salesforce initiated a dividend, so like who knows? Or just increasing sort of investment in the business?

Yes, no change in strategy with capital allocation, which is we do some acquisitions. We do tend to repurchase the equity we've given in employee equity programs. And then there's occasional opportunistic additional buyback. Last year, there were some. The stock got down to what I would consider a very naturally low place, and so we bought back more. The algorithm we have to buy back every day in the market, embedded in it is that it buys back more in a nonlinear way when the stock price is low. And so on balance, on average, over the last 10 years, we've bought back an extra 1% per year. So you think of the total number of shares outstanding today, ballpark, 10% less than the outstanding shares 10 years ago.

Got it. Got it. I want to shift gears a little bit to margins. Gross margins came in a little bit in FY '23, the Linode acquisition being a part of that. Can you help us understand like the puts and takes on that gross margin line on a go-forward basis in terms of -- like if we think about the cloud hosting costs and colocation costs and versus maybe a higher mix of like high-value solutions like security, like where does it net out for us on a go-forward basis?

Yes. The big hit last year was in the accounting rules for colo. So as we built out the big core Linode regions or the new Akamai Connected Cloud regions, we secure colo contracts over an extended period, many years into the future, to allow for growth in the same facility because that's more efficient. And we don't pay for the higher levels of colo until we actually use it, say, 4 years from now, but the accounting rules linearize it. So that hurt our gross margins by 1 point or 2 for expenses that we're not paying any cash for. So a little bit unnatural. We also have increased investment, which would show up on the OpEx line, which comes into operating margins. Favorable to us is we're reducing our cloud spend because we're using our Connected Cloud. And we had a lot of cloud spend that was growing rapidly, and now a lot of that has moved on to Akamai Connected Cloud. So last year, our third-party cloud spend went down instead of up, 50%. And this year, it's going to go down a lot more, so a pretty material difference. And so last year, you see our op margins actually improved 1 point, got back to 30, which is where we like to be but even though we're doing a lot of investment in new security capabilities and a lot of investment in compute. So we're balancing those things out.

Got it. So some of the cost savings initiatives are particularly around cloud that were put in place this year. Yield this year but you're taking that excess, sort of the efficiencies, looking to keep operating margins flat and reinvest that into the business in terms of -- and can you talk to us a little bit more like the nature of investments in security? The cloud side of the equation is more obvious, is expanded capacity is going into those 100 cities.

Yes. For security, the investments are largely people. There's developers, both through the acquisition and then we've added a lot more to develop the new capabilities and the specialized go-to-market teams. So our model is the rep owns the entire account for all the products, but we do have specialist teams for some of the security products and now for compute. On the compute side, we do have a lot of development that's new and some operational folks. And also, we have a new compute specialist team, by and large, new, so there's increased OpEx there. And then as we talked about as we're doing this build out, you got a lot of direct costs for colo or allocated costs and of course, with a big CapEx build, the depreciation now comes in the [ front ] on that.

Got it. And the last question I was going to ask was around sort of CapEx and free cash flows. With the increased investment in cloud, you saw CapEx intensity go up about 7 points and in FY '23, it was about 19% of revenues. The initial guidance seems to imply some stabilization in terms of the mid-teens. Can you help us like triangulate like where are we going in terms of the steady state? Like is Gecko in the Connected Cloud going to make this an inherently more CapEx-intensive business over the next couple of years as you build that out? Or is the majority of that, the investment going to get behind us sooner?

Yes, great question, and it depends on revenue growth for our enterprise compute business. So a big tranche last year to get the core Linode capacity in place. This year, smaller but still meaningful investment. We're rebuilding and upgrading some of the original Linode data centers, and in some cases, that means a new data center in the same city. And we're doing Gecko, which much less expensive. We are already in those locations. So the CapEx investment this year is less than a big number last year. Now we're using some of that capacity for our own compute usage, and the rest is now available for customers to use. So we're -- hope to fill that up with revenue at -- to fill that up as no extra CapEx cost. Once that happens, then there would be more CapEx for additional revenue. And the faster we can take on more revenue, the faster we would have more CapEx spend. On balance, it's a very profitable business for us at anything like the current pricing environment. So we want that to happen. Ultimately, when it's at scale, the compute business is CapEx intensive. That's what it is, but the margins are accretive for Akamai as a whole. And now you can -- how you look at the accounting depends -- how much you're spending depends on how fast the revenue is growing. On balance, you want it to grow fast.

So there's been some catch-up investment in the past year, maybe a little bit this year. But on a go-forward basis, CapEx should be a good leading indicator of what you're seeing on the cloud demand side of the equation.

Yes. And you'll see our revenue reported for cloud computing at the same time.

Excellent. Outstanding. Tom, thank you so much for joining us. It's a great conversation.

Okay. Thank you.