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

All right. Good morning, everyone, and welcome to TD Cowen's 53rd Annual TMT Conference. My name is Michael Elias. I'm the Communications Infrastructure analyst. For this session, it's structured as a fireside chat. We have about 30 minutes for this. I've said this a few times. I promise I'll open it up for questions. I just get captivated and end up not opening up for questions. But for this session, I'm honored, we have Akamai and from Akamai, we have their CEO, Tom Leighton. I have a bunch of questions prepared. Like I said, I'll pause and take questions. But Tom, thank you so much for being here. I always appreciate it.

No. Thank you.

I want to kick things off talking about strategy, particularly the go-to-market transition that you guys were talking about earlier this year. I believe the specific language that you're talking about is you wanted to increase the ratio of hunters to farmers. So my question for you is how has the go-to-market transition gone? And as part of that, maybe points of friction and points of upside where things have gone better than expected?

Yes, we're in an evolution to get more focus on hunting. With our newer products, we have access to a broader landscape of customers that weren't CDN customers or didn't care about CDN, but they need the new security capabilities. They're interested, obviously, in cloud computing. They care about the cost of that. And so there's more of a focus on hunting for those customers and more of a focus on the specialists that can help us sell Guardicore segmentation, which is -- it's a little bit more complicated to introduce segmentation into a company. API security is brand new. And so we have specialists there. I think over a couple of year time frame, that will be the kind of thing that any rep at Akamai can sell. And cloud computing. We're hiring folks with a lot of experience selling cloud infrastructure services. And I think that helps with our revenue growth. So that's the evolution that's underway, making good progress there.

So I want to build on that point. So when I think about Akamai historically, you were CDN expanded into security. And at the Analyst Days that I've attended in the past, it was very much of, hey, we want to sell the security solution set or penetrate the CDN base with the security. Now as I think of compute, I would argue it's a different buyer within the organization of the compute relative to security. So as part of that, you need to kind of go -- you can still penetrate the installed base on CDN and security, but you do need to essentially go out and hunt for new customers. As you think about going out and hunting, I mean, what are the explicit avenues that you're finding have been most effective? Is it through a channel partner? Like how are you identifying these new customers?

Yes. Channels is the majority by far of new customers and bookings. So we have channel partners, both on the security and the compute and it's very valuable for us. So that's the lead.

So it's pretty much working it through the channel is the idea...

Yes. And that's different than CDN wasn't so much that way.

CDN wasn't a channel business historically?

Not really. We had channel partners, but it's -- you didn't really exercise the channel that much for this. But the security sales and compute, very channel friendly.

I want to talk a little bit about -- on that same earnings call where you were talking about the go-to-market transition. That was a very interesting earnings call for me. And as I was listening to it, I heard you guys -- you were essentially kind of changing your stance on growth for the business. And from my perspective, it was very much in line with the things you've been talking about. You have been saying for security that the larger product we're getting to the plan of maturity. And then even within delivery, we have seen the trends for that. So my question for you is what led you to kind of reset the bar and decide, hey, this is the moment. And as part of that, I was a little bit surprised with kind of the step down in compute, particularly. So I'm just curious, what was the catalyst for that?

Well, we haven't given long-term guidance for over 3 years.

Since May 2022?

Yes. And a lot has changed since then. And we thought it was important to get our current view out to the community. And yes, we took security. We think going forward about 10% growth and the factors there that we do have a mature -- more mature product set for a lot of the revenue. We have the market-leading WAF, but with good penetration, market-leading bot management solution, again, pretty good penetration. You look at Prolexic and scrubbing, again, mature technology. There's still growth in all these areas. We're the market leader, but pretty good penetration. So the growth rates there are not going to be what they were 3 years ago. Now you look at the new capabilities, which have a lot of runway, very strong growth. We've called out with Guardicore and segmentation, we're the market leader. And with API security, we're the market leader, a lot of potential there. We think we're going to grow the ARR 30% to 35% this year. Just announced firewall for AI, just starting, but boy, there's a lot of interest there. And so we've got this dynamic where you've got the newer things, lot of potential, growing very rapidly, but it takes a little while for them to swing a $2 billion number which today, that mass is dominated by leading products that have a little bit more saturation, I would say, and so less future growth. And we thought it was important to update the community with what we see going forward.

And then as part of this -- so that explains the revenue dynamic today. But also -- the other part was about the operating margins. We've been talking about 30% operating margins for a while. To me, the way I interpret it is like saying we're going to get back to 30% over time as there -- as you invest in these new platforms, there's requisite investments that have to be made and as part of that, that comes as the near-term hit to the margins. So as we think of the path back there, what are the critical investments that need to be made for you to get to, let's say, that 10% growth that you're talking about at the end of the decade?

Yes. So well, first, we hit 30% in Q1. So we're not far. We have guided to a little bit less for the year. We've been close to 30%. We do think it's important to operate in that range. In terms of investment, in terms of new security products, there's investment there in the technology. There's investment in our compute infrastructure to increase the TAM we can go after. There's investment in the infrastructure for compute. So there's the CapEx, which we depreciate over 6 years, but that has an impact on the operating margin. And also the way the co-lo works for that, we get hit by the accounting. So gross margins take a little hit there. So those are sort of the factors. Also, as we talked about with the sales force, we'll be growing that over time to increase the capacity and the bookings as we have more products to sell. So those are the areas that we're investing in, and we're doing it in a judicious way. So there is very modest growth overall in our head count as our revenue grows.

To that point, you have taken resources from the delivery business and reallocated them over to compute. And it does seem to me like that's compute and obviously, security given how big it is as a percentage of the business. But compute seems to be the big incremental focus for Akamai. Would you say that growing the compute business is the top priority for Akamai?

Well, yes, we love all our children. Security is really important for us, and compute is really important for us. And we still have the world's largest and best CDN. So we're investing a lot less in it than we used to. So we can plow more of the investment into security and compute. But I would say, they are twin focuses for us.

Okay. I appreciate that. Now let's talk a little bit about compute and then we'll talk about security. And hopefully, if we have a little time, we're going to -- we'll talk about delivery. But for compute, key topic of this conference today has been -- what we're seeing on the data center side with AI. Before we go to AI, what I would ask you is, within the compute business, what are the biggest opportunities that you're seeing? And I'm curious if it's actually related to AI or if it's just the core cloud computing that has been the base of, I'd argue, the data center market for the last 10 years?

Huge opportunity in core cloud computing. I would say, the center of mass for us, the sweet spot is media. Because big media, big Akamai customers, they like us a lot. They're spending a fortune in the cloud. They can't afford it and they're cutting huge checks to their biggest competitor. And they pretty much all have directives that on the top down, they got to stop doing that. And they also care about performance. And with our compute solution today, we can offer them better performance at a much lower price point. And by the way, we don't compete with them. We've also built out a media workflow ecosystem of partners that I think very competitive with what the hyperscalers can offer. So that's the sweet spot and where center of mass of the revenue and early adoption is. That said, there's interest across the board. Customers that want to save money or want lower latency by getting their containers closer to end users are perfect for our cloud.

On the media side, I'm curious, is it a function of, given the volume of traffic that they're moving, they're getting killed on the ingress egress...

That's the big part.

And that's the point of differentiation for you guys?

Big part of it and then also the performance associated with that.

Okay. I can see that. As part of that, one of the thing that -- when I think about the compute business, it's a CapEx-intensive business or can at least, right? You're fitting out data centers that you're leasing or if you're the hyperscalers, you're building out your capacity and fitting that out. So as part of that, I'm curious, like how do you think about the ultimate CapEx investment that needs to be made in order for you to get to the scale of the compute business that you ultimately envision, right? How do you think about that CapEx trajectory?

Yes. So we had an initial build-out to get started. And the strategy is as we fill that up, and we see the demand, we add more capacity to accommodate the demand that we see in the next 6 to 12 months. And so the faster that grows, then the more investment we make as we go along. At a very, very approximate level, $1 of CapEx can generate ballpark $1 of revenue per year. And so it's a profitable business for us.

That makes sense. Now I want to take this a step further. We are seeing major changes on the data center side in terms of how the data center is architected. And specifically, I think it's something that Satya said, which is, we don't want to have an AI fleet and a cloud fleet, we want to have an everything fleet. And to that point, as I think about inference, it increasingly does seem like it is going to run at a higher rack density. And as part of that, it kind of needs to live in the new purpose-built facilities. When I think of your compute platform, you've done this Gecko initiative, which is to take those 4,000 sites that you have within the carrier networks and then integrate those within the dedicated data centers that you got from Linode, plus the expansion ones. As I think of the vendors that you use for data center capacity, I mean, those are -- tend to be interconnection facilities that I would argue aren't equipped for 150 kW per cap. So the reason I'm asking all this is as you think about capturing the AI opportunity, particularly inference, do you view the data centers that you currently occupy as a limiting factor for your ability to penetrate that market?

No. So we have 2 kinds of data centers, if you will, for the compute business. We have the core data centers, which are very much like you'd see a hyperscaler have. And we're in 36 cities, 41 data centers and they're perfectly well equipped. And then we have which called Gecko, it's sort of the internal name, the commercial name will be managed container service. And that's where we're deploying customer containers into our edge [pops]. And we're not going to deploy it in all 4,000. Today, we're now running in 100 cities with that and that's not for the heavy-duty core computing. That's more lightweight containers where the customer wants to have their business logic close to the end user for performance reasons. And you don't need the super heavy data center for that. Now obviously, inference, I think, is rapidly becoming more efficient that -- you don't need state-of-the-art GPUs and zillions of them like you would for training a massive model. It's much lighter weight, might be a medium-sized model, probably running GPUs, but we're running some today on CPU just fine. And I think you'll see continued improvements in the algorithms underlying what an enterprise needs for its particular AI application. It's going to get a lot more efficient. Just it was -- there was never a reason to make that a more efficient before and the only use for this kind of technology was the graphics for gaming, but now you're using it for everything. There's going to be continued, I think, improvement in the underlying algorithm.

So my next question was going to be, over the long term, do you think -- because I personally do it if my interaction with AI has been any indication, I do think this is a generational technology and it's here to stay. Now as part of that, one of the things that I think through is we have power constraints in these markets, so things need to become more efficient. It sounds like you believe that they will become efficient to the degree where you can essentially run them on CPUs, potentially even at a rack density comparable to that of cloud AI, which I'd say, is like 14 to 17 kW per cap. That's what I'm getting from...

Yes, today, I think the spectrum is CPU to mid-range GPU for most applications.

And as part of that, the footprint that you have would be suited in your view to accommodate those workloads?

Yes.

Okay. That's very helpful. One of the things that was interesting to me is you guys in-source a lot of third-party compute that you're doing with others. I think part of it is the ingress egress dynamic, right, that drove you to like, hey, let's bring this in-house. Having brought it in-house, has there been anything that you've learned that has ultimately translated into success with customers where you can say, hey, look, we just did this and this is what came out of it. And as a result, we can drive similar benefits for you. Has there been anything to that degree?

Well, I think eating your own dog food or drinking your own champagne, whatever it is, is always good to do because you do find issues with the product, and you get those fixed before a customer has to go through that. And you do learn to appreciate what the customer needs to do to use your capability. So I think that's been very helpful for us.

All right. Let's -- I want to transition over and talk about security a bit. So obviously, we hit it in the beginning in terms of the shift there. But of those products, particularly like your Guardicore offering, I'm curious how are those growing relative to your expectation when you acquired them? Like how has the platform performed? And then as part of that, as you think of the next cycle of growth, particularly as you align the go-to-market, how do you think of the upside opportunity for those, just as we try and frame it for investors?

We're very pleased. And we've forecasted the ARR to grow 30% to 35% this year. So very, very strong growth. So both Guardicore and the no-name acquisitions, we're very happy with.

So along those lines, we talk about how those businesses aren't large enough yet to move the needle. As we think of the road map, this could be fun with math, if you will, as we think of the road map for them becoming contributors or meaningful contributors, how far away do you think there are? Is there a magic number that you get to once these businesses are scaled up where, okay, we crossed this revenue threshold, then it puts us in striking distance of driving that top line growth that you've been talking about at the end of the decade?

Yes, pretty quickly because when you're growing at 30% to 35%, you get big pretty fast. And so certainly, by the end of the decade, those are major contributors.

And do you think the business can grow at 35%? Because obviously, you run into a law of large...

Yes. Right now, we've projected this year for the ARR. And we think over the longer term, 10% is a good target. And there'll be new products that factor into that over that time frame as well.

I want to ask you because, of course, we have to talk about AI with everything we do. I want to talk about AI as it relates to security. One of -- I was having a conversation with somebody recently, and I was saying that I believe the future belongs to the person who or the people who could most effectively weaponize AI. But when I said weaponize, I didn't actually....

So the future belongs to the bad guys, is that...

That's -- I'm making this point is not because I didn't mean it like that, I meant it like who can most effectively harness the technology for productivity. But then that brought up a separate conversation, and that's why I'm bringing this up to you around threats from AI.

They're very real.

So to that point, how do you see the AI and kind of the state of what you've seen from AI and how it's impacted the security landscape? how do you see that being a potential growth driver for the business? And can that be almost like a generational driver for the security business?

Yes. It's already a generational benefit for the attackers, both in terms of deep fakes, generating malware, that's why we're seeing, I think, more penetration, more compromised devices used as bot armies. So very helpful to the attacker. Also, the widespread use of AI today already has greatly increased the attack surface. It is surprisingly easy to trick your AI application or agent, to give up information or to do bad things. And that's why you need a special firewall just for your AI applications. Just a regular web app firewall doesn't solve the problems you've got that are unique to AI. So that creates demand for that. And then just using AI is, I think, very beneficial. We've been using traditional AI from the beginning in our security products. Now we're using it, especially with the interfaces. So an operator can see human language discourse what's going on inside the enterprise, what firewall rules aren't up to date, what is happening that shouldn't be, either an attack or just inefficiencies, what's happening inside their enterprise infrastructure, very useful. And that's using the new Gen AI capabilities.

Yesterday, I was thinking about this. And one of the things that I think AI could be useful for is as we humans think about things, we tend to think about things on a granular level and then build up to the system level. But where I think we're blinded is that we don't see the interdependency within systems. And I think that AI can be very helpful in architecting a system and helping call out essentially the vulnerabilities there. So that resonates with -- what you just said just resonates with me. But I wanted to ask you, how does the -- how has that tax changed as a function of AI? Is that the volume has increased? Is it the sophistication?

Yes.

Oh, my gosh. Like, yes, I'm just curious how that...

You use it to write code.

So you're using it the write code that then...

You teach it to get around the defenses that are out there.

And it can learn faster and faster?

No, yes, it's very good.

Got it. Okay. All right. I would imagine there's a corollary on the security side as well, though, just like it is in iterating, you are also iterating there. So perhaps the 2-day -- because I do get worried this, that's why I'm asking...

There's a scary side to the AI, yes.

Okay. So I do want to transition and talk a little bit about the delivery business. I think the latest revenue guidance was you're going back to -- the plan is to get back to flat, I want to say...

Well, yes, we'd like to get it totally stabilized. And in the meantime, we're trying to get it to be -- that the declines are much less deep than they were last year.

So for me, it's price and quantity, it's price and volume. That's it. So from that perspective I want to talk about the quality, the volume side. It seems like post COVID, there was a rerating of growth, right? And at least a period where things have to normalize and then from there, you see a reacceleration. How have the traffic patterns changed, not only in terms of the growth, but then also underlying traffic patterns? Have you seen changes in the way that traffic is moving around in a post-COVID world?

In Q1, as we talked about, traffic growth got better. We're keeping a close eye on Q2 to see where things go. There's reasons why it should get better. There's a lot of traffic still to come online. A lot of improvements in quality that can lead to more traffic. We'll see about AI, but if everybody starts generating lots of videos and stuff like that, anything that's video or big software generates traffic. And that's helpful. I think probably saw share gains, 4 of our leading competitors are out of business. And that also probably helps the pricing dynamic. It's still competitive. Not as crazy as it was or there's fewer players being crazy with pricing.

Do you think -- I appreciate the point on the video side. Do you think the usage of like as you think of like ChatGPT Grok, obviously, it's feeding you text. It's outputting text quite quickly. So my question for you is, do you see the actual like ChatGPT Grok being a benefit to the traffic? Or is it just so small in the grand...

Text is too small to worry about. If it's generating video, that's a whole different game.

So that can be a needle mover?

Yes.

Okay. I feel like that gets into the territory of like multimodal AI, and you tell -- you give it a prompt and it's giving you a complete video. We'll see when we get there. I feel like it will happen. Okay. So that's on the traffic growth side. Just as you look out, I do feel like we've gotten through the normalization period post COVID, do you think so?

I would think, yes.

All right. So traffic growth should improve. Now on the pricing front. Pricing has just continued -- from my perspective, continually declined within the CDN space. Any reason to believe that dynamic changes?

It's declining a little bit less steeply, and there's fewer players out there that are selling below their cost, which is what contributed, I think, to the tough market conditions. We'll see what investors do. But for a long time, investors awarded companies for growth even if they were losing money at it. Maybe there'll be a little bit of a shift there, we'll see. So there certainly are fewer players that are upside down that way. And that helps with the pricing dynamics.

So that's what I wanted to ask, which is that, obviously, you have had a role in consolidating some of those peers. I appreciate the media companies. They do -- they have their own CDNs, a bunch of them do. Do you think that as the market rationalizes further, I guess, one, do you expect the market to rationalize further? And then the second question as part of that is, does that accrue to the benefit of pricing power and changing the rate of declines? They'll still decline, just instead of declining at 20% or 25%, they'll decline at like 10%?

Well, that's what we'd like to see happen.

Do you think there's more consolidation to come though?

There is, potentially.

Okay. All right. where I'm getting at is like you mentioned the upside down dynamic. And the CDN business is obviously a large business that's generating a lot of cash flow for you guys. So part of that, the question that we get from investors is, is this a business that gets back to flat? Or is this just a perpetual leaky bucket is the way that it's been described by investors to me? So from that perspective, I'm just trying to get at -- I can see the pathway for the traffic growth. I think where I'm struggling is on the -- if you have periods where the traffic growth isn't enough to countervail the price declines that how you get...

Yes. As we've guided over the medium to longer term, we would like to see continued improvement there. And the goal is to get close to flat for that business. And I think there's good reasons why that can take place. And I do think some of these -- 4 other companies going out of business are good examples of why the ones that are still in business should not continue to sell it below their cost because they'll go the same way as the others did and that should lead to a better overall environment. Delivery is not going away. Traffic is not going away. I don't believe. I think it will continue to increase. And so we just got to get pricing rationalized.

I want to add, this is my last question. This is a philosophical question for you because you've been with the company since the beginning. So initially CDN business. You started to diversify way. I mean, it makes sense to me just given the -- what we just talked about with the delivery business. So you went into security. And you had a twin engine story for a long time. And this is from my perspective. It felt like you went at it, the market didn't quite give you the multiple of the security business, of like a pure-play security business. So then you pivoted again and then you now instead of 2 engines, it's 2 engines deemphasizing CDN, now it's compute and security. I guess what I'm getting at with this is as we think of the long-term picture, is compute the stop and you run out that opportunity set and you keep growing security and that's it? Or do you add other verticals and Akamai effectively becomes like a conglomerate of different tech verticals?

No, I think you want to Akamai as a cloud company. The hyperscalers have compute, that's their biggest business. They also do delivery and they do security. So basically, what we're doing today and focused on is exactly what you'd see the big cloud companies doing. And I think it makes sense. It's very synergistic. For security, we're applying the security at the exact same time and the exact same processor as we're doing the delivery. Compute is very synergistic with delivery. We talked about the big media companies being our big early compute customers for a good reason. It's very synergistic. And so I think really the compute is sort of the last big leg of the stool in the platform. We've always done compute in terms of the lighter weight stuff. JavaScript at the edge, cloudlets, now we're getting into the core compute and we're putting containers on the edge. It's really the last missing piece. So I don't think you'll see us do some fourth thing and it's not meant to be a conglomerate. It's what makes sense for the customer, what they need us to be doing. And it makes us fully competitive with the hyperscalers.

So it was an adjacency that you felt was appropriate to add and then that's where you're going after?

Yes.

Understood. Got it. I'm sorry, I promised I'd open up for questions, and I just got captivated. So I apologize. With that, Tom, we're out of time. Thank you so much for being here.

Very good. Thank you.