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

All right. I'm going to go ahead and get it started just to keep the trains running somewhat on time. I think, we're little bit late here. But my name is Rod Hall. I'm the infrastructure technology analyst at Goldman Sachs. I've got with me John McCool, who's the Chief Platform Officer and SVP of Engineering Operations at Arista and a storied networking person in general. So welcome, John. Thanks for coming.

Thanks, Rod.

So before we get started, I'll just say that Arista's reporting on Thursday, so we're going to avoid any financial questions but all things technology-oriented, and I guess more broad, our open game for John as we get going.

With that being said, I just thought I'd kick off and say how the quarter is going, how the earnings are looking. No, I'm just kidding.

Stay tuned.

Yes. So I thought maybe we could kick off with -- because you know the campus networking side of the business so well. And that's a new growth area for Arista. A lot of investors may not fully understand why Arista would bother with this because the company has built itself on a hyperscale world, large-scale data centers. Can you just say a word about why this makes sense for Arista and help people understand kind of how it fits into that networking business that you've already built?

Sure. Absolutely. So if you think about Arista, we certainly made our name with the hyperscale and cloud networking in general. Even before that, we started in the financial vertical segment with low-latency switching and built out a really strong footprint inside the enterprise for, I would say, people who really care about their network to monetize something, financial services, media and entertainment. And they still have large data centers, but they also have a piece of their network that's not going away. It's not going to ever move to the cloud, it's the on-premise piece that connects the users into their applications. Same buying center. So typically, there's a VP of Networking in a Fortune 2000 company that runs both the campus and the data center side. So it was very obvious technology segment for us to go after. A lot of us have worked in that segment before. A lot of the base technology is the same for a data center, for a router or a campus switch. So I think it was a pretty obvious next step for us.

Can you say a word or two about the person that makes a buying decision in these places? Because I think that's a misconception as well. There tend to be 2 teams, 1 running campus, 1 running data center. But why you see it as a same person in many cases?

That's a good point. I mean from an operational perspective, sometimes the campus folks may be a little bit more oriented towards the desktop help desk arena, and the people that work on the network in the data center around kind of that server application focus and work with that team day to day. But people are looking at their network end to end. And typically, those 2 towers report into someone who is responsible for the total networking spend in an enterprise. That's very common. And as you go down into a smaller company, it could be 5 people that run the network end to end.

Okay. And just to finish up on the distribution side of things before we get into technology a little bit. Could you say a word about the channel? Like how important is the channel to all of this? I know there's a view amongst some investors that the channel is costly, that Arista needs a channel developed right from the get-go to get this going. Is that really true? Or is the channel something that comes along later?

Yes, I think it has a lot to do with how you think about the channel over what segment that we're focused in. And we're clearly focused on the Fortune 2000 and folks that really care about their network. We still believe we have to make the value proposition to those customers and pull the channel in to executing on deals. We're looking for channel partners who have kind of a full breadth of technical capabilities that reach network storage and put that whole solution together. And as we've been successful with these Tier 1 accounts, we'll find a region or geography that come together and help us build that and actually go find other new deals.

So it's more about direct distribution, direct sales and they bring the channel and to help to keep the deal.

That's the way that we're dealing it, at least with the segment we're going after. I think it's important when we say campus, a lot of people associate that with mid-market, that's not our target.

Right. All right. Let's talk about wireless a little bit, where you see that going? You've made an acquisition in wireless. But what differentiates your product there? Are you seeing, I guess, demand for Arista switching hardware and other people's wireless? Or do you pretty much see most of these deals going on -- both your switching hardware and your wireless solutions?

Yes. Let me start with the second part of that question first. So just in general, as we're going into enterprises with Arista technology, whether it's data center, routing, data center interconnect campus, we're promoting and espousing the concept of interoperability. All our cloud providers have multiple sources and there's some mythology, I would say, in the enterprise that you have to have a monolithic, end-to-end 1 supplier architecture. And people quite aren't happy with that. I mean when you look at storage market, people are used to having multiple suppliers or with compute. Enterprise networking has been a bit obfuscated in this conception that you need end to end. So we'll go in and we'll find a project. It could be a wireless project that uses existing switches in the access layer. It could be a data center project that takes out a data center and keeps the access layer the same. So we'll engage with an enterprise kind of end-to-end Arista value proposition and then find those insertion points. And we found several in the enterprise where it's been wireless only. And a bit of a surprise perhaps to us because we've been talking to some of those customers for some time. They're interested in the data center, but the first project that moved happened to be wireless and vice versa.

[Audio Gap]

Just APs. APs in our back-end cloud technology. We've also seen kind of distribution switches where they're retaining the brownfield access layer, maybe come back to Arista in the future, but that was their first project they executed.

Why do you think they came to you for wireless-only type projects like that? What is it that made them want to come to Arista?

I think one thing that's been important, we've been known as a data center company. We added routing. And especially in the enterprise, that routing part of the spend is perhaps a lot less than the actual connections to the servers, but very strategic. If you're thinking about data replication or connectivity to the hybrid cloud, and now with campus, I think we're being viewed as an enterprise networking company and being looked as an alternate to the incumbent, more from an end-to-end perspective than we ever have before.

So people do see you as, say, there's a knife just trying to your tear your -- their company.

There's another game in town where -- did well in the Gartner Quadrant with data center, why aren't we talking to these people?

Okay. What about challenges or learnings in campus networking so far? It's a new area for you. Could you talk a little bit about maybe some of the things that you've come across, and maybe didn't expect to come across or other challenges that have cropped up?

I don't think there's been any huge surprises. There's certainly some new areas that we have to focus on. It's the Wild West of devices in the campus. Everybody here has a lot of devices. How do they connect. Some of them are well known. You have Android devices, Apple devices. You go into health care, there's monitors and heart pumps and all kinds of interesting things. So we've added technology in CloudVision that can visualize all the devices that are connected to the network and then how the traffic is flowing either to devices in the data center or devices that are going out to the Internet and being able to drill out -- drill down on that front. [Audio Gap] this can create some interesting interoperability challenges that sometimes we have to go in and figure out as well. So that's a big piece of maybe what's moved from a technology area in the campus that we didn't have to deal with as much in the data center.

Okay. And you talked a little bit about wireless as a starting point in some accounts. Do you think -- is it an 80% of the accounts started in the data center and moved toward campus? Or do you think it -- maybe it's not like -- that's how I perceive it to be, but I'm not sure if that's accurate?

We've been pleasantly surprised by the mix, but certainly, kind of the highlight of it is we have data center customers that have been -- enterprises that have used Arista in the data center and looking to build out with Arista. So that's really kind of a home base. And then I would say also, we've had strong verticals based on our data center in financial services, media and entertainment. We've seen a lot of interest in campus from health care, which perhaps in the data center alone, was not probably the strongest vertical. But now with the campus portfolio, a lot more interest.

Is that because the health care people are leveraging AI, larger compute installations on-site and maybe would naturally want to think about campus or what?

Maybe -- not really. I think one of the things we did with this cloud principle is we saw the cloud providers building out a management framework for their entire networking infrastructure by putting agents on the switches that come back to a big data set that then can be managed. We emulate that with CloudVision. And for a lot of health care providers, they have mission-critical networks. They don't have huge staffs. They're difficult to manage. And that CloudVision component of being able to aggregate the entire estate and manage it very simply, I think, has been -- have mundane operational concerns.

So we -- I mean, you guys are just [Audio Gap] just monitoring what broadly is a technology in your products because you've started that subject a little bit. Could you talk a little bit more about why is that an important technological direction? It would seem like something that maybe you could already do, but then you're kind of expanding your technological capabilities there. So could you tell us a little bit more about what you're thinking there?

Yes. So just specifically, we did not announce an acquisition. But -- so I won't get into that.

Okay. Well, it came out in the press.

There's been some news reports as such. But we have some technology today that lives inside our switches called TAP Aggregation, so you can monitor multiple points coming in from a variety of switch routers into a TAP Aggregation point. And then effectively push that data into CloudVision, which is our management center to analyze that data. It's very critical for financial services, if you do trading, time stamping packets, understanding the jitter. It's interesting for voice applications or realtime applications. We see a lot of that in service provider kind of settings. And we think that's growing from a market that was kind of a separate segment, so people that sell physical TAPs and analyzation infrastructure. They're just something that you have as part of the network and the tools that you're using to manage your network devices should have that monitoring capability just built in.

And are you seeing expanded demand for this kind of monitoring? What is it that is going on out there in the data center world that might be driving an increased interest in this sort of technology?

Yes, I think there's different use cases now emerging, I talked about hospitals and campus. I only know what my devices are. Those devices sometimes don't have certificates. So how do I secure them? Possibly the only way to know what's happening is to be able to monitor the traffic flows? Is it going to a known IP address? Is it going to the video camera, proper server? Are you starting to see traffic flows from that camera to some place in outer space? That might be in a different country, that could be a little concerning. So the monitoring capability and mapping for security becomes more of a forefront in the IoT world.

So it's much more of the proliferation of devices than a hyperscale thing. It's already going on in hyperscale, but it's just campus proliferation of devices that are driving it?

Correct. Yes. Hyperscale will be more around application, latency, utilization of resources. But it manifests itself in different ways and different verticals.

Could you maybe talk about campus wins a little bit? I don't know if there are generic examples you can give or specific examples, if you want. But things -- places where you've won, how you won, why you think you've won these deals?

Yes, I'll talk about some maybe generic examples. I think we've done [Audio Gap] that was a traditional enterprise stronghold for us in the data center. So we have reputational capability within that market segment. People move around. They've used Arista in the data center, they go to a new place. And some of those wins have not been data center first. They have been campus wins first. Some attraction and some of those to our segmentation technology, which allows users in the campus to connect to various segments that then can be controlled through traditional firewalls. And what crosses those segments are the security aspect of that company, not the network provider, but the person who's running the firewall. So that's been attractive in a couple of deals. Health care, probably a bit of one that we were going after, but probably found more success than we expected early on. And that has been around CloudVision, the simplicity of managing kind of everything Arista, and not having to deploy as many services and people to analyze defects and security alerts in the infrastructure.

Okay. Let's talk about 400-gig a little bit, if you don't mind.

Switching gears.

Switch gears to data center, hyperscale, a little bit more. Could you talk a little bit about the 400-gig technology cycle, just kind of, A, what is it really? I mean we have -- there's a lot of misconception, I think, and definitional fuzziness about what 400-gig really means. Some people mean a 400-gig switching chip and then potentially serving 100-gig ports with it. Some people mean actual 400-gig ports in aggregation, which I think would be more Arista and where Arista is. Could you just talk a little bit about -- maybe why don't you start with the 400-gig chips, and what is going into the actual 400-gig ports and optics and all that?

That's a good point. It's kind of a big umbrella. And I think what people think about what 400-gig is, when does it all flip, right? When does the base turn? So all those pieces have to come together in kind of an ecosystem to make that happen. From a switch point of view, I think things are actually in pretty good shape. So we've seen our announcements and other announcements around chips that are 400-gig native. Now those chips can be used to build out switches that are 100-gig. For example, our 7368 is a 100-gig switch. But in its core, it's a switch that people would recognize to build out a 32-port, 400-gig Petabox. And we've used that to actually get better power consumption in a 100-gig setting. So I think that ecosystem around 400-gig chips is pretty, has a good footing and kind of ready to go. Certainly, there was a lot of angst probably last year when we talked about 50-gig PAM4? And is that going to work? I think those challenges have been resolved. When you wait sort of on an optical ecosystem, and there's 2 pieces to that. The ZR is one that gets talked about a lot, apart that from in it, but then there's kind of the garden-variety 400-gig optics that live inside the data center, and you're starting to see some of those in production, but the bulk of them coming together probably this year, mid this year. ZR, which is a long, long reach optics. It is very interesting because you can interconnect multiple data centers by using these optics inside a switch or a router, without the need for additional optical equipment. People are very excited about that. That's probably more into kind of production mode, the back half of this year, with samples kind of starting now. Those kind of transitions tend to lead from the backbone first, so I get the highest performance in my backbone and then I build out the leads. Does that help?

Yes, that's helpful. I wonder if you could talk a little bit about why you think some competitors in the industry keep talking about 400-gig as an insertion point. What do people mean by that? Because it seems to me, at least some of the people we've talked to, really are talking about 100-gig ports. They're talking about top of rack, maybe competing with more white box, Broadcom-based solutions and things like that. I wonder if you maybe could help us understand why you think some people...

Well, we lead in 100-gig, right, by many market share components. We're the port leader in 100-gig today. So if you're not us, you want to find an insertion point. So 400-gig is something, I think, folks are rallying around to try to kind of unseat our success there.

Do you think it is an insertion point, I guess, is the question?