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

Great. Good morning, everybody. Tim Horan, I am the Digital Infrastructure Communications and now Satellite analyst at Oppenheimer. My pleasure to be hosting Ed McGowan the CFO -- long-time CFO of Akamai and long-time Akamai employee. Thank you so much for joining us, Ed.

And I guess the company has evolved dramatically over the last -- your career there, last 5 years, 10 years, 15 years, you have a hell of a lot of new products. Cloud is a major business for you guys now and growing rapidly. Security is a huge, major business kind of growing rapidly. How do you explain what the company is now or define what it is? And kind of what is the strategy tying all this together here longer term?

Yes. Well, Tim, thanks for having me today. Good to see you again. Yes. So I've been here 25 years, and we've changed quite a bit. What's interesting, though, I think if you go back to even our IPO slides, there was a view of how the company was going to evolve, starting with CDN going to security, compute was on the horizon as well. So it's kind of played out how we had thought it would over time. Obviously, it's -- each business is in a very different state at the moment. But all of the different services do leverage a common platform. And even our compute data centers, even though they're separate from the CDN, we have connected our backbone to the location. So that serves as a competitive advantage, right? You've got better performance, but also basically 0 cost because of how much traffic we serve. And if you look at every hyperscaler, they've got some form of a delivery platform. It's required if your users are displaced and your applications are further away, you need some type of a platform. We did a great job of moving from a CDN to a security company over, say, call it, from 2011 through continuing on today, it's over half our revenue. And there, we leverage the platform, the data that we got from the platform. We saw every Internet user multiple times a day, and we built a very big web security business. And now we're moving into the enterprise security space with our Guardicore and API security, which does a little bit of both web and enterprise. So it all does fit together. Our strategy in terms of where the business is going. I think CDN is kind of a flattish to down slightly type business. Obviously, it's gone through a couple of years of some challenges in terms of traffic growth rates and the pricing dynamics, but things are starting to get a little bit better there. We don't look at that as a significant growth driver, but it is a very strategic asset for us required in both security and in compute. And with the enterprise compute business, I think that's a significant area of growth, a long way to go there with innovation and M&A. So we think that can continue to be a good source of growth for us. And I think the real big opportunity and where -- if we execute well, the company will find its largest source of revenue over time coming from compute.

And how do you define your platform? I know you said it's a common platform, but what are they all leveraging together? And how has that platform kind of evolved here?

Yes. So I mean, at a basic level, when we designed the platform, it's now 400,000 machines and over 4,000 locations. The concept originally was the Internet wasn't designed for performance. So we built an overlay network effectively that is performance-based, right? Now when I say we leverage the platform, if I think about the security products today, let's say, web application firewall, that same server that's delivering, say, a video experience to your house might be blocking an attack coming from your neighbor's house, let's say they have an infected device and they're part of a bot army or something. So we are able to get a lot of leverage off of that platform for the security and delivery business, both with the physical infrastructure as well as the data. And also, if you think about being able to, say, for denial service attacks, block attacks where they emanate instead of bringing them into a central location, you get significantly more capacity out at the edge to be able to block those attacks, but also it's very cost effective for us, too, because we're using those machines for multiple purposes. So that's the platform in terms of security and compute from the physical side. Then with -- like as I talked about -- sorry, delivering security, with compute, it's tying the CDN network into the core data centers. And also, we just launched our container service, which effectively is being able to run a container in any one of those 4,000 locations to the extent that it makes sense, if there's enough capacity there and the economics make sense for us and there's customer demand. So there's that physical side in terms of the hardware and then also some of the core technology, the routing technology and all that kind of stuff. But then there's also from an engineering perspective and operational perspective, the same people that build out the delivery platform or building out the cloud infrastructure platform. We get a lot of similar engineering efforts across the different product lines. And then from a go-to-market perspective, there's scale where we're selling to the same customers, multiple products, you're getting pretty good scale in the go-to-market. There are some overlay functions that are specialized in selling some of these advanced capabilities. But in general, the whole field force can sell pretty much every product. So you get some scale there as well. So that's what I mean when I talk about sort of the overall platform from a like the technical platform, but also just the operational platform that we have as a company and the ability to have people multi...

So you kind of -- you've integrated Linode and a bunch of your cloud computing capabilities into the same platform at this point?

Yes.

Got it. So just switching gears a little bit. One of the themes I'm getting from the conference is that AI is actually starting to happen. I know it's been happening for a while, but people are -- enterprises, I should say, are deploying it, and they're seeing productivity improvements, improvements in products on and on. And apparently, this is also creating new demands for different -- well, lower latency networking, different forms of networking, maybe forms of edge compute. So kind of related to all this, what does AI mean for your business from a revenue and a technology perspective? And then I'd love to understand after that, what does it mean for your own operations and how you're...

Yes. So if I think about the security business, we just launched a product, an AI firewall product, which we weren't even thinking about several years ago. So a new category, which effectively helps, say, a commerce company or a travel site has a chatbot of some sort and there's request going in and outbound, being able to put a firewall in place to ensure what's going in is what should be going in and you're not getting hit with denial of service attacks or bots, you name it. And then also what's coming out, the responses are not hallucinations or something that might be offensive or personal identifiable information or trade secrets and that sort of thing. So that's a new area. So there's a new revenue source there. We just launched that. A lot of demand so far from customers, a lot of proof of concepts, a couple of paying customers. So it's still early days there. I think there'll be more security-related products to come there. In terms of security demand, I think it's a massive driver of security demand. Think about how sophisticated the attacks are getting even -- I was talking with my HR team, we just went through a training on how to identify that you're actually talking to a real person when you're interviewing them remotely, that these AI technology got so advanced that you can actually fake someone up when you think you're actually talking to a human, it's actually not a human. So there's -- just the sophistication of these attacks are going to get even more and more sophisticated. And a product like Guardicore becomes even more crucial because all your external defenses, your endpoint detection, your secure web gateways, all that stuff, something is going to get in because the attackers are getting so much better. So that helps the demand environment. In terms of compute, we are seeing some early days with GPU as a service. We do offer that in the platform. We do have some folks running inference engines, kind of very early days on that. So there's not a ton of revenue yet, but I do think that's a good opportunity for us.

As an aside, how do you tell you're interviewing AI instead of a human? What is the trick?

I didn't really pay attention because I don't really interview anybody, but there was a whole training that we just launched for it. But yes, there's like tricks you can ask them certain questions and look for eye contact and all these different things, but it is becoming a challenge. It's something I would not have thought. I mean, certainly doing deep fakes, like you could find your voice or my voice with all the earnings calls that we've done and make phone calls and tell my employees to do something using my voice, right? Now how do my employees going to know that? So you have to make sure you putting controls in place to be able to ensure that my treasurer doesn't go and launch a wire out to somebody, and we just make sure we have controls around that. But security environment is going to be the same thing, right, where things will get much better at faking people out.

Yes, that's incredible. So is AI -- it sounds like -- is it driving more security attacks and quite different security attacks?

Yes, definitely, definitely. Yes. So we're seeing -- just even if you think about your own world of the text that you get or voice mails that are fakes and the e-mails, go back 10, 12 years ago, it used to be that I have an uncle over in somewhere in Africa or whatever and some misspellings and poor language and you knew like, okay, this obviously isn't real. But some people would fall for it and they would give their bank account information and get their money stolen. Now it's a lot more sophisticated and it does look like it's maybe something coming from your boss or my boss and it's something we should act upon or a link that we should click on. Like I've seen some fakes -- one of the bigger targets in my environment and some of the CFO, the finance departments usually get hit with a ton of e-mail fakes. And they're starting to get really good with the workflow. So it look like, let's say, coming from an Oracle application or a Microsoft application or something like that where it does look like it's legitimate workflow that you would click on to go perform an action. And as soon as you do that, the bad guy in, your machine is infected and now you've got ransomware running around in your environment.

Yes, incredible. And also, a lot of companies -- a lot of your customers, in particular, have content that, that content unknown to them has been used to train AI models, and they would rather get paid for it.

No, it's a great point. That's something that the publishers in particular, are trying to figure out. And obviously, search is going to be potentially turned upside down in terms of the model. I think we're in a pretty good position to participate in that. Obviously, certainly, if there's protections that customers want in terms of not allowing AI bots to get in and get their information, we can identify those and block those. That's just one example. But as the commerce model changes, there's a potential that we could be involved somehow. I don't have an answer to that yet today, but it does create an opportunity. I know some people are worried that does that mean there's less Internet traffic. Maybe on the margin, you might see a little less traffic to some publishers, but I think the streaming, that's not going to be replaced by AI. That's where most of the traffic is, the software downloads and that sort of stuff. So I don't think it will have a material impact on traffic on the Internet. It certainly will have some, but I do think it opens up new business models for you.

Well, I definitely wanted to talk about the traffic and a little bit because I've heard different theories about it. But just getting back to your ability to kind of host models from an inferencing perspective, it would seem like your infrastructure is really, really well positioned to do that in some form of edge compute with lower latency. And I know you're starting to see that a little bit, but are you starting to create products or infrastructure that can really support that?

Yes. So I mean, we are -- we -- the managed container service is probably the best example of something we launched recently, where if you wanted to run in, say, hundreds of locations, we certainly have that capability now where you can -- whether it's running GPUs or CPU, running your model in multiple locations, very low cost. It's the same locations where we would have some CDN infrastructure. Obviously, you're not talking about a training model out there. It's a very lightweight application where latency and cost is a big issue. So that's a good example of something that we recently launched. But we're working on other things, too. Like I said, there's some demand for GPUs today. We're being pretty smart about how we're thinking about rolling that out. But that's another area potentially of some interesting revenue for us. I don't think you'll see us get into the training model anytime soon, if at all, because it's -- the hyperscalers pretty much are going to lock that market up.

So then on the traffic side, there's a lot -- it does feel like Internet traffic has been slowing here a bit the last couple of years. Well, do you think that's correct? And do you have a sense of what's driving that and maybe what can reaccelerate? I guess there's some thinking that AI could maybe reaccelerate the traffic growth again, but...

Yes. I've heard both sides of the argument, right? Some say it will take traffic away, some say it might accelerate. It depends on the application, right? If it's something that's going to have consumers spending more time online, consuming high-bandwidth applications, certainly, that's going to drive a lot of traffic. But in terms of demand, I think we saw over the last, say, 18 months, 2 years, obviously, out of the pandemic, extremely strong volume even the year after the first year of the pandemic, but then it started to taper off. One thing that we noted from our customers, especially the streaming customers, one, there was the writer strike. So content -- new content was not as readily available. And there's a lag between when you produce a movie and when it actually gets released or a show, there's usually a year lag or sometimes 2 years for a movie. So there was less new exciting content. The streaming folks like the Disney's of the world and NBC's, they were focusing on cost savings. They had kind of rolled out aggressively, got a lot of consumers and then they started cracking down on things like password sharing. There was some experimental stuff going on with the new codec so that you would play around with the bit rates to try to get good quality with fewer bits, Fewer bits means less traffic, obviously. I think that's kind of largely played itself out. There'll always be new advances in technology, but we are seeing much better streaming traffic, especially over the last two quarters and continuing here in the month of July and early August. So we're seeing better traffic. It's not the typical 30% plus traffic growth we're seeing on the Internet as a whole, but it's a much better environment. Some of that could also be there's less competition, too. You had four sort of scale players on the large volume side exit the market. We were the benefactor of a lot of that. So the demand environment is better. There's still do-it-yourself with the really large players, and that can throw things around a bit. But I think in general, the trends are looking better.

That's interesting. So you think it was a bit of a onetime hiccup in traffic growth. We didn't have a lot of huge new content and some of the password sharing cracked down was a bit of a onetime issue. Yes, interesting. And so is traffic for streaming -- has it improved? Like is it like 50% better than it was, 25% better? Just...

Yes, that's a good question. I would say it's not 50% better, but it's noticeably better. We're seeing much -- like looking at my model coming into the year, we're doing a lot better from an overall aggregate growth standpoint, maybe 1/3 better or something like that. That's...

And I guess while we're on it, your overall revenues on CDN have improved a bit. I know the Edgio acquisition helped there. But are you seeing -- what's going on with pricing trends? I mean they were brutal there for about a decade on and off. Have they improved from the trends?

Yes. So there's -- I always sort of break the business into two components. There's a super high-volume big media streaming software downloads, gaming customers where they put the most downward pressure on pricing for the first 15, 20 years. And that was just volumes were significantly growing and just you do a deal for a year and then you build out a table and the volume a year later is much higher than you ever contemplated. So that put a lot of downward pressure on pricing. That's starting to moderate. And you can see that because it's tens of customers that are in that category. That's been improving. You're not seeing the type of annual discounts like you saw in the past. So that's good. There's also less people chasing it, too. So that helps a bit. And part of that has to do with volumes are not growing nearly as fast. We literally, for the first 10 or 15 years, our traffic doubled every single year. And we're not seeing that now. We're talking about hundreds of terabits of traffic on the platform. But then there's the rest of the base. That's your commerce, your travel, your financial services, manufacturing, just kind of auto, all that rest of the traffic, small percentage in the grand scheme of things. But there, it's probably a little bit more price sensitive than the upper end. On aggregate, if I look at my pricing trends, they are moderating. So look, if I look at my year-over-year price declines, they are heading in the right direction. They're starting to moderate, and that's been helping with sort of flattening out the business. If you look at our revenue in the last 3 quarters, it's been between $318 million and $320 million. So we're kind of flattening out a bit, which is good. You'll always have renewals. There'll always be a big customer or two that's going to bounce you around, maybe decline a couple of million over quarter-over-quarter, but definitely much healthier than both pricing and volume.

And I guess in that regard, in that particular segment, you had a new competitor in -- well, not that new anymore, Fastly, seemed for an initial period have some features and functionality that you guys didn't have. Do you think you have kind of similar latency to them now or maybe even better and a better overall product at this point?

Yes. It's interesting. The -- where they started off was going after the -- that second group of customers I was talking about, they come after the commerce and the less traffic type customer that higher price point, lower traffic, focused on acceleration, not just delivering and cashing. And they find it's tough to grow. We don't lose a lot of customers. So it was sort of a futile endeavor. Where they started to focus after that was on the big delivery. So you see them in at the Disney's, TikTok, Sony, Microsoft's of the world, and they pick off a percentage of the share. A lot of those folks will load balance between sometimes their own infrastructure and 3 or 4 CDNs. So they're picking up growth there. And that's -- you can grow pretty quickly for a year doing that. If you become a new entrant at Microsoft or Apple or whoever, and you pick up 10% or 15%, you're going to show growth for a year, and that can be meaningful. But that's a tough business, right? It's heavy CapEx, it's -- the margins are much worse than ours. I don't think I would say they had significant technology advantages. They were, in some cases, they had some tooling that was a little bit easier for customers to onboard themselves. We had more sophisticated products to solve more sophisticated and difficult challenges with enterprises. And we had a much bigger services organization to very sticky with the customer because a lot of times, they would outsource a lot of what running the sites and running the security like web app firewall, we have a pretty big services business that runs the -- runs whether it's bot management or web app firewall for our customers. We handle all the rules, rule changes and things like that. But they had little things like Fast Purge was a thing that they had at one point. They had an origin offload product that we created something very, very similar. So I wouldn't say it was anything significant. In terms of performance, they perform decent in some of the bigger Internet locations. But when you get a heavy traffic day, that can cause some serious concerns. And we tend to see more share shift our way in that particular time because being on the other side of the choke point or the congestion point when you're deployed in many locations in an ISP, you're going to get better performance. And a lot of these sophisticated load balancers will balance off of performance.

Good. And are you seeing much impact from Cloudflare? Do you compete against them much?

Yes. So Cloudflare doesn't compete on the big media streaming and gaming yet. I think at some point, they might. I think it will be probably a bad decision because it's got a very different growth profile, and it's got a very different set of economics, and it's very -- much more capital intensive to get in that business. We do see them compete for web app firewall to some extent, probably lose a handful of customers a year. We take about a handful of customers. Our churn in terms of lost customer annualized revenue for customers that churn off the platform is less than 0.5%, and it's been like that for years. So they don't take a lot of customers and more of a nuisance where they'll go to a procurement and say, "Hey, we can do it for 1/3 of the price" and you're just in there dealing with price declines on renewals as a result. But don't lose a ton of business there. We do -- on CDN, there's -- they focused on the down -- like the lower part of the market. So like they have 1 million free customers or something like that and 100,000 or so paying customers. We focus more on the top of the pyramid. They've been trying to come up market into the enterprise space. So we don't see them a ton. There's also some traffic we don't touch, but I know that they do deliver some of the adult content and some of the other stuff that -- dark web stuff that we don't touch. But that -- I have no idea what the economics look like there. I assume they're pretty good. But they also have a pretty good secure web gateway business from what I understand. We don't see them that often. They compete quite a bit with Zscaler. Our Swig is not a huge revenue generator for us. They don't have an API security solution, and they don't have micro segmentation. And in terms of compute, maybe a bit on the Functions as a Service business, but in terms of the big compute opportunities, we don't see them there. They're more of a kind of legacy edge compute. They do have a storage platform, but yes, so they're not a huge source of competition for us.

So when I talk to investors about you guys, I guess the key question people are asking me to ask is, can you maintain double-digit revenue growth in security? It seems to be a pretty big focus. I know you're entering the enterprise security market, which is substantially larger. But I think you've been entering that for quite a few years. I think you have some very differentiated products now. But are those products enough in enterprise security to maintain double-digit revenue growth?

Yes. So we broke out the growth rates for you guys this last quarter. We bundled API security with Guardicore platforms, which includes Secure Web Gateway and our enterprise access solution. And there, that business was growing 32% year-over-year, $67 million in revenue. That's so approaching $300 million, growing at a very nice clip. Obviously, that's not enough to offset the rest of the product slowdown. That rest of the products, WAF, DDoS, Prolexic, that's growing 7% roughly as a total category. Some products growing a little bit faster, some growing a little bit slower, but that's a fairly decent number to think about. We've always said that 10% would include acquisitions. So in order to maintain the 10% over a longer period of time, there'll be acquisitions as we've always done. The enterprise category is probably the area that has the most opportunity for growth for us. We've had some fits and starts. And I think now we're getting our sea legs much better there. And starting to get some real traction and we think that there's more to do from a platform perspective. So there'll be some homegrown products. We just launched, like I said, the AI firewall earlier in our conversation, and we'll be launching new features and functionality. I think the security landscape is going to change much faster and very quickly over time as a result of AI. So I think there's certainly lots of opportunity for both continued growth with the stuff we have now, but also new innovation and acquisition.

Yes, that's really helpful. And I would think it's a different sales motion, enterprise security than web security. And that obviously has taken some time to develop. Where are you with that sales motion and go-to-market?

Yes. So we're continuing to invest in our -- what we call our overlay sales team. When we do an acquisition, we tend to keep, especially in the enterprise space, where even with Noname, we kept their sales force and with Guardicore, we kept their sales force in their channel and kept the channel for API as well. But we have a gentleman who is Senior Vice President who reports directly to P.J., who runs sales, runs an overlay team that has license to hunt in all of our existing accounts and has a big hunting effort outside of our existing accounts because obviously, enterprise security is a much bigger opportunity than web security because there's certain verticals where websites really don't matter, and they have lots of employees and spend a lot on IT security. So we do a fair bit of hunting there. There's more sophisticated channel operation there. We do, I think, almost all of our business with the exception of some stuff we sell to our internal -- our existing customers goes through the channel with Guardicore. So I'd say we're probably in the later innings there in terms of -- we'll be adding more capabilities to that, looking to add a few more channel partners. And to the extent that we do acquisitions, if it makes sense, we'll keep the sales specialists that come along with an acquisition and just roll them under that team. But it's a smaller team than our regular field force, obviously. And they work in conjunction with our -- the owners of our major accounts, and we have compensation to make sure that both sides get paid for a deal if it goes, we have multiple folks working on it.

And I know you used the term platform. Is your web security and enterprise security, are they kind of converging onto the similar platform? And I guess, what do you mean by kind of platform? And how do you...

Yes. So with -- it depends. In some cases, yes. So for example, with API security and our web app firewall, we've built what we call a connector. So if you're using Noname security and you're a web application firewall customer, you can use our web app firewall to put in rule sets to block malicious things and basically protect your API traffic, if you will. With the Guardicore, there's really not a connection to the web products per se. There is a platform where you've got enterprise access and secure web gateway along with micro segmentation. So that's all one pane of glass. You can set one set of rules and that sort of thing. That's an area that we've consolidated over the last, I think it was 18 months ago now under one leader out of Israel, who came from Guardicore that's making that more of a whole platform. So we'll be adding capabilities to that over time.

So just switching gears, and thank you to cloud infrastructure. Cloud, broadly speaking, had an acceleration of growth, it seemed like in the industry. And you guys had very, very strong growth. Is your growth sustainable, you think? And how much investment will that require? And I guess even before that, do you agree that the overall cloud sector saw some accelerating growth? And what do you think kind of drove that?

Yes. So there's sort of two pieces to our cloud business. There's the legacy, what we call other cloud applications, which if you think about the origin of that, that was really stuff that we would build or offer as a way of getting more delivery business. So for example, we had an origin -- sorry, an object storage offering for some of our big media companies that wanted to get better offload. They might say, go to the storage platform before you come back to our origin if you have a cache miss, and so that was kind of a highly redundant object storage platform, about a $50 million business. It's not growing anymore. We've -- we're end of lifeing that. We have a storage offering in our compute business that's both -- we have a block storage offering as well as an object storage offering. So we'll be selling that now. Hopefully, most of those customers will migrate to that platform. I assume some will just end their contract with us. But -- so we've signaled that, that business will decline over time. We had some cloudlets that we developed, say, waiting room applications, image management, video management, where we're looking for partners to offer that as a cloud service to drive cloud business. And in some cases, we are migrating away from our own solution to a new -- to a partner's products in exchange for them moving some of their cloud spend off the hyperscaler to us, and then we jointly go to market, sell that product. We don't have to develop it anymore and that sort of thing. So we do expect that business to decline over the next 18 to 24 months. It grew about 3% this year. So that's not a growth engine. The cloud infrastructure services business that grew 30%, that is a significant growth engine, and that's where we're really investing significant dollars. And there, we talked about 30% growth this quarter with expected accelerating growth going into the back half of this year and into next year. We do have a lot of big customer contracts that we've signed that have not started to generate revenue yet. So we have very good line of sight. We did tell investors that we may have a bit of a timing issue this year where we may not be able to get enough months of revenue to hit the 15%. We might miss it by a point or two potentially, but that will just make the growth next year better. So it's really just a question of how quickly can you migrate traffic over from an existing application to ours. In some cases, we had to do some development. In some cases, we had to do some build-out. So there was a bit of a lag that we knew coming into the year, and now we're really at the whim of the customer in terms of how quickly they'll migrate.

So I have two questions from the audience. I definitely want to hit on. There's a question on the fourth quarter, Ed. Will the Noname acquisition -- when does that acquisition lap and that -- just...

So June -- yes, we had a couple of weeks of June, revenue in June last year. I called it out for you. It was $8 million of inorganic contribution. So if you look at that category of API and Guardicore, 48% growth, back that out, 32% growth if you take the inorganic contribution, which was a quarter less a couple of weeks of revenue that we had last year.

Okay. Got it. And then are you still expecting some revenue losses from the U.S. Fed in the fourth quarter?

Yes. That's a good question. We've seen some. It hasn't been significant so far. We are making a big investment in becoming FedRAMP high. So we do expect the federal government business to grow over time, certainly. We're already FedRAMP medium, I think, is what they call it, but we're investing to become FedRAMP high. We have good line of sight to some pretty good opportunities there. But there could be some additional -- like say, for example, Department of Education was a customer. So that's going away. So obviously, that revenue will go away. Nothing material, but it's a headwind.

And then lastly, a lot of different businesses, a lot of moving parts. Can you give us a sense of what you think this business can grow at longer term of revenue? Is it a mid-single-digit revenue grower or high single? Can you get to double-digit revenue growth overall at some point?

Yes. I mean I think if you look at the ingredients to get to double-digit growth, say, 10% or better, one of the big impediments to that was the delivery business was shrinking 8% to 10%. So now I think that's hopefully in the low single digit, kind of 0% to 4% as we had negative 4% as we had sort of set as a long-term target. I think that's probably right. There's some risk to that, obviously, DIY can sometimes maybe knock you off course for a year. But let's say you get that to that sort of level of 0 to minus 4%. If security kicks in at around 10%, that certainly is going to help. But I think compute, especially as compute infrastructure services becomes a much larger portion of the business, certainly a much bigger market, you have the opportunity to be a high single-digit, low double-digit growing business sustainably. I think we're not quite there yet. I think we have the ingredients to get there. There's a lot of execution that's required, but that's certainly our goal. That's what we're investing for. That's what we're trying to get to. And we think we can ultimately have the end markets to get to hopefully a sustainable double-digit growing business. That's the goal.

And can you do that on a relatively stable EBITDA and free cash flow margins? Or will that come down over time?

Yes, good question. So I think what we're learning with the gross margin in the business for compute with sort of the partner dynamics and how we're going to be going to market there with partners versus us building our own sort of applications, think of like media workflow. We're not going to build our own. We'll partner with somebody. So if you resell it, you get a lighter margin. And then also some of the dynamics around colo and the lease accounting, we're probably not going to see expansion in gross margin. I see maybe a point, but I'm not anticipating that even though the mix will shift. Really, you'll see it on the operating side in terms of getting operating leverage. And you actually saw that this quarter, right? We beat revenue and there's a significant flow-through to the bottom line. We delivered 30% op margin, EPS by $0.15. So we are definitely set up for really good operating leverage. And I think that margins can expand beyond 30% over time. There's more investment to go, both with capital as well as some engineering efforts. And obviously, if you do an acquisition, you're dilutive for a year. But yes, I think there's opportunity there. And then from a free cash flow perspective, I think the growth rate in compute will ultimately dictate where free cash flow lands. I think over the next few years, as we grow from $300 million run rate roughly in CIS to a couple of billion, you probably can do it at the same type of dynamic in terms of low 70s gross margin, CapEx somewhere in that 18%, 19%, 20% range and EBITDA margins kind of where they're in the low 40s and then maybe the operating margin is either side of 30%, maybe it expands a bit. And then as you get beyond that, the -- you may be making much bigger bets on infrastructure. And at that point, it's a very different company. And there, you're talking about a significantly larger growth in terms of dollars. So that -- you just kind of watch the progression of the hyperscalers in the very early days. They sort of ran kind of the economics we're talking about now in terms of like the cash flow -- the investment in CapEx to revenue. And obviously, they got much bigger and started building their own power plants and stuff like that. We're many, many, many years away from that if we ever get there. So I think what you see now for margins and cash flow will probably be roughly the same and then hopefully expand a bit. And then if there's big opportunities, we'll certainly tell you if there's a big investment we're making that would lead to faster growth and more revenue, certainly let you know about that.

We're out of time, Ed. Really appreciate it. And thank you, Mark, for lending us Ed for 45 minutes here. It was great.

Thanks, Tim. Really appreciate it.

Absolutely appreciate it.