Arista Networks, Inc. (ANET) Earnings Call Transcript & Summary

Good afternoon, everybody, and maybe good evening if -- for those of you who are still following us from Europe. And if you do, thank you for that. This must have been a long day for you. So I am Pierre Ferragu. I lead the technology research team at New Street. And I'm very, very happy to host for this session, Arista, Arista Networks. And we have here on the call in -- on the stage -- on the virtual stage, Ashwin Kohli, who is Senior Vice President in charge of customer engineering; and Ken Kiser, Group Vice Present of -- in sales, who has like a -- as part of his responsibilities, the sales responsibilities for service -- service provider clients for Arista. So I will have like a fireside chat -- a virtual fireside chat with Ashwin and Ken. I have a lot of questions for them. I'm sure many of you in the audience would have questions. If you have them, just type them in into your conference window or actually shoot me an e-mail as well, if you want, it works well for me, and I'll do my best to integrate your questions in the floor as well. Well, Ashwin, again, thanks a lot for making the time to join us.

And as a starting point, let me ask you a fairly broad question: What is Arista's technology? And what is its role in 5G?

Sure. Perfect. Great. And thank you, Pierre, and everyone. I really appreciate you guys joining our fireside chat today. I'm actually going to punch it over to Ken. Ken is closer to the customers as well as I mentioned before, and then I'm going to have him answer that question, and I'm going to give you a little bit of my background as well on that answer. That works for you? Yes?

Yes.

Ken, over to you.

Sure. So Arista Networks is a technology company based in Silicon Valley. We make networking devices, commonly referred to as switches and routers. Initially, the company got started making very specialized switch and equipment targeted at the low-latency financial market. So the distinguishing characteristics of those products were very, very high levels of reliability, given the criticality of those services, and most importantly, the latency, right? Obviously, moving the information between the trading servers provided a big advantage when you do algorithmic trading. So that was kind of the entry to market. It got us going. It basically generated a very rapid profitability for the company. But the aspirations were always around solving the problem for the cloud. I mean Arista was founded in 2004 with the premise that the cloud, back then, it wasn't called the cloud, but it was early stages of Google and social networking, that the cloud was going to have a demand for networking products that was far different than anything that had been done before in terms of the performance of those products, the way you manage those products, the way you evolve those products in a rapid nature to drive Moore's law type of economics. So as it's pretty well-known now, Arista was very successful with that journey and became a very large and continues to be a very large supplier of networking equipment to the largest cloud operators on the planet. Since then, right, we've been on our journey to diversify our business. Certainly, we don't have any desire to not continue to do business with those very large customers. As you know, 2 of those hyperscalers are top 10 customers for us. But we've done a lot of work over the last few years to expand our offering outside of the cloud. Now we feel very proud, and we've gotten a lot of great advantages by working with these companies, the innovation, the knowledge, the direction they've helped us to drive our products. So a lot of the things that we built in conjunction with a Facebook or a Microsoft, we've been able to take those learnings and apply them to products that we have been successful selling to general enterprises, big banks, big hospitals, state and local government, is one, as you probably heard recently about our progress in the campus business. But for this discussion, really, it's about the third leg of that stool, which is the service providers. So we've been on a journey. And I've been kind of the steward of the strategy from the very beginning of my time at Arista in 2012. And initially, it was basically just a direct, take the same sort of data center switching products that we built for the high-frequency traders and the cloud operators and basically take them to solve similar problems in service provider. Every service provider is a business as well. They have IT infrastructure, they have data center infrastructure. And we were very successful at basically just selling the same products to service providers, it's just another customer type. But all along, the goal has been to basically continuing to evolve our product to become more of what the industry calls a full-blown router. Jayshree, and we, joke a lot, that our goal is to take router as a noun to routing as a verb because we believe routing is a feature set, it's not a class of product. But it was always the intention to take the journey, both with our hardware partners, companies like Broadcom and Intel that build our silicon, as well as the software set continually down that very long and difficult journey to add the entire feature stack to make it applicable to take an Arista product and stick it in the middle of the largest, most complicated, high-end service provider networks there are. And we are on that journey. I mean we -- we're probably further along than we have to go. There's still a lot of work to do. But that journey is what has allowed us to be relevant in this 5G trend. So first part is, we've made the investment, we've built the relationships with the customers, we've got the foundational wins in the data center and other parts of the network to allow us to compete for 5G. So that's an Arista-driven thing. The other side of it, which your question is, is what's going on that would allow Arista to actually participate in that. And what's going on there is if you understand how 4G networks were built. They were very classic telco, right, it was one vendor, it was typically proprietary that each market, each 4G market, like Verizon went into New York, right? They picked one vendor, Nokia, whoever, Samsung, Ericsson. And when Vodafone went into London, right, they would pick one vendor to build out the radio access network, the antennas and the net backhaul, that would be a single-vendor thing. And because of the way the technology was developed, you had to do it that way. And then the way you got diversity is you did Cisco in London and then you do Juniper or Ericsson in Frankfurt and then so on and so on and so on. But what's happened with 5G is those interfaces between all the various components of the radios and the core and the backhaul are now standard, right? They're using standard Ethernet technology, standard IP technology. So no longer do you have to buy it from one place. It doesn't mean you won't. I mean there's a lot of business reasons that you might want to do it. But that change in architecture has allowed companies like Arista, who have a long history and a lot of intellectual property around Ethernet and IP, to be able to compete in this space. And the service providers are under a lot of pressure, right? They haven't had great growth. They certainly have a very capital-intensive -- capital-intensive business that's got a lot of competitive dynamics coming out of them. And they've been forced and are becoming much more open-minded about searching out new partners, new vendors to help them drive innovation as this 5G turning takes place. So to summarize, right, it's a 2-part answer: One is Arista's been at this thing for long time, driving towards building the right features, building the right product, building the right credibility, building the right mass to make us credible in this space; at the same time, the technology has evolved from 4G to 5G from kind of narrowband to broadband that will allow new vendors that are more standards-based to insert into the telecom infrastructure.

And then, Pierre, if I can add to what Ken has said, to answer your question very specifically in 5G in terms of technology, if you break it down into very -- 3 high-level buckets, right, so you've got the radio access networks, which is right at the edge, then those radios are connected to the aggregation points, which are then connected to the core. So if you actually look at -- if you break those 3 things down, starting at the core, and this is what Ken was talking about, traditionally, a lot of the service providers needed, what Ken and I joke very often is what's called a god-like box, right, which is a routing box. It has to have a whole bunch of legacy features in order for us to go sell into that space. Well, in the last 4, 5 years, Arista has put a lot of R&D investment in our software in order for us to go get into that space in the core side, right? The middle piece is where 5G really comes into place, where this is all the aggregation points where the radio is connected. And if you actually look at that space specifically and you just peel the onion out, a lot of that looks like data centers, right? And so the reason why it's prevalent is because for 5G, when you look at the application types, you've got couple of different types, right, you've got the infrastructure that's necessary to make the call and then there are applications that are necessary for 5G, like IoT or smart cities or smart driving, those kinds of applications that need to be closer to the edge. When you're building that infrastructure in the middle, it looks like a data center, the reason why service providers are looking at us is because they want to be more like cloud-like. They want to have the same principles, low power, low footprint, high bandwidth, more automation, open APIs to go build a data center to give them the flexibility for that. And so because Arista's been really good at building out cloud infrastructure, they're now naturally having conversations with Ken and my team in the field in order to go build that middle piece, right, which is -- gives us a really good edge to break into the 5G space. I wanted to make sure I answered your question there.

Yes. I think you very clearly answered the question and actually stepped over the next one. I wanted to hear your thoughts about Telco cloud. So this idea that the telecom network is becoming like a distributed cloud. So this is exactly where you are positioned. So can you tell us about what you're learning from that? What's different between a cloud and a telecom network today? And what does it take to transpose the way you do networking in the cloud into the way you do networking for telecom? And how is that journey looking like?

Let me take it.

Go ahead. Yes.

So it's interesting, right? And if you're watching the industry, you see, there's a lot of announcements between some of the cloud operators and the telecom operators, AT&T and Microsoft, AT&T and IBM have made announcements, Verizon and Amazon have made announcements. And a couple of things, right? So clearly, the business is all driven by the applications. But one of the things that's changing in 5G is that you got to understand how the networks are architected. In the old way, 4G world, you had a radio, right, you had a backhaul, right, all closed. And then some far pretty deep in the network, basically, the data and the voice would get unpacked in 4G, and then that would be where the IP connectivity was. And then that IP, you would then peer, right? They would connect to the Internet and their backbone and whatnot. Like I said, it was pretty deep in the network. What happens with 5G is that native IP interface now it's much, much, much closer to the radio, much, much, much closer to the antenna. So where before the operator had to own everything through the RAN because it was closed, now you can insert basically third-party data centers and applications, much, much closer to the end. In some cases, the actual tower itself. So what that allows for is it allows for a much more robust application infrastructure. So if you are in the business of hosting applications, like the cloud guys are, getting that cloud data center, that Microsoft data center or that Amazon data center connected closer to the tower than in past is the objective, right? There's good value there. And of course, the operator, service buyers want to capitalize on that asset, and they're looking for ways to do deals and make those connections. So that is part of it. The other thing is there's this idea that by putting these applications right at the tower, there's going to be this magical revenue stream that's just going to appear because latency is everything, right? And this is a big debate. And I've heard this come up in calls today and elsewhere. And there's this idea that there's this beachfront property at the bottom of the tower that you can then deploy compute and storage and deploy these, like I said, these magical applications that no one really understands yet that would create this new revenue stream. So that may happen, right? We always have seen in our industry. When you create infrastructure or create the opportunity to innovate, people rush into that space and business models get created. But the idea that latency is this magic thing and that lower -- lesser latency is always better. You got to understand, right? An application's got a lot of parts to it, right? There's the handset, there's the database. There's all these things that have nothing to do with the network that drives the performance of that app. So what's going on and what should be interesting to watch and what I would encourage Pierre, you, and the people that are listening is, where is the right place, right? Does it really make sense to try to build out a massive-scale data center underneath every cell tower? Or is there some middle ground, right? So we know that one data center in the middle of a continent is also not the best answer, right? We've seen that learnt -- the whole concept of content delivery network has shown that some amount of distributed nature of application infrastructure is a really good thing. So the debate is, is where is that. And like 4G, there were some hard technology, architectural things that determined that the connectivity to the IP and the applications was pretty deep in the network. 5G gives you the option of putting it closer, but what will be interesting to see as this moves along, where does it go, right? And then to answer your question, where is Arista's play in it? We are ambivalent, right? I mean we are a data center company. So whether those servers are connected to our devices that are underneath the tower, whether those servers are connected to our switches somewhere in the middle, whether those services are connected to our huge sets of switches inside of these largest data centers ever built, we -- our products and technology fit in each of those markets. So it's really what the business driver and let's -- what the business driver is, what the application driver is that will determine where the best place to put those things are. And I think, in my opinion, it's like anything, there'll be a little bit everywhere, right? There'll be some magic or some appropriate trade-off where the majority of it will go, but we will absolutely see some compute and storage at the farthest end of the network because there'll be -- I'm sure there'll be some application that could do that. But the economics, I think, you'll find niches that will cause the stuff to be everywhere, and then you'll find something that kind of fits in the middle, where the vast majority we deploy. So that -- go ahead, Ashwin.

Yes. And Ken, I just wanted to add just a small thought to that. Right, Pierre, just to summarize what Ken said, if you think about this, right, the edge, which is where the radios are, the middle and you've got the large core, think of them almost as 3 types of Telco clouds or data centers, right? So the core is a very large data center. The middle is a medium-sized data center, and then the edge is a very, very small, maybe 1 or 2 switches over here. So Ken is absolutely right. The value for Arista is to be in all 3 spaces. But I want to stress a point over here. The additional value is today, in a 4G environment, if you look at any other vendors today, they sell different type of products to answer 3 use cases. Different hardware, different software, different ways to go monetize them, different ways to go operate them, to go get telemetry out of them. The value for Arista is, it's exactly the same merchant silicon that we can go provide in the large, medium and the small data centers. It's the same operating system. It's the same way for the same service provider to go provision, monitor, configure, troubleshoot, provide visibility. So then if you think about this as a business person, what's the value? It's a massive total cost of ownership reduction, and it's an OpEx reduction. So the reason we get that is because it's the same box provided in multiple use cases. And then the service providers can say, "Oh, wait a second. If I have to now go automate this," and they have to go make it programmable, very much like the cloud providers, that's the value that Arista is bringing in each one of these use cases. Same exact way across all the 3 types of data centers or what we call Telco cloud in the SP space. Hopefully, that makes sense.

Yes, it does. It completely does. And actually, it makes a lot of sense, and I understand well this idea that 5G is expanding the IP domain is a very, very bottom -- is very like...

The tower itself, yes.