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

Good afternoon. My name is Fahad Najam, I'm the senior analyst covering comm equipment here at MKM. And today, I have the privilege of having Mark Foss, Vice President of Products at Arista Networks. So Mark, thank you for joining us. I'll quickly turn it over to you for any forward-looking remarks that you may have to -- or disclosures you may have to disclose. And then we'll get started.

Yes. No, we can just get started.

All right. I'd like to remind investors, this is your time, so feel free to ask any questions via the chat session. I have plenty of questions prepared that will take us to the rest of the evening. But please ask questions, and I'll ask them on your behalf. So Mark, to get us started, it looks like we have the vaccine here, and we got the first dose in New York City. So maybe '21 is a normal year, God willing. And if that's the case, I want to take a time and reflect on what you think how COVID-19 has changed the world, especially for Arista.

Yes. Well, I mean, when COVID, I guess, happened here in the U.S., I guess, starting in March, really, we weren't really sure what was going to happen to be quite honest. And we weren't sure if people were just going to hold spending or whatever. But I think the move to online, I think, really helped our -- the overall spend in the cloud and in the service provider side of the business. The bandwidth increases for people doing remote working and things like that, didn't actually impact spending on those areas that I mentioned. So we actually saw some strength in those areas. The place that probably did get impacted, which was a new market for us is the enterprise campus market, and this is connecting to workspaces and desktops because people are not in the office. So many customers were like, well, I'm not going to upgrade the latest and greatest when I don't have any -- don't have anybody in the office. But that's a very new market for us. Like we only did $100 million in that market for the first year, which was middle of 2019 to middle of 2020. And now we're saying we can do $200 million over the next 6 quarters after that. So that area of the market, it's very small for us, fast-growing, but I think that part of the market was probably most impacted by COVID. But the rest of the market, actually, we saw an uptick because of COVID.

Interesting. And I think you've previously highlighted that you still see it and eventually $1 billion revenue plus opportunity from campus. Is that the view and...

Yes. I mean, if you look at the campus market, which, depending on the industry analyst you look at, it could be anywhere between a $10 billion or a $12 billion market. With Arista being only $100 million, that's less than 1% of the market right now. If you look at what we did in data center, when we started there in '08, we went from 0% market share in '08 to about 18% share today in data centers. So if we can, over the course of many years, if we can replicate our success in data centers as we did in campus and gain share there, the TAM is there. So we believe that this could be $1 billion market for us, but it's going to take a few years to get there.

As we look towards calendar '21, and it seems like if it is indeed going to be a somewhat of a normalized year, but we return to the office where we return kind of in a hybrid mode, some employees working remote, some working in the office, I think my feeling is that a lot of enterprises have not invested in the infrastructure. And there's a big gap. Do you think that -- or are you beginning to have those kind of discussions with your customers when they're not scrambling to adjust their enterprise IT budgets and say, well, this is -- infrastructure is a big spend area, we need to narrow this gap?

Yes. A little bit. In the U.S., at least in California, I know that like office is probably going to be -- remain closed until September. So if anything, it may return to normal maybe in Q4. I think the next 6 or 9 months are a little bit -- are a bit sketchy. Other countries, for example, like Australia or New Zealand, these countries are actually back to normal already because they've been able to contain the virus. So it really depends on the country you're talking to. But in the U.S., in general, I think there could be -- if these -- if they've left their networks, not really done too much to their campus networks over the last year, it may be a good time to upgrade before they bring a lot of people back, generally starting in September depending on the customer. But yes, it would -- could potentially be a good time to start looking at that, but time will tell.

Okay. Now in terms of -- moving a little bit of focus. Out your area of purview, which is enterprise looking into cloud, how do you think COVID-19 impacted your cloud customers?

Yes. For the Tier 1 cloud, I think we saw -- we probably saw a little bit of an uptick there. There were -- we had -- Microsoft is already at a 10% customer, Facebook was a 10% customer. I think we saw the most increase really in the Tier 2 cloud. So the Tier 2 cloud customers, they generally have a very lumpy buying pattern. They'll -- unlike the regular cloud titans. The cloud titans, they buy every quarter. It's pretty consistent that they buy every quarter. The Tier 2 clouds, they'll generally buy a lot in one quarter, then they go away for like a year, then they come back. So we found that with the kind of the stay-at-home orders, we -- it's got us talking of like, a lot of the Tier 2 cloud comes -- cloud companies come back and buy all at once. So we really saw the Tier 2 cloud vertical increase because of that. And it could be just -- they were -- it could just be timing because of COVID and the increased bandwidth requirements. But that would be the one place I would see where we saw strength.

And do you think that strength continues in '21 because now a lot of enterprises have essentially moved apps to those Tier 2 cloud-delivered applications, right? So maybe that demand remains steady and continues to be strong. Is that how you're thinking about it?

I think the world is continuing to move to the cloud. I don't know when we're going to be at that point, if ever, I guess, to where you're going to see a slowdown or people moving applications out of the cloud. But at this point, from what we could see is the world is still moving to the cloud. We probably won't see the abrupt moves that we saw when COVID, which was a catalyst, but I think we're still going to see a long-term trend of companies moving more applications to the cloud. This may make them actually more comfortable in the fact that they've moved more there, and now they're feeling more comfortable with how it's working. So I still think you're going to see that trend continue.

And then lastly, I mean, I guess, looking at the impact of COVID-19 on the supply. I noticed that on your last earnings call, you had almost doubling of your inventory levels. And I suspect the global supply chain is somewhat normalized, and hopefully would normalize. But any thoughts you can share with us on the outlook in the supply chain?

Yes. No, we -- a big chunk of components come from China. And we saw the first shutdowns of component suppliers in January, February when China was shutting down. Those component suppliers have since come back online. What we're seeing as well is even we're seeing capacity reductions, oftentimes in supplier factories where they'll have like a COVID breakout within the factory and after quarantine. So you'll see reduced capacity. We've actually seen a couple of bouts of this in our own factories from time to time. So you'll see reduction in capacity when you do see COVID hit these certain companies. And there's also the risk of, obviously, a government shutdown could limit the impact of factory output. But in most cases, I think the factories are considered essential. So I don't that think they've been hit with government shutdown orders. But I think you saw the increase in the inventory of Arista because we want to mitigate any kind of risk to component suppliers or any kind of factories having reduction in capacity. So for us, building up inventories and mitigation for us to be able to keep our customers happy and keep lead times as low as possible. But it's still uncertain. I mean the vaccine is here, but as we all know, it's going to take 6 to 9 months for it to become effective. And we're just mitigating risk.

All right. So it looks like things, hopefully, are better than 90 days ago, but you're still -- have a massive amount of inventory on your balance sheet that allows you to meet customer demands without...

Absolutely.

Money on the table?

Absolutely, yes.

All right. That's good. So let's dive a little bit deeper into enterprise and talk a little bit about the trends you're seeing. The first question I often get from clients is, why do you -- or why do I think that Arista should have meaningful success in campus networking? I have my views indiscernible a significant installed base of Cisco cameras, which is not all the customers are on board with Cisco's move to a subscription model being shoved down to them. And so talk to us a little bit about your design wins, wins with your customer, why do you win? And what do you see the runway ahead of you?

Yes. Well, I think it's a repeat of really our success in data center, to be quite honest. When we started data center from 0 in '08, people were saying, Cisco has 75% or 80% share in the data center space and why are they going to choose you? So we actually have a lot of data as to why we were successful, and I think we can replicate that. And I think it all comes down to the software. Arista designed our EOS. It was -- it's a 21st century operating system. If you look at the operating systems of the other incumbent vendors in data center and in campus, these were 1980s, 1990s technologies. So we're a true 21st century operating system, which means not only does it give you a lot of flexibility in automation, but it gives you very high quality. These older operating systems are based on 1980s, 1990s software architecture and technology. And they have bugs and these softwares -- the software can fail and it can bring down your entire network. The Arista EOS solves a lot of those issues. So our quality has been very high. And for us to start as a start-up back in '08, people were always saying, "Well, what's your quality like? You're a start-up. Your quality -- you guys are probably working on your things to your quality, right?" And we're like, "No. Actually, our quality is like unbelievable." So I think that was a really big thing that really helped us get traction there. And also our service and support, we're very customer focused. And we're very focused on customer first, servicing our customers and just being very diligent of that and also developing best-of-breed products as well. So our customers -- we have roughly 7,000 data center customers. They actually asked us in like 2016, and they said, "Hey, we love your software. We love your products. We love the server and support in the data center. Can we please have something that we can use in our campus? Because our campus, we're still living in fear of doing software upgrades. We have no automation. We have no visibility on what's going on. Can you guys help us out?" So that was the real catalyst for us to get into campus. And so we developed some products that were specific to campus and leveraged our EOS. And now we have a consistent operating system, which you can deploy in the campus and the data center. And we're seeing traction in that campus for the very same reasons that we saw in the data center in the first place. So so far, it's been good.

Now a lot of skeptical investors often ask why Arista doesn't have a large sales force of aftermarket compared to its larger peers, like Cisco, Juniper, HPE. So why would you be necessarily more successful? And I think part of the answer is you're already selling it to your embedded installed base. The large enterprises that were the customers for your data centers, but I'll let you answer that question.

Yes. Sure. Yes, no, if you look at the enterprise campus market, it's about, like I said, it's about $10 billion, $12 billion. If you look at the specifics of that market, about half of that market is large enterprise, which is Global 2000. And that is right in our sweet spot for -- we do -- so our model is direct touch to the end customer and then we use channel fulfillment for that part of the market. And of course, we have 7,000 customers already in that space. But really, the Global 2000 is kind of our sweet spot for that. Then if you look at the other half of the market, this is kind of the medium business and the SMB. The difference in the campus market and the data center market is data center is a very concentrated market. There's maybe tens of thousands of -- low tens of thousands of customers globally. The enterprise campus market, there's hundreds of thousands of customers. You can't possibly reach all these customers. So we're going to focus on the Global 2000 with our direct sales force. And then we're leveraging the channel for the mid-market and SMB. Now our first products are more optimized for the large enterprise because that's -- it is our main target. And you can use these products in the medium enterprise. But really, our go-to-market is pretty defined with, again, our direct sales force on the large and then leveraging the channel for the medium and small.

Now in terms of -- when you're selling solutions into these large enterprises, oftentimes, Cisco would maybe argue and that these customers want an end-to-end portfolio solution, WiFi, campus switching, closet switches, security, the whole gamut of product offerings combined as one single platform. How are you competing in that scenario? And how do you render?

We found that there's a lot of, I guess, legacy vendor fatigue in that space. If you look at the large enterprise segment specifically, Cisco has between 85% and 90% share of that market. I think HPE is #2. And then you've got Huawei, who's really based in China for the Chinese side. But we're seeing a lot of vendor fatigue, our incumbent vendor fatigue, and customers want a choice. And they haven't had what they view as a viable choice. So there -- many customers are kind of, thank God, we have someone else that we can actually look at and perhaps dual source or perhaps even consider moving away from the incumbent vendor in many times because like I said, there's -- these incumbent vendors have been running on older software for years and years and years. And some people have actually built their careers off of managing buggy software. But sometimes they see it as a breadth of fresh air when we come in, and our quality is great. We don't have all these issues they've had to spend a lot of money on firefighting for years. So it's actually been pretty good.

So just to kind of, not putting words in your mouth, but you're saying your enterprise customers tend to buy best-of-breed. So they will like best-of-breed campus search and best-of-breed WiFi, best-of-breed security solutions. So it's not necessarily one package solution from one vendor.

Right. Some like best of breed, but I think fundamentally, they just want quality software. I mean here's one reliable software that won't crash. I mean let's be quite honest. I mean, I don't -- I don't -- I like my Mac that I use here because it's very high quality. I had some PCs back -- way back when they would crash all the time. It infuriates me. I want something stable. So I think customers -- they just want to have a network that works. They can automate it and they can -- they have a turnkey solution that they can manage it with a minimal staff. It really -- they just want something that works.

Very interesting. All right, shifting focus within enterprise to WiFi. Edge as -- I think the industry has called it in the past, is a very promising new frontier for Arista. And I think in this post-COVID-19 pandemic world, I think WiFi is becoming a critical tool kit for the enterprise, whether it's enforcing contact tracing or enforcing social distancing moves. So can you talk to us a little bit about the trend you're seeing in the WiFi market? What are you seeing there in terms of demand trends? And then I want to also ask you a little bit about this new solution that you're offering called P-Tracer?

Sure. Sure, yes. I think the trend we're seeing, I mean, we're seeing kind of a move in how WiFi a number of years ago was all on-prem controller-based solutions. What you're seeing is you're seeing a move towards a shift towards cloud-based WiFi solutions. And that's what we have with -- when we acquired Mojo a couple of years ago. It's a cloud-based solution. And that's where the market is currently moving to is currently cloud-based. Within the WiFi itself, we're seeing a move towards WiFi 6, which is, I guess, the sixth generation 802.11ac, where we're seeing higher bandwidth at the edge to allow for IoT and all the different things that connect to the WiFi and the campus. I think the P-Tracer technology you're talking about is a feature we introduced not too long ago, and it is for contact tracing within the enterprise campus. So we can actually -- it can actually be used to track people's movements and find out exactly if people that have maybe -- that have been exposed to COVID are interacting with other people. So it's just a way using the WiFi technology to be able to understand contact tracing. So people can use that as a tool as they navigate through COVID.

So how is the engagement? Are you increasingly engaged by enterprises and your customers who are necessarily looking at this as primarily the biggest problem that they're trying to address? And is that -- how is the engagement for this new solution there for WiFi? How oftentimes do you sell with the WiFi first and then the -- pulled in for the other campus equipment?

Yes. It's hard to say. It's hard to say. We have a number of existing WiFi customers that Mojo had that we've been educating on our software and our campus switching lineup. There's sort of some of that happening. Then there's also -- but Mojo was a little more medium business focused, if you will. They relied in the channel, which is great because that's a market that we're cultivating as well. Arista was obviously more focused on the large enterprise. So for the large enterprise customers, we're actually bringing them into our WiFi solution. And in a lot of times, they haven't done cloud WiFi before and it's something that they're interested in looking at. So I think it's kind of a little bit of both on just educating both sides on the switches to the WiFi side.

All right. I have a question from an audience member. We're not asking you about 400-gig data center.

We left 400 gig. What's the question?

So let's switch topics to data center and talk to us a little bit about the trends you're seeing. When do we expect the 400-gig ramp to start? Where is Facebook and Microsoft in their upgrade cycle? Or is it -- you can talk -- give a little color on the cloud trends.

Yes. Well, if you look at the 400-gig market right now, pretty much, it's -- the uptick is dominated right now by Google and Amazon. So Google and Amazon have actually already started their 400G upgrade cycle, but they're -- they use their own homegrown equipment, which they've been doing ever since before Arista was a company. So that kind of dominates the 400-gig numbers right now because Google and Amazon have done that. The rest of the market, which, I guess, would be vendor-based 400 gig, it's very small right now. Like I think the entire market in Q1 was -- or sorry, in Q3, it was like less -- I mean, it's less than 20,000 ports globally, I think. It was very, very small. And there are companies from all verticals trying it and -- even enterprises are looking at 400 gig, may not be using the 400-gig interfaces, but you could use them at 100 gig as well. So they are buying the 400-gig ports. But I don't think you're going to see the significant ramp in the vendor side until some of the large cloud customers, like you're talking about, Microsoft or Facebook or other that don't build their own, decide to adopt that. And I think you'll probably start seeing that probably in like second half of 2021. But it's -- I have to warn people that it's not going to be the really massive upgrade cycle that we saw when we saw the shift from 40 gig to 100 gig because that was like a one-for-one replacement. You had the same optics form factor. It was just a lot cleaner of a transition. 400 gig, I think, is going to be a lot more targeted. I think data center interconnect is probably going to be the first place where these large cloud customers are going to upgrade. But I don't think you're going to see this massive upgrade cycle that you saw when it moved to 100 gig. It's going to be more targeted, and it's going to take a longer time okay.

Okay. So staying on 400 gig for a second, just kind of to help investors put into perspective. I'm a former network engineer myself.

Awesome.

Google and Amazon, and Facebook and Microsoft, I guess, Google, Microsoft and Facebook have something common in their architectures. They have 2 cores. One is a core that just connects their proprietary data centers. And the other core is the one that connects to the rest of the world. I think in the rest of the world for say Cisco, Juniper, and increasingly, Arista and Cisco's and Google's B2 network, right? So you've been winning share there.

Yes.

So you -- and that's routing, right? And that's a lot of routing use cases. Can you talk to us a little bit about where you're seeing routing in the core? And how is that -- is that 400 gig? Is that something you included in the 400-gig numbers or is that something extremes?

Yes. Yes, no, that's a very good point. If you look at how these big cloud customers, a lot of them, they -- if they use white box, they use white box. They want to have the -- it's -- I guess the closer you get to the server, the simpler the network is. And it's a lot more economical for you to make your own or homegrown stuff when it's simpler. So I would say that directly connecting the servers is where you see the most white boxes. But as you go up the network stack, you'll see more -- the network gets more complex. And then that's where you see more vendor representation. And specifically, the part you mentioned in the routing layer is a very -- routing is very, very complex. And what you'll see is you'll see, yes, Cisco and Juniper there, most often, and they're also looking at Arista in a lot of cases for this because it doesn't make sense for them to make white boxes in this area. And with these cloud providers having all -- very large data centers around the world and a lot of global spend, it is still a pretty significant spend there. So definitely, we're involved in those. I think it's primarily mostly 100 gig today. I think over time, I think this will move to 400 gig, but it will take some time. I think there's some optics that we're still waiting on to be able to go extended distances. But data center interconnect for routing is definitely a big piece for us. And the cloud providers, they like us because we were used in the spine layer already, so they know us. And just using those same products for the routing layer and just turning on a few more features isn't that big of a change for them. So it's actually been pretty good.

Right. So just because that's an incremental opportunity for Arista above and beyond your core DCI lease by -- your leaf spine intra-data center in the proprietary DCI network that they have.

Yes. Routing is definitely an incremental opportunity for us. Now you have to be careful because there's this large legacy routing TAM that everyone loves to look at, which is whatever, it's like $10 billion or something like that. But that's based on the legacy routing pricing. And I guess one of the appealing things to go to a -- to do a switch for routing is that the cost structure is a lot less per port. So that TAM doesn't necessarily apply to switching because the TAM -- the price per port is so much lower in switching, which makes us a lot more appealing. Now for us to be able to measure the route, and it's very difficult because the same products that we use for our spine layer could also be used for routing. So it's very hard for us to be able to accurately measure how much we're doing in routing. We look at it in the number of customers that we're doing that buy the advanced routing licenses sometimes. But for us to be able to make and put a dollar value on that is very difficult because it's very intertwined with the spine layer, if you will.

So can you speak to the momentum you're exhausting in the routing use cases? How can we, the investment community, follow your success there?

Yes. Like I said, I think you could look at it in terms of the number of customers that we have. So I think in our earnings report, oftentimes, we'll kind of mention how many routing customers we have, that would be one thing. I think some industry analysts that I've been talking to have been talking about breaking out kind of a switch routing segment because right now, they have the traditional switching, traditional routing, and they know that there's a blurring between the switching and routing. So I think you may start seeing some analyst reports, maybe starting as early as next year, where they're going to try to attempt to measure the routing portion of switching and create a new segment. So that's another aspect as well.

All right. And then lastly, on this cloud topic, competitive dynamics. I mean Juniper has talked a lot about entering into the 400 gig. Our own checks are saying that Cisco's Silicon One strategy, they're selling their Q1 into Facebook and Microsoft, and those trials are going very well. How do you think about the competitive dynamic?

Yes. I think if you look at these cloud customers, they trial everything. Like if something is new, they will take a look at it absolutely. So I don't doubt that everything is trialing. But again, it all comes down to the software. Companies -- these cloud companies don't buy based on hardware. They based -- they'll buy based on the software. And if you don't have a robust software stack that they have flexibility to program on and has very high quality, it's very difficult to insert yourself in there just based on a hardware play. I mean the hardware -- a lot of these -- a lot of these vendors are going to be using the exact same silicon that we are. So we don't differentiate ourselves in the hardware. It's really -- it really comes down to the software. And that's really what it comes down to. And then also having joint engineering, we actually partner with the cloud providers jointly, engineer to engineer, to actually develop next-generation solutions together. And when you have that level of partnership, it gets pretty sticky, and it's very difficult for others to come in unless they also develop a partnership with that kind.

All right. Well, I have a lot more questions to ask, but we are, unfortunately, out of time.

All right. Thank you. It was fun.

Thanks, everyone. Thank you, Mark, for joining us, and wishing you all a good success. And hopefully, this vaccine does as its promise and we are back in to a normal '21.

I hope so too. Thank you very much.

All right. Take care. Thank you.

Take care.