I think we're live. Awesome. I'm Roger Boyd. I cover Cybersecurity and Infrastructure Software here at UBS. I'm very happy to be hosting the Akamai team. Tom Leighton, who's CEO and Co-Founder of the company. So thank you for being here, Tom.
Ah, it's very nice to be here.
Yeah. Awesome. I think maybe we could just start with the kind of cliché summary of the Akamai journey, 'cause it has evolved quite a bit in the past couple of years, and really a lot over the last quarter of a century, really, being maybe the first edge platform to really be out there. If you could just start with kind of the overall kind of summary of what's happened over the last couple of years as you evolved to support more edge use cases.
Yeah, we started with content delivery and the first, and today still by far, the largest edge platform, with 4,000 POPs in over 750 cities. And then about a little over 10 years ago, we created the web app firewall as a cloud service marketplace. We're the leader there by far today, and API Security is now our largest product line, bigger than delivery. And most recently, we've been putting a lot of investment and focus around compute. We've always done compute, think of function as a service on our edge platform, but now we're supporting containers, VMs, more mainstream compute. Initially, in core data centers, and now we're in the process of moving that support out into some of our edge POPs.
So that, you know, the goal will be to have really fully distributed support for containers and VMs, which is unique in the marketplace.
Yeah. That's interesting. Maybe just to follow on with the delivery business, and as you pointed out, until this year, it was the largest revenue segment. It's been slowly declining for a couple of years. Can you just talk about what's happened in the CDN market? We've gone from a somewhat concentrated market to a point of extreme diversification and now back to some consolidation from an end user perspective. What's the latest there? And, as a leader in that space, what's the health check on the overall delivery market as it relates to pricing, traffic growth, demand?
Yeah, delivery is a very competitive market. You know, it's a marketplace we created 25 years ago. So you can imagine it's a lot of commoditization has taken place. We're the market leader by far. Some firms have us with a majority market share, but there's a lot of competition. And you know, there's price pressure associated with that, including the hyperscalers, who in many cases will give delivery away for free if you use their compute. In fact, as we get into compute, we'll do the same thing. You know, you give us your compute, which often will be 10x the revenue associated with delivery-
Mm-hmm.
Yeah, it makes sense for us to, you know, give discounts on delivery as well. You know, one change for us in delivery is we're doing less of the really spiky traffic, where we don't get paid enough for it. So that, you know, hurt revenue a little bit, but improved profitability. A lot less spend on CapEx as a result, and we're using that investment actually for compute now, which gives us a better ROI.
Makes sense. To that point of consolidation, Akamai recently purchased the CDN business from two smaller vendors who decided to leave that. What do you think, or what is your view on what those exits from the CDN business suggest about the competitiveness in that market, and particularly the challenges delivering a profitable business at scale, in this environment?
Yeah, there's a lot of competition. I don't think any of them is profitable.
Yeah.
You know, Akamai is very profitable. We're unique because of our architecture. You know, we're deployed inside, you know, about 1,400 other networks. Most of our POPs, we get free colo, free power, free bandwidth, and so it generates a lot of cash for us. We're very profitable at it, but the competitors don't have that capability, and so they're not able to generate profit. And in this case, you know, two of them independently decided they wanted to terminate their CDN business. They wanted to keep their customers for other products, and so they wanted a way that they could get the CDN out but not hurt the customers. So they both independently approached us as the market leader and said, "Look, we want you to take over our CDN customers.
You know, give them good service so that their customers stay whole. The economics were very compelling for us. It made a lot of sense for us to purchase the contracts. We get about 200 new customers that now we can cross-sell other services. So financially, it was good for us, it was good for the customer, and it was good for those companies.
Yeah. Yeah. I read that as from a customer acquisition perspective, a lot more favorable than you going after those customers-
Yeah
in the market.
It was very easy. It'll be, you know, three months migration period, and very very profitable, you know, transfer to us.
Yeah. On that cross-sell opportunity, 200 new customers, I think the overlap between those two customer bases and the core Akamai base is fairly low. Can you just talk about how the customer bases differ from what you typically service? And, on the cross-sell side, what how much of an opportunity do you see to sell things like compute and security?
Yeah, you know, I would say at a high level, they are comparable to our base.
Yeah.
It's just accounts that, you know, they'd had for whatever reason, maybe, you know, they were a large Lumen customer for other reasons and had taken on the compute or StackPath, similar story. So they're not a fundamentally different kind of customer.
Mm-hmm.
They're in our target customer base, which is good because then it's very natural for us to sell them our security solutions and ultimately compute as well. You know, big delivery customers are our first target for compute.
Yeah. It's a good segue to the security business. Again, as you pointed out, it's now the largest revenue segment for Akamai. Historically, that had been growing kind of mid-20s growth. It slowed a little bit last year and has now re-accelerated. Can you just unpack? I know you went through this on the third quarter call, but what's contributing to that re-acceleration? What's working well in security, and what are you seeing competitively there?
Yeah, we have several market-leading products. The attack landscape has increased, which has increased the need for our capabilities. You know, we have been pleased this year that, you know, our security business is doing a little bit better than we'd initially forecasted. Now we're projecting for the year, 15% growth. You know, most of our revenue is in app and API security. That's led by our web app firewall, which is the leading solution in the marketplace by far.
Mm-hmm.
Still have, you know, reasonable growth there, even though we've been in that space now for over 10 years. We built bot management on top of it, now at about a $250 million a year run rate, with good growth. Again, market-leading solution by a wide margin. You know, there's other capabilities that are smaller that we've built, that are growing rapidly but on small numbers. I think the exciting thing for the future in that segment is API security-
Mm-hmm
... which is, I think, just being recognized as something major enterprises have to have. You know, our leading competitors there are startups. I think we're in a very, you know, good position. Zero revenue today, effectively. It's a new capability, but we've already made a very nice integration with our web app firewall, something we call the Easy On button. So very quickly-
Mm-hmm
... we can, you know, set it up with a proof of concept and get it going for customers. And I'm optimistic about, you know, substantial future growth there. We also have the scrubbing segment, the anti-DDoS, you know, solutions. They're very mature. You know, we're starting with Prolexic, and so generally, we don't see as much growth. But, you know, from time to time, things happen. You know, Killnet, for example, here in the U.S., took out a lot of medical centers that became large Akamai customers. And that was a global, you know, attack base that, you know, helped with the Prolexic revenue this year. And then the third, you know, pillar of security is enterprise security, led by Guardicore.
Again, the market leader by a good margin, nearly doubled revenue year-over-year in Q3. I think with, you know, the more and more penetrations and now GenAI being used by the attackers, everybody's pretty much everybody's getting penetrated. And so there's a much stronger recognition that you need that interior layer of defense, and that's what Guardicore provides. So yeah, malware getting in, the key is to identify it quickly and proactively wall it off, not let it spread. And that's what Guardicore does really well. And I think that's, you know, leading to the, you know, substantial growth there, now at about $100 million a year.
Yeah.
We've been very pleased with that.
On Guardicore, I kind of think about that as... I mean, you're obviously a very big, large and successful business in the web app firewall and DDoS, but I think about that as kind of the new age crown jewel in the portfolio right now. Lots of growth there. It's a market-leading solution. Can you just talk about how important that product is from a growth perspective, but I guess more so from a go-to-market perspective and a channel presence, a product that can be sold outside of the typical CDN sales motion, and how that's helped you kind of along in that segment?
Yeah. I think you captured the advantages, you know, very well. All those things are pluses for Guardicore. In fact, you think of the go-to-market motion with our API security also is independent of CDN. Really unrelated to CDN, we're leveraging our Guardicore go-to-market motion for API security now-
Mm-hmm
... which has been very successful for Guardicore, and we want to, you know, carry that over as we grow API security. Also, Guardicore will help us with our other enterprise products, like Enterprise Application Access. Early in the year, in Q1, we're gonna have Enterprise Application Access on the same control plane as Guardicore, and that will enable our customers to unify their approach to north- south and east- west, as those solutions are known, and that'll be a unique capability in the marketplace. So I think Guardicore very important for us.
Yeah. I'll echo the comments on API security. It seems like it's getting a lot more attention from CISOs and operators as a growing area of risk. How big do you think that market could be relative to something like web app firewall? It seems like over time, that should be something that's pretty sizable.
Yeah, I think that's the right compare, and it seems like it ought to be at least as large as web app firewall. At a high level, it's very comparable.
The whole world is, in a short period of time, moving to APIs. So this provides the defenses there that web app firewall provides for web apps. So should be as large, you know, and, and we want to become the market leader there, too. We've got a good, I think, head start in, in doing that.
I guess on that point, you pointed out that a lot of your competitors in that space are startup companies. What do you think positions Akamai from a delivery standpoint, from a web app firewall standpoint, that allows you to compete better there? And can you talk a little bit about the asset you bought in terms of the breadth that Neosec provides, everything from visibility to detection? That's one of the more broad solutions out there.
Yeah, we thought it was the best solution. And it's the only solution that records, you know, the entire transaction history so that you can call that up and get the replay. You know, customers first need it to get visibility into all the APIs they have. You know, most big enterprises, the CISO or the CIO will tell you they don't even know all the APIs they have, 'cause their developers have created them, put them out there, and who knew? And so first thing we do is tell you what you got. Then we tell you which ones have vulnerabilities, then which ones the vulnerabilities are being actively exploited, and then, of course, ultimately block that. So yeah, very important.
It's, you know, a few years ago, didn't matter because people weren't doing so much of the APIs, but now that enterprises are relying on the APIs for everything, the attackers have learned how to exploit them, and there's a lot of vulnerabilities there.
Who do you think about as the buyer of a solution for API security? Is that the team that's buying web app firewall? Is it the developers? Is it kind of a DevOps mix?
It's like a web app firewall, so it's your CISO or CIO, depending how you're structured.
Maybe shifting to compute, it's been a pretty big part of the story over the last year or so. Can you just outline again the vision of a Connected Cloud, leveraging the fact that you are the most distributed edge presence out there, the importance of connecting that back to more centralized compute? What does that look like over time, and where you are today in that building out that vision?
Yeah. So today, on our edge platform, in 4,000 POPs, we run JavaScript. So we'll spin up JavaScript apps, function as a service in a few milliseconds based on user demand. And that's very valuable. You could do a lot of things with that. But 99.9% of the compute revenue is in sort of the core cloud compute containers, VMs, you know, as a service. And that's what we now are able to support, starting with the acquisition of Linode. Now, Linode was focused on small and medium business, which is not our target market, and we're maintaining their business. It's growing in the mid-teens, you know, so we've maintained that.
But we made a very large investment to upgrade it so that it can be used by major enterprises for the mission-critical apps, and that's where all the money is. And so initially, we're supporting that in core data centers. We've added 13 new ones. Linode had 11. We're working on upgrading those 11. We're in the early days of actually now migrating the compute capability into our edge regions. Now, we're not going to go to all 4,000. You know, initially, you know, this year we'll be in about a dozen. We're in beta now with a couple customers. Next year, we'd like to add dozens more. But the idea there is that we're going to support containers and VMs as a service in many cities around the world where the hyperscalers don't have a presence, and that'll be pretty cool.
And then after that, you know, we're working on it now, but it won't be available next year, but we want to spin those up on demand, based on user demand, just like we do with JavaScript. So you don't have to pre-plan and pre-provision manually, in many cases, where your VMs are, how many you've got, where your containers are, but it all happens automatically. And that'll be a, I think, a pretty exciting development. So we get better performance, and of course, we're also at a lower price point than the hyperscalers. And, you know, some of our customers really need to save money. For example, big media-
Yeah
... you know, I think, is our first target segment. They care about performance. They're already using us for a lot of their delivery. They have huge bills with the hyperscalers. They need to save money, and we're a very natural choice. In fact, it was through conversations with many of them that really led us to decide that, you know, we wanted to embark on this.
Yeah. I want to touch on some of those early use cases you're seeing, but you've famously been dogfooding it yourself, and Akamai's been moving some of your own cloud costs into your own data centers. What's the status there, and what have you learned from kind of reliability, scalability perspective that you think you can take to other customers?
Yeah, we're our first big customer. You know, our cloud bills just were spinning out of control, like our customers tell me is happening to them. And, you know, so this year, without, you know, migrating our own apps into our, our own cloud, would've spent over $150 million. We'll probably spend a little over $100 million now, saving ballpark $50 million, which for us is a lot. You know, and next year probably would've been over $200 million, as the rate it was growing, was growing at a fast clip. And, you know, we want to cut our spend in half from this year, so huge, huge savings. We've learned a ton about it. You know, it's not, it's not trivial to do. You know, you don't flip the switch.
You know, so we have that understanding with working with our customers. You know, I can tell you the... You talk to the CXO at the, the big media companies, and that's an easy conversation. 'Cause, yeah, better performance at much lower cost, they want to flip the switch, but it takes time. You got to get the development teams engaged. It takes effort to do. You know, if you look at us, and it's really over the course of the year that this is happening, but massive savings and better performance, and so very, very worth it. Very pleased.
Yeah. From a customer perspective, I think you've talked about some of the early use cases being a very small fraction of your customers' overall cloud spend, but in terms of materiality to Akamai, it's fairly substantial.
Yeah.
What are the initial use cases you're seeing? You talked a lot about the fact that the media vertical is kind of the prime opportunity off the bat, but what other verticals are you thinking about as kind of the next steps there?
Yeah. It really goes across the board, but it's applications where performance matters. If you're doing QA or some central back office thing and performance doesn't matter, okay, fine. Probably not an initial target. It's customers that are spending a lot, especially on egress fees or per hit charges, 'cause those costs will go way down with us. If you're not doing that, okay, not very compelling. If you've got applications that are on containers and VMs and don't use a lot of third-party apps as managed services, that's a good candidate. If you've got something that is really locked in deep to the hyperscaler, forget it, 'cause that's gonna be a lot of effort and there's just easier things to do.
Yeah, our goal is initially to get to, say, 1% of the cloud spend, which probably a couple billion dollars, which is huge for us, trivial in the marketplace. We're focusing initially on big media. Commerce is related. We're PCI compliant today for our use, but it won't be till next year we're PCI compliant for general use, but then commerce becomes viable. Also, media and commerce are good because those companies compete with their hyperscaler quite often, and we don't. And we'll protect their data. We're not gonna use their data for other purposes, and so that helps. And they are already big Akamai customers. They know us and trust us.
How do you think about the competitive environment for what you're building? 'Cause I think it's... I mean, it's clearly primarily the hyperscalers who are trying to build edge presence. It's a long road to that, long road ahead for them, but you have other competitors with edge clouds that are talking about this environment. Just how do you think about who's really able to compete and have the uniqueness of the centralized compute that you now have embedded into the-
Yeah, for that, where we're headed, it's 100% the hyperscalers.
Yeah.
You know, obviously, Linode competed with DigitalOcean and, and other folks that focus on SMB. We, we compete there, but that's not our focus. There's other CDNs that have some flavors of compute, but they don't support containers and VMs, so they're not competitive.
Okay. From an investment standpoint, I think the messaging coming out of this last quarter was from a capacity build-out standpoint, you're at a good spot. It sounds like some of the regulatory aspects are coming next year. You've added 13 new core compute regions this year. The CapEx assumptions, I think, moderate from here. Is that the right perspective? And, are we in a position where heading into the calendar 2024, you're really an opportunity to go out and start winning some business?
Yeah, I think that you characterized it well. The initial tranche of CapEx, basically done. There's a little more for upgrading existing Linode regions. Some of those we're upgrading. There's a little more for, you know, porting the technology into some of our edge regions, as I mentioned, but that's much smaller than the initial build-out for the core data centers. From here, we need to fill that with revenue. We're filling it with Akamai business, you know, you know, and that takes a chunk of it, but it's, you know, filling it with revenue. And as we do that and fill that up, then, you know, we'll come back and buy more CapEx. But that's a problem I wanna have.
Yeah. One of the use cases for edge that's been highlighted over the past couple of quarters has been the opportunity to run some level of generative AI inference at the edge. And I think you've been fairly measured in how big of an opportunity. Some of your competitors have been a little less. But how do you think about the role that edge can play in machine learning inference use cases? And at a very high level, where do you expect AI /ML inference to be done, if you think about centralized cloud, distributed edge on device?
Yeah, we've been doing, you know, AI and ML on our edge platform for a long time. You know, you think of something like Bot Manager, and that is a big inference engine. It's deciding, is it a human, is it a bot? You know, for commerce companies, you know, what's the first page you present to a user before you know who they are? You know, we have multiple customers today in apparel, and you know, what we'll do is go get the local weather report. You know, if the weather report says it's 80 degrees outside, we'll present them a different page for the customer than if the weather report says it's snowing.
You know, 'cause if you present the page of parkas because it's snowing, you're gonna lose the sale on the first impression from the person, you know, who's in 80 degrees. You know, and so that's an example. That's all through edge computing. And we've, you know, been doing that for a long time, and it's... And there's inference engines, you know, for once you know something about the user, what's the next thing to show them? So that's something we've done for a long time. Now, I think where you're seeing a lot of the conversation now is around big LLMs, you know, and Gen AI, and GPUs. You know, you know, and that's a different kind of thing. To build a big model and initial training, yeah, today that would be GPU, and that's not at the edge. Shouldn't...
I, I don't think it makes any sense at the edge, 'cause you're collecting a vast amount of data. A one-time, more or less, big series of computations makes sense in a core data center. Once you have your engine and you're using it in an inference mode, there are situations like I described, okay, that makes more sense at the edge. Probably, that's probably CPU-based whereas an inference engine. Much more economical to do it that way. You know, with our compute, we do support GPUs. We are buying some. It's not the focus area for us, but that would be for the, the core. You know, for our edge regions, that'll be CPU-based. And initially, we're not focused on GenAI. I think... GenAI is very cool.
It'll generate the need for a lot of compute, but we're focused on, you know, big media, their workflow. Very economical for us, you know, to do that on, on CPU. So that's where we're, you know, our money is initially going.
Yeah, I think it makes sense. I mean, maybe just going back to the overall edge compute opportunity, and I think there's been questions about how that market evolves over time. Regardless of whether it's a significant impact to your business, do you buy this idea that generative AI applications are driving more customization in applications, and that inherently is gonna drive the need for more edge compute applications running closer to the edge?
Well, I think it depends on what you really mean by that. Inference engines, yeah, as I mentioned, been doing that on the edge for a long, long time. I think GenAI, you know, people are doing a lot of experimenting with it, and I think ultimately, it will drive a lot of need for compute. I also think, probably more and more over time, people—now that they see how important it is and what it can do, there'll be a lot of work on the algorithms and to make it much more efficient to do than it is today, and maybe a lot more use on, on CPU. You know, we'll see, going forward.
Maybe just zooming out and thinking about next year, 2024. How do you think about... I guess, on the delivery side, I think we're in a place where traffic's starting to stabilize in a pre-COVID era. Other than that, like, is there, are there, like, agenda items that you're thinking about from a go-to-market perspective that you think you need to change? I mean, with Compute, it seems a lot of that's gonna be very hands-on, but how are you thinking about any changes to that going into next year?
In delivery or go-to-market, or? You know, I'd say delivery, yeah, you're right. I think we're now seeing traffic growth levels more like, you know, pre-COVID, which is nice to see.
Mm-hmm.
As I mentioned, you know, we're being, you know, less willing to take spiky traffic and to give, you know, large discounts. That makes it more profitable for us and less spend on CapEx. We want to use that for compute, you know, instead. Internally, we've moved a lot of resources from the delivery teams into the compute teams. You know, go-to-market, I think, will look pretty similar, you know, next year to how it does this year, how we go about it. Security is, you know, very heavily channels-based. Delivery is heavily direct. Y ou know, compute is a blend. You know, the big media customers, generally, it's a direct conversation. We'll have channel partners, you know, to help with the lift and shift.
And we also have channel partners that do a lot of the compute work already for customers that we're becoming Akamai partners now to do that. So that'll be a mix, I think, going forward.
Yeah. I guess, maybe last question. How are you thinking about the M&A environment? You've obviously had a lot of success with Guardicore and excitement around Neosec, and Linode is a game-changing acquisition. How do you think about that as a capital allocation priority over the next year?
I think, you know, the philosophy is the same. We're always looking for acquisitions that can make a difference for us and our customers. You know, security is a very rapidly moving landscape, so, you know, we're always looking at, you know, the companies that come along there. A lot of startups are interesting. I think pricing is still pretty crazy for a lot of those companies and, you know, we're not gonna do anything crazy. You know, big acquisitions, we're especially careful with. We want to be really sure it's game-changing, like a Guardicore or a Linode, and really sure we're gonna make it be successful. So we're. You don't see that happen very often. We've been thrilled with Neosec. You know, that was much less expensive, but I think has a ton of potential.
Y ou know, for us, so we're making a lot of investment around it.
Yeah. Actually, one final question. I mean, the success you've had with Guardicore, it feels a lot like what you're trying to do at Neosec. Like, is there one or two lessons from what you've done with Guardicore that you're looking to kind of apply to Neosec to build out that product?
You know, we want to get a good cultural match with the team. In that kind of situation, we want the management team to want to stay.
Mm-hmm.
We're very happy with the channels' go-to-market operation with Guardicore, so actually, we're extending that for API security with Neosec. You know, we do want to integrate them in, into the Akamai family, but take the best of what they do and use that for more of Akamai going forward.
Yeah. Awesome. I will wrap it there. Tom, thank you very much for joining us. Thank you all for joining the session, and hope everyone has a good rest of the conference.
Great. Thank you.