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

All right. Good afternoon, everyone. I'm Rod Hall. Welcome to the Goldman Sachs Technology Conference. Really appreciate you coming along to this. I've got the pleasure of hosting Arista Networks again this year. We've got John McCool, the Chief Platform Officer, also SVP of Engineering and Ops. So John, welcome. Really good to have you.

Good to be here. Thank you very much.

And as is always the case, Arista is good enough to attend this conference, but they're in their quiet period because they haven't reported earnings yet. So we always restrict this conversation strictly -- just the technology, what's going on with the business, that way, try to stay away from financials and numbers and so on, given it's in the quiet period. So hopefully, everybody out there understands that, and we can still learn some really interesting things from John, who's been in the industry a long time. So just a reminder on that. There will be a chance to ask questions as well. I'll go through some Q&A with John, and then we'll have a chance for people out in the audience to ask questions. There's a system you can put those in to, and I'll pass the questions in onto John. So with that, let me get going. I don't know, John, if you want to kick off with a couple of comments about what's going on with the platforms and systems work that you've been doing, maybe just a little quick intro for everybody, especially people that maybe aren't as familiar with Arista, and then I'll jump into some questions.

Sure. Just a quick update on Arista. I think we've started as a cloud networking company, I think, before cloud networking was even a concept and, actually, the cloud market was relatively small. And have grown rapidly through the rise of the cloud-native companies, but have expanded into enterprise and service providers over the last 10 years. And they're adopting more of this cloud-based approach. And with that, we've expanded our portfolio into routing, into campus networking, WiFi and with our recent acquisition of Awake into network security.

Great. Okay. That's good. Well, let's start off talking about cloud campus networking, which is always a hot topic with Arista. It's even more interesting this year with lockdowns, and the question as to whether we need campuses or not. But I think -- I suspect we do. It seems like campus sales for people like Cisco and so on are still going pretty well. You launched your new 750 Series campus chassis back in November. Andy's briefly talked about the product's throughput and performance advantages, et cetera. Could you just talk a little bit about what drives differentiation, in your mind, for Arista products versus other campus switching products that are out there?

Sure. Maybe I'll just start with the 750. We entered into the campus space largely looking at how we differentiate with some of the trends we see there. The modular campus space, I think, we believe, has been largely neglected. This is -- if you think about large modular chassis for enterprise access, these go into large organizations that have a lot of people or equipment that's networked. We've seen industry trends that have taken kind of commercial or stacking products and tried to leverage them into modular architectures. So we still have an opportunity to really build from the ground up a new modular access architecture that's purpose built, advantages around density and also bandwidth. We have 150 gigabits per slot, which seems outrageous in some ways when you think about a typical desktop user. But people are connecting new types of devices, WiFi 6 IoT devices and moving from 1 gig to 2.5 gig. So that performance becomes very important. The one thing that differentiates us from a whole system architecture, though, is that our operating system and binary that we released for our routing products or for our data center products or our campus products is the same. So you can qualify an Arista release and deploy it in all these various use cases, and our management system, CloudVision, runs across all of these different products in the portfolio. So if you use Arista in one segment of your enterprise network, you're ready to go and to deploy in another area with relative ease. And that's highly differentiated from the incumbent who has used more bespoke architectures for these different roles with different protocols, different ASICs and different management stacks.

Great. Okay. Can you talk a little bit about go-to-marketing campus? And I'm going to circle back at the end of this and ask you to talk a little bit about COVID and how your philosophy on campus is changing, what you're hearing back from companies as well. But let's go talk about go-to-market first because that's been a big point of, I think, investor focus. After you announced campus, there's an assumption it was going to be a really expensive go-to-market. That hasn't really proven to be the case yet, but you still could end up needing to expand the channel partners and so on. So just kind of walk us through like what the go-to-market is today, what it needs to be as you continue to grow the campus business?

Yes. We don't consider campus a distinct market segment. We're not going for mid-market. There's a lot of business in commercial. We've stayed very true to our enterprise focus of kind of Fortune 2000 companies. So we've expanded the channel presence around that segment, that customer segment, and not distinct from selling campus, but also selling data center to selling routing and the new pieces that we brought into the portfolio with Big Switch and Awake. So think of our investment in the channel really about driving the portfolio in the enterprise. We now have over 7,000 customers, which has been -- most of them are enterprise customers, and that increase has been with the channel. And we're going to continue to drive that in years ahead. So we think we're doing well. There's definitely room for expansion as we grow.

So how many -- could you talk about roughly how many channel partners are out there that matter for these Fortune 2000 companies? And are you -- have you expanded the number of channel partners since you launched campus? Or are you pretty much just using the same channel partners you had before and having them sell a little different product?

Yes. We definitely have expanded the number of partners. I don't have a good number for you. But again, we're not looking for a campus partner. We're looking for someone that can be aligned to the top customers, these Fortune 2000 customers that are part of our strategy and can adopt the Arista portfolio.

Okay. And is it -- just back to this partner thing. I don't know if I've ever asked you guys this before, but is it -- could you give us some kind of ballpark on how much the number of partners has changed? So before pre campus and after campus are now -- are we talking about double the partners? Are we talking about a 10% increase? What -- is it a smallish increase? Is it a great, big increase in the number of partners you're dealing with?

I don't have a good number for that, Rod.

No, no. Okay, okay. Let's go on and talk about your learnings in campus. You've been doing it for a little while. You've seen some growth. Talk to me about what you've learned in terms of the product resonates, what maybe -- what kind of features were you missing or product have you been missing that maybe you need to have? Kind of what have you learned as you've gone out there and deployed now with some customers?

We started a few years ago with a pretty limited portfolio. 24- and 48-port switches was our first introduction as well as a distribution switch. We needed to fill that out, and we took a different approach. Instead of kind of building in stacking, which we have been told by our customers is inherently not very reliable and very different from our cloud approach, we've built bigger building blocks. So if you have more users, you build -- buy a 96-port switch that we came out with in Q4. Now with the 750, we can address all the densities that might be needed by a customer. So we feel like we're in a good place there from a hardware SKU product perspective. Obviously, we have more on the road map, and we'll be filling that out. And the key from a customer requirement is, can you deal from a software perspective with data, voice, video deployments? And we believe we've achieved that with the feature set that we have. Now we're moving into really software differentiation and, obviously, kind of tying in the campus with the management of data center and routing in and of itself is differentiating. But we've taken some of the concepts we had in data center around tying in with partners and security and made them applicable to the campus. So we came out with a technology we call macro segmentation in the data center which connected into firewalls. So if you're using Palo Alto, you could use it to deploy security policy to prevent East-West traffic and leverage the network to prevent East-West traffic based on your security and firewall policy. We've taken that concept now into the campus with a macro segmentation approach based upon groups. So a customer would like to identify maybe the finance group as a separate group that can only communicate together, or PoE cameras are in their own segment and can't communicate with anything except the PoE camera controller. And we've done some integration with Forescout, where they can set policies based on these groups and then enforce them through the switches. So that's new on top of what we've done to further differentiate the campus solution.

Okay. All right. That's great. I also wanted to ask you about wireless LAN. How often a campus wireless -- campus switching sale includes wireless LAN and your cognitive WiFi solutions? How often -- and it includes wireless LAN that you didn't sell somebody else's? And how often it's just a straight ahead switching sale?

It's a mixed bag. We've had opportunities come at us where the customer was looking to refresh the whole campus, and we've won some of those opportunities end-to-end on campus switching and wireless. Oftentimes, they're looking kind of at the road map, but they're only going to upgrade the WiFi or the switch. And that's why it's been important for us to offer WiFi, because people want to see that you have that approach. And we want them as well, and come back in and got some switching or WiFi later. So it's really kind of a mixed bag. We're obviously coming in with the ability to mix and match, given that we're not the incumbent and not trying to tie these together in some artificial way through protocols to make things work.

Right. Okay. Well, let's talk about COVID. There's all kinds of interesting -- your first reaction to COVID is well, there ought to be less campus switching. And then you start really thinking about it, and you realize, well, wait a minute. I've got to do Zoom from my desk. I need a lot more capacity. And then we've had people point out, maybe people don't want to share desks anymore either with all this stuff. So I'm curious what your take on all this is. Like what do you see changing out there in the way people are thinking about campus networking as a result of all this?

Yes. I think it's still difficult to peg the work-from-home long-term trends, and people are talking about flexibility. And sometimes, you come in, and sometimes you don't. I think we're seeing that as kind of a dominant theme of, yes, there will be an office, but there'll be more flexibility and maybe less office space. So you could argue that diminishes the number of ports. And -- but we're just beginning. So from a business perspective, I think we still have a lot to be gained independent of what the overall TAM is for campus switching coming in against a strong incumbent. The other piece, though, that I don't think is represented as much is IoT and other devices. We have a substantial investment in cameras and access devices and other things that hang on the network: TVs, monitors, et cetera. So that still will be a very relevant piece independent of the number of employees that are coming in. Our view is employees will be connecting through wireless and devices and fixed assets, more connecting through the wired part, which will continue to exist if employees are there or not in the office.

Yes. I'm not. I mean, I think we -- I don't know about you, but I like the flexibility. I sure like to see some people more often.

Exactly. But the novelty has worn off.

It's worn off. Yes, for sure. I think it's worn off. That's a good way to put it. The other thing I wanted to ask you is, we thought about you getting into the data center first enterprise, and then the enterprise saying, "Hey, could you do campus for us because we really like what you're doing in the data center." I'm wondering, have you ever seen a reverse situation where now that you get further into it, where people look at what you're doing in campus and want to look at you for the data center?

Yes. We've had opportunities. Our kind of motto is land and expand. So we're trying to insert Arista into these accounts in whatever way we can. Obviously, data center is the tradition and the strength and where we can lead more often, but we've had routing opportunities and campus opportunities come up. Campus has definitely expanded the interest. I'd also say the bigger breadth of the portfolio makes us a more interesting supplier. We can be looked as an alternate for enterprise networking, where in the past, we were an alternate just in data center.

Do you feel like you've got all the technologies you need for campus now? Is there anything -- I mean, I...

The fundamentals, yes. I think there's more rounding out that we need to do. But it's starting to fall into just the traditional things you would do once you're in any segment.

Okay. Let's talk about routing a little bit. I want to make sure we touch on that. So you've expanded into both WAN and some edge routing use cases. You guys have talked about routing traction in telco cloud use cases. But I wonder if you could talk about the broader traction you're seeing in routing, what that looks like over the next 12, 18, 24 months?

Yes. Sure. I think it's good to kind of break it down by the segments. So if we look at cloud networking and cloud providers, right, that was really the impetus for us to enter into routing and taking a very IPP approach to that with BGP. And what we did was enable our customers to take multiple data centers and aggregate them so they functioned as a single data center. Traditionally, 2 routers were used to interconnect. With this lower-cost approach with merchant silicon, you could build out a broader routing layer across your network and have a higher number of interconnects. So that's going really well. Data center, we move from connecting your servers to being able to interconnect data centers and interconnect an enterprise WAN. And I think we're doing well there as well. They're smaller portions of the overall routing market. The dominant market for routers are service providers. And there, we've had a stronghold in the telco cloud, but are pushing into other use cases in the service provider area.

Okay. You've mentioned in the past how the service providers, especially the Tier 1s, would want the full set of routing features, more the full routing table or a much more extensive routing table than a cloud provider would want. So can you talk a little bit about -- is that still the case? Are they -- are the Tier 1s starting to think more like the cloud providers and architectural design? And then also maybe talk a little bit about Tier 2 and Tier 3 interest.

Yes. So I think when we -- we got to a point probably around 2017, '18, where we were able in merchant silicon, with more of a switching paste architecture, to put full routing tables in. And we actually got some good traction with a couple of key designs early on, but unfortunately, weren't as replicable as we had liked. We realized looking back now, they were a bit more greenfield than the traditional opportunity. We have since built a lot of functionality, putting things in like RSDT, to allow us to connect to a service provider legacy core. And last year, I think we've achieved kind of a minimal viable set for the use cases that we're going after. An evidence of that is we've gotten some good wins in a couple of new logos that we hadn't had before. The traditional use case we have with telco cloud, I think we believe we've actually strengthened that position. I think there was a lot of concern in the telco cloud. As they move to 5G, they would also transition to white box and commercial operating systems. That threat has somewhat diminished. Without a viable cloud OS, we're seeing more of a trend back to traditional OEMs like Arista. And then the other use cases, it's a combination of us getting more functionality, but also we're seeing the service provider revenue growth is on more IP-based services. So if you think about COVID, that accelerates it, right? The concept of site-to-site VPNs and us connecting into headquarters through the VPN and then out to the Internet becomes less relevant, right? I'm connected today through kind of split tunnels. If I need to access something in our corporate data center, I go through a VPN. If not, I'm out to the cloud. And the cloud trend is exacerbating that. So we believe the puck is still moving in our direction, and I think we're feeling like COVID might accelerate some of that transition.

Okay. Since you've been around in the industry a long time, I -- one of the things that I've been thinking about a lot lately is the hyperscale, the cloud players and the networks they've built, the backbone networks and the cost of provision of those networks, which just got to be probably world-class in terms of price because they're highly automated and super high-scale networks. And then you compare that to the carriers and the backbone networks, they're running the core parts of their networks. And you wonder if maybe the direction we're headed here is that carriers eventually just end up leasing capacity on top of -- now maybe it would be complex the way they would do that, but that capacity on top of these hyperscale networks, why should we have duplicated capacity, especially when we have these super high-scale automated things? I'm just curious what you think about that. Do you think that your ultimate customer in that core part of the network ends up being more and more and more of these cloud guys? Or do you find much interest or resonance with that kind of an idea with some of the carriers out there?

I think it's an interesting concept. I think it would be tough for a service provider to get their head around that because you're effectively -- it's a make-buy decision with your competitor. And that would be tough. That would be tough. So I don't think in the short term, we'll see that -- anything like that happen.

It's interesting to me because it's this concept of who is your real competitor when you're a carrier, and the carrier expertise is accessed. It's connecting that end user of the business in the network.

Yes. Yes. Yes.

But then how much is it -- how much does carriers really make on the core network, it's more of a cost center than it is a revenue generator. There's some revenue generated, of course, but -- I don't know. I -- we keep thinking about it, but it seems like it's maybe a little -- an idea a little ahead of its time. What about access routing for Arista? Could you talk a little bit about that? Especially we'll see more and more access routing deployment with 5G. Do you have an opportunity to participate there, do you think? How interesting a market is that to you?

It's -- I mean, it's a substantial market we're not participating today. I think when you look at -- look back and look at markets that we have entered like campus, we got to believe that we can have substantial differentiation that would appeal to our existing customer base and an opportunity to expand spend with our current customers and also conviction to a multiyear road map to be successful. So not announcing anything today, but they're sort of the elements that we see. And if you look at things like we did with WiFi and Mojo, we saw a transition to cloud-based WiFi, and kind of bypassed the head-to-head competition with the incumbency and try to lead in a new market transition. That's kind of what we think about as we enter our new space.

Yes. I just think -- what would you say some of the negatives of that market might be? Like what are some of the things that would maybe not be as attractive about it?

I mean, you enter a new market like that, you really have to think about how you differentiate versus the incumbents, right? It's a competitive market with a lot of incumbency and a lot of folks that have applied a lot of experience to that, so.

Yes. It seems like it's an esoteric market in that way. But then on the other hand, it's with -- particularly with 5G and some of the architectural shifts that are occurring. It's also a really interesting market from an automation and...

Well, that's where...

Some of the things you guys are good at.