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

All right. Great. All right. Thanks, everybody, for being here. My name is Frank Louthan, and I'm the senior wireline analyst here at Raymond James, covering Akamai. Very pleased to have Ed McGowan, Chief Financial Officer from Akamai, here with us today. We'll jump through a few questions and take some -- see if folks have some questions from the audience.

So Ed, let's talk about Akamai. So you guys had a great year of execution. Kind of walk through what's changed here this year that's kind of put you in this enviable position. What do you -- how would you characterize it?

Well, first of all, thanks for having me, Frank. Great to see everybody. Yes, we've had great execution here. I think it's a good way to describe it. We're pretty much across the board on everything from -- probably the biggest change is our investments we made in our compute business. We've rolled out 12 to 13 new major data centers. That's gone very, very well. We started to move some of our applications off of the public clouds onto that infrastructure. It was a pretty ambitious goal. We've done very well there. We talked about last quarter, our cloud expense is down 23% year-over-year, which is pretty impressive. Security is going great. We're exceeding our own expectations. We rolled out some new bundles that are selling very, very well. Guardicore is selling very well. We're starting to see the delivery business improve. So we're starting to see traffic come back. We've done a good job of being a bit more stingy with pricing declines, I would say, and being a little bit more selective with how much peak will go on our networks. So just great execution across the board. But the biggest change is the compute investments.

Great. So with that, let's talk about Linode being part of that success and kind of give us an overview of kind of where you fit in the compute in the cloud market. When we talk to a lot of investors, they balk at this. They think you're trying to compete head on with Amazon or something like that. Walk us through how your product is different and kind of where you fit into the compute and cloud landscape.

Yes. I mean, first of all, it is comparable to the hyperscalers. It's just at a smaller scale. We don't have all of the features and functionality. And our goal isn't to compete head-to-head, and it's not a binary win everything or lose everything. I think as people go into a multi-cloud environment, we can be one of those options. We're a lot cheaper. We have a more distributed architecture. We're going to continue to get a lot more distributed. So for those low-latency use cases, I think it's a perfect opportunity for customers that a few data centers isn't enough for them. They need to be closer to where their users are. We'd be a great option for that. But it's not a case that it's win all -- winner-take-all in this business. And we think there's plenty of room for us in this market, and we're proving it by moving our own spend. And we spend, as we've talked about publicly several times, well over $100 million. And we're moving that to our own platform. It works just fine. And obviously, we save a ton of money by doing that.

So kind of a follow-up on that. So what have you learned, right? So that's a big step to take, consolidate your cloud and eat your own cooking. And that's a really -- to jump in there, what have you learned from that? And how is that -- how are you able to then sell that product to your customers from -- with that example?

Yes. What's interesting, it's actually a really good exercise I think a lot of companies should go through, just even if they're not moving. Just figure out how they're using cloud. Is there a smarter way to do it? If you look at your -- say, if you're looking at data, you're sampling everything versus storing everything, how long do you keep data, that sort of stuff. So just that hygiene of going through and looking at the architecture and how we're using the cloud, bringing it on to our own platform, and we just are using it a lot smarter now and actually using less capacity than we thought we would need to initially. But also realize that it's not as impossible as everybody thinks. It was -- going in, you always take on a big journey like this. You always wonder how successful you'll be. And I'm pretty amazed at some of the applications that we were able to move and how quickly we're able to move them.

Yes. So talk to us a little bit about the enterprise readiness part of it. That was one of the things that took a little longer to get together, but that's starting to come together. Where are we on having all those piece parts to become fully ready for enterprise deployment? And when will that be -- that project be completely done?

Yes. I mean we'll always be adding functionality, adding new partners, building out capacity and that sort of thing. But Linode built a great technology. So that -- it was something that is super simple, easy to use. It was designed for somebody, a developer, to effectively fall into a virtual machine in a matter of minutes, right? Now we didn't want to lose any of that, but they didn't design it for the major enterprises. They didn't have -- they had some pretty good sized customers. But for even our needs, we needed to have a lot more redundancy, resiliency. We connected the -- all the data centers we acquired plus the new ones we built out to our backbone. So it's a lot more fault-tolerant, robust. We added more capacity. That's one of the things that we realized is that for some of these workloads that we're going after, the data centers need to be larger and have more capacity. So that was probably the biggest thing was the build-out of the capacity and making the service a lot more resilient and reliable. And then the other thing is around compliance. So adding SOC 2, PCI and other compliance, we'll continue to add that as we go. And those are really sort of the basic building blocks that we need to put in place just to be able to start to take some of the workloads. Obviously, there's some companies that have bought some of the lock-in features and some proprietary things. That took a lot longer to be able to compete there. And some things, we just won't compete for. But for someone who's using basic raw compute and storage, we certainly have the capability to take on pretty much anything.

And when did you have all that in place and what's sort of the sales cycle? So when can we expect that to kind of be fully going and hitting its stride?

Yes. So we're selling now. We've landed a bunch of impressive logos, and we've got some pretty good customers that are starting to ramp up pretty significantly. So we're out selling today. It's really just a question of bringing people on. The sales cycle can be a bit long in terms of people will pick a workload that they want to move. They'll start testing it at first. And then once they get comfortable, start to move it over. So it's not something that I can sell a month later. I've got the capacity all sold out. So it will take a while. But we're out selling now, and we're focused primarily in the media and entertainment vertical and we're focused in commerce. Those are the two areas that we're very strong in and have long tested relationships with and also believe that that's a good opportunity for us. There's certain unique characteristics in each one of those markets, whether you're a commerce customer that has data that's moving around a lot, and it's very expensive. If you have data that's being pulled out of the system, egress fees are something that a lot of our customers complain about. Given that we're connecting it to our backbone, that won't be a problem. If you use Akamai, it will be free for you. For the media customers, a lot of the major media customers haven't locked into some of the proprietary workflow capabilities. Workflow is something that tends to be latency sensitive. You need a lot of compute for that. That's something that I think will be -- probably some of our larger customers will come in those two verticals.

So walk us through kind of the egress fees and that kind of stuff that -- we hear about this a lot. And again, some investors don't seem to fully appreciate that. Walk us through what kind of struggle that puts a customer under and how you help them get out from under that for workloads that are -- that you can get outside there.

Yes. So let's take a simple use case. Let's say you're a media company that has a video library. You've got, say, 10,000 different titles. Obviously, there's going to be somewhat of an 80-20 rule, where there'll be 20% of that content is very popular. And as a CDN, most big media companies will use a CDN. With caching technology, you can cache that popular 20%. And we've been fighting this battle for a while with both commerce and media customers, where they'll ask us to get a very high cache hit rate. Meaning for every request that comes to the CDN, you want to get 99-plus percent offload. That's hard to do when you've got a pretty big library, right? There's going to be some cold content. Or if you're a retailer, there's -- you might have tens of thousands of SKUs with pictures and things like that. It's hard to get that. So we've built cache hierarchies and things like that. If I were to tell you that someone who has a 99% cache hit rate, you say that's pretty good. It's only 1% of the time you have to go back to the origin. A lot of these companies will use the cloud providers as their origin. So that's where the files will stay. And if I have to go back one out of every 100 times, you think that's pretty good offload. That's pretty acceptable. That could actually double or triple your CDN bill. The amount of fees that it costs you, the egress, the data going out to the CDN or to the user, if they're not using a CDN, it could be that expensive. So the storage is cheap. The egress fees are very high. You might say, well, why don't the clouds just waive the egress fees? It's a massive profit pool for them. So it'd be extremely disruptive to do that. So we've been attacking this problem by building -- caching hierarchies and strategies to get as high of an offload as we can. In some cases, it's 99-plus percent we can get with customers. But if they're using us, given that we've connected this to our backbone and we've got the largest -- one of the largest backbones in the world, we deliver hundreds of terabits per second. So the fees that would cost us to deliver that, any kind of cache misses, et cetera, would be insignificant. So we can pass that along to our customers and offer that as a value add if you use us.

Great. So what would a tech conference be without an AI-based question? So the spend, we had an interesting conversation after the quarter. Obviously, you've got a lot of -- you're servicing to benefit from AI-based applications. The general thought is that's basically run on GPUs. You have some plans for that, but you also have found a lot of places where you can use CPUs and so forth that still handle some AI workloads or aspects of them. Talk to us about that. And then if you're not looking to make a massive investment in GPU farms and so forth, how do you -- talk to us about how you're going to play in AI and that's going to benefit you.

Yes. So I mean, we -- first of all, the Linode platform does support GPUs. We do have GPUs in the Linode platform. It hasn't been a huge source of revenue today, but it can be. We're obviously going to look at any opportunity through the lens of, does it make sense from a profitability perspective? You won't see us going out and building massive GPU farms to say, look, if we build it, they will come. That's not going to be our strategy. But if there's customers that want to use us for that, we can -- technically, there's no reason why we can't. Now there's some interesting work being done with very sophisticated algorithms and CPUs to try to get maybe not the type of processing you can get out of a GPU but get pretty significant utilization out of a CPU. And where we sort of envision a potential differentiator for Akamai would be in the inference engines, right? So you can imagine someone may be using a big cloud provider with GPUs for their training models. But the inference engines could run out at the edge in hundreds of locations where some of the decisioning -- it would make sense to do the decisioning there. And potentially, you could run it on a cheaper CPU-like platform. Again, it depends on the workload. And so this is something that's evolving over time, but that is the potential. I think we could even run GPUs out at the edge if it economically made sense. There's no technical reason why we wouldn't. But I do think that, that is a potential for us to be very differentiated in the marketplace because we have the most distributed cloud platform that can do some processing. Maybe not as heavy as a training model, but you can do some pretty interesting stuff.

And does that -- is that a customer request that comes to you? Or is that a solution that you then pitch to the customer to say, hey, we can do this workload for you for less because we've got it on this -- on these CPUs? Or is that the customer is looking for this?

Yes, I think it's a combination, right? You can have customers that come to us and ask us for that. I mean we see it a little bit today at a much lower scale with our Functions as a Service, our edge computing. Like you can envision, let's take advertising decisioning, right? You see that today where customers will use edge computing to do A/B testing. I show an ad to one person, get this type of response. Send it to another person, get a different type of response. You could envision having more compute power to be able to do other things. Maybe you're looking at certain demographic information or making other decisions by having that CPU and that capability to do more at the edge. And maybe you're using AI to inform how you're going to make those decisions. So that could be an example of where someone might come to us to say, hey, I'm interested in doing that, but latency is a big problem. I'm trying to make a decision for somebody who's going to make a split-second decision to buy something or to either transact or not transact. I want that to be quick, and I want it to be cost-effective, and I want to do it at the edge. So that's a potential example. So we're seeing some lighter-weight examples of that today. With even personalization, you can envision that getting a lot more sophisticated. So there's, I think, a lot of opportunity. So it would be a combination of we're pitching an idea where somebody comes to us and says, I've got an idea. I want to try something. I can't do it in a cloud provider today. We have examples of that. There's a good case study about what we're doing with Apple in Private Relay. It's a perfect example where the cloud providers didn't provide enough of a distributed network to be able to run that service, so they came to us to help them out.

Okay. Great. Switch gears a little bit to the delivery business. Talk to us about kind of Q4, always a good seasonal quarter. But you've got some tough comps, World Cup last year and so forth. How should we think about that? And any room for upside in the quarter?

Yes. I mean, Q4 has always the potential for upside. I won't know that until I get to the end of the quarter, obviously. You have two seasons I'd like to talk about. One is the retail commerce season. We do have a lot of customers that have taken advantage of our 0 overage product. So we basically used to have customers that would spend, let's say, $1.2 million with us, but $700,000 was in Q4. They want to flatten their bill. So we have a contract structure to enable them to flatten out their spend throughout the year. It limits some of the upside. We still have -- not every customer buys that way, so there's still some commerce upside. But what we've seen over the last few years is a bigger, what I'll call, media cycle, where you see typically strong gaming releases that coincide with a holiday season. And then you have new devices coming online that require lots of firmware updates. So you've got new PCs, new laptops, new Apple devices, phones, you name it, that need to get firmware sent to them. So we tend to see a lot of software downloads and also a lot of video usage in the last couple of weeks of the season as well. So generally speaking, Q4s have been -- provided upside for us. Hard to call it because you don't get to see the season before you have the call, and the season is not done yet.

All right. Great. And with that, you just bought sort of the remnants of a couple of competitors, generally kind of at the low end or disengaged from the market. Talk to us about that opportunity and how that's going? And do you see any more of those -- more of that capacity coming out of the industry?

Yes, good question. So we did -- we had an opportunistic acquisition come to us on an inbound. So both StackPath and Lumen came to us and talked about how they wanted to exit the business. It wasn't really working. It wasn't a growing business. They were struggling to make money in it. So we looked at it and said, look, we can transact. Here's an opportunity. We're not interested in everything. But potentially, there's a piece of that business that we would acquire and we did. So we bought some contracts from them, the majority of the revenue basically. And I think it's -- number one, economically, it makes a lot of sense. You saw the guidance we provided. It's very accretive. And number two, for the customers, it's much less disruptive than just saying, hey, look, we're going out. Let's have a jump ball in the market. Some of these customers are new to us. Some of these customers are existing customers, which is really a share expansion for us. We did one of these several years ago. Instart Logic was a company that was in the market that exited. They don't come around very often. It's usually very opportunistic. But I'm not totally surprised that we had a few entrants leave the market. When you focus primarily on just the delivery business, especially the high-volume media delivery business, it can be challenging to make money there. We've gotten to scale, so we can make money. Hard to predict if it will happen again. If it does and the opportunity arises, we'll evaluate it.

Okay. Great. So Guardicore team has been executing really well as part of that driving some growth. Kind of walk us through the changes there and kind of how does their leadership help you guys and the success you've had this year, especially in the security side.

Yes, it's been amazing. What a great acquisition. The team is phenomenal. It's almost entirely intact. We've pretty much given them a lot of runway in terms of investment dollars. We have a -- we've built out the overlay sales team. It's much larger now under the leadership of the sales leader who came over as part of the acquisition. Pavel, the former CEO, now not only has just the Guardicore but all of our enterprise solutions in his engineering team. But they've done just amazing. We got to $100 million run rate. The business last quarter grew at over 90%. There was some onetime license revenue. But even when you back that out, it's growing at 60% year-over-year. It makes a lot of sense. The solution is great for anyone who -- no matter how good your perimeter security is, there's always something that gets in. And micro segmentation can give you the visibility that there's something unusual in your network. You can identify ransomware and malware and stop it before it becomes a big headline-grabbing event. I often get asked a lot of questions why -- that if people are using all the best security perimeter, why would they need this? Well, the problem is no matter how good your security perimeter is, no matter how well you train people not to click on links, it happens. And stuff gets in. It's always going to get in. This is a great tool to prevent it from becoming a big disaster. And so we've been doing a phenomenal job. What impresses me the most about this is the new customer acquisition from this business. A lot of people would think this is primarily being driven by our installed base. Actually, most of the growth is coming from new customer acquisitions.

So with that, walk us through kind of connection with those different products and how the cross-selling is kind of going with you guys in the different services that you have and why that just continues to be an important part of the business.

Yes. So with Guardicore, in particular, I'd say that probably the best connection would be the access control product. So oftentimes, when you are -- start sitting down and reflecting on a breach, you say, okay, well, what happened? Something got in. Well, how did it get in? Well, somebody had an infected device. They logged in through the VPN, and then they have horizontal access. So the virus started spreading and found somewhere. Okay, why do we allow people to have horizontal access? What if we were to not use a VPN, use another type of product where the device never actually gets in behind the firewall? Our access control product does that. Think of the application effectively dialing out so that if a device is infected, it can't cause as much damage as it would if it was using a VPN. That, along with the Guardicore solution, makes a lot of sense. It's a much safer security posture to have -- limit the -- what gets in behind the firewall and then also be able to detect anything that does get in and look for unusual movement of data as it's traversing through your network. So that's a pretty natural synergy there. With the web security products, our -- think of your web application firewall, your bot management detection. That's a very natural synergy with your selling motion for CDN. So if you're launching a new video service, the last thing you want is bots to come, overwhelm your site, steal information or knock you offline, a distributed denial of service attack takes you out, somebody comes and tries to defame your website, somebody tries to steal data. So it's natural to have those protections for anything you're doing on the web. So it's -- all our sellers today can sell all of our web application services. Sometimes, we'll get the CISO who is responsible for all the web properties, also owns the IT environment. He can bring us in and give us a chance at some of our enterprise security products. So there's some pretty decent synergy across the board.

Okay. And so you built a lot of this business through M&A. It's been a big part of the strategy. How do you think about M&A going forward? Any larger deals? Or how do you view that as part of the strategy and especially on the security side to bolt on different products that you don't have in-house?

Yes. So M&A has been a great tool for us. We've been pretty active shoppers, pretty disciplined buyers. Our acquisitions are either in the flavor of a tech tuck-in, smallish $100 million or less. We've done some adjacencies. Guardicore is a good example. It was a little bit larger. Linode obviously got us into new categories, the largest deal we did at $900 million. That's probably the largest deal we've done going all the way back to when we used stock for what was -- is it -- who do we think -- back into 1999, I'm kind of drawing a blank now. But it got us into the streaming business. But so we do some consolidation when it's opportunistic and it comes our way. But I don't see us doing anything transformative. We don't need to do a merger of equals, bet the company kind of acquisition. So I could see us doing more of the bolting on some capabilities, getting us into a market to accelerate us. API security is a perfect example, right? That's a huge opportunity. A lot of threats vectors are being exploited now for open API gateways, and there's not a lot of good solutions out there. So there's something you could either build it yourself or jump start your time to market. If you think about the time it takes to build a product and bring it to market to get it to scale, sometimes they can take 3 to 5 to 7 years. If you could buy something like Guardicore, we brought that from less than $30 million to over -- a run rate of over $100 million in a little over 2 years. That's a great tool for us, and we'll continue to look at M&A and continue the same posture of being very disciplined in terms of what we're going to buy and look at a lot of different opportunities.

Okay. So back to AI, you talked about the threat of AI-orchestrated attacks and how you think you're positioned to protect from that. How is that going? Are you seeing those in the marketplace today? And then how -- what are customers coming to you, looking to you, to provide on that?

Yes. So I think just in general, the sophistication of the type of attacks that you see are only going to get greater. For example, think about this conversation now. It's being taped. It's public. Somebody could stitch together my voice, send a voice mail to somebody on my team, asking them to wire some money somewhere, right? You can probably find the words to put that together. Send that off. It sounds like it's me. They could maybe figure out what my phone number is, spoof the phone to make it look like it came from my phone. E-mail attacks now are getting much more sophisticated. You see a lot of workflow-type attacks, where some of it looks like a DocuSign or something that would be -- maybe looks like a expense request. So e-mail phishing is getting a lot more sophisticated. And so all that does is really just increase the vulnerability for companies, and they're always looking for ways to solve whatever the most recent attack is. I think in particular, what I just described really opens up the door for Guardicore in particular, right? Because there again, the world is moving so quickly that the risk of something getting in, no matter how good you think you're prepared, more likely than not in the next -- pick your time frame, something is going to get in. So this just -- it gives you a great way of preventing massive damage.

Great. I got a couple more questions, but see if anybody have any questions in the audience?

I was just curious [indiscernible]

It does.

[indiscernible]

Yes. So we've had a few announcements that we put out with some of the major SIs. I think with Guardicore, in particular, the system integrators, it's a pretty interesting business. If you think about -- they're in the business of selling services. The journey to roll out micro segmentation across a really large sophisticated IT environment is -- it can be a pretty big lift. And there's -- you get into the sort of awareness phase of getting all the information and determining what flows of data are going across your network and you want to set up rules, and you have to adopt those rules so you're blocking certain traffic. Businesses aren't static. They change and evolve. So there's the need to maintain those rule sets and things like that. So I think there's an opportunity for some of the system integrators who did some digital transformation sales, including a Guardicore, where say like a web application firewall is probably more of a week or 2 of installation. It's not really as interesting. But almost all of the sales from Guardicore are coming through the channel right now. And we've been able to add to that channel. And as I talked about a few minutes ago, we're getting a lot of new logos in certain verticals that we're not always that strong in as a result of that channel. So it's been a really nice addition as part of the acquisition.

All right. Great. Well, got the CFO, you got to ask the capital allocation question. Kind of walk us through the priorities there between capital intensity and share repurchases. And then, of course, what's the thoughts on long-term paying the dividends, something like that?

Sure. So obviously, a great business. We produce a ton of cash. So in terms of fueling the CapEx needs, that's just -- that's not a problem today. Maybe down the road, if it becomes a massive business for us, I'd love to have that challenge someday. But that's not really a big thought in terms of the capital allocation that just comes from operations. So far, stock buybacks have been the primary use of capital allocation in terms of returning that to shareholders. We've reduced our share count pretty significantly over the last few years, but averaging a little over 1% a year for the last 10 years. So we'll continue to buy back stock. M&A is a great use of capital. So we'll continue to look to put our capital to use for smart M&A. Dividend, we haven't paid a dividend yet today. Something we can -- we always look at and discuss, but haven't done that yet. But we're certainly the type of company that produces a lot of cash. And maybe someday if that makes sense, we'll do it. Something we talk to our Board about all the time. And we think right now, the use of capital going to M&A and to buybacks is probably the best use of our capital at this point.

All right. Great. Any last questions from the audience? All right. Ed, thank you very much. We really appreciate it. Thanks, everybody, for being here.