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

Good afternoon, everybody, and thank you again for joining us here at Day 1 of the Stifel Cross Sector Industry Conference. We're very fortunate with us today is Arista Networks, and we've got John McCool, who is the Chief Platform Officer as well as the SVP of Engineering and Operations.

And John, the place I'd really like to start first off, if you will, is a little bit on the cloud side of the business. Obviously, last year was a challenge across a lot of different areas, but certainly, the Titan has felt that, I think, maybe even a little bit more than some others. Can you talk a little bit about what you're seeing now from that customer set? You've had a couple of good quarters here over the last 2. Is it a case of just things opening up in general? Are we -- A, is it some pent-up demand, combined with a better operating environment? Just help frame, I guess, some of the return to growth that we've seen in that segment here over the last several months?

Sure. Yes. I mean, if you look at 2016 to 2019, right, just a phenomenal streak of growth and build out of those cloud platforms. And I think we're seeing kind of -- first of all, they're seeing a ton of demand based on COVID. We're all home doing this, and there's a lot of focus on the cloud and build out of their businesses and just the relative strength of what they need to keep those networks running. And then obviously, I think we're at the early stages, but the beginning of a build out on 400-gig, which is in front of us, but that's also an exciting piece of the cloud story.

So maybe there on 400, what's -- I guess what's the right way to bring that sort of opportunity? Obviously, at 100, particularly with those Titans, you had really this greenfield market that was largely left to you. Should we think of 400 as really an upgrade cycle, where those existing customers will be replacing some of their older 100-gig systems at certain points of their networks? And is there anything, I guess, to make some of the competition being a little bit more prepared this time, if you will, relative to where they were when 100-gig was first coming out?

You packed a lot into that question. I've to disaggregate that a little bit. Let me start with the competition point and sort of this framing that we had it to ourselves last time. It was highly competitive. I mean there are -- the same competitors that we went up against in the 100-gig cycle are coming at us again. And if you step back and look at the strategy, very similar strategy, right? I think we're focused on merchant silicon exclusively, and our competition either solely homegrown silicon, vertical integration or some hybrid mix. So in terms of the strategic point, I think, still very much the same. We are the -- incumbent now, which we weren't the last time around. So from a competitive framing, I think that's where we are. In terms of the build-out, there's a couple of differences. With 100-gig, customers were forced from the legacy 40-gig to really use 100-gig port, either for the new 100-gig optics or they could backfill it with 40-gig optics. And from a market reporting, you saw a lot of 100-gig ports go into the market that were actually operating at 40-gig. So we saw a big transition from a market standpoint, where I think the time, the customers themselves were kind of wasting 60% of the bandwidth. With this new cycle with 400-gig is more elegant, if you will, from a technology point of view, they made a technology so you can have 1 400-gig port, you could break it down to a few, 2 100-gig ports or 4 100-gig ports, but still use the high-density chipsets. So you're going to see customer choice being more spread out against those different speed grades. So we look at the opportunity as the 400-, 200-, 100-gig cycle and then upgrading them to our new products to take advantage of that densi . Does that make sense?

It does. And I think that's an interesting sort of clarification. One of the things that I think is somewhat, I guess, unique about that market as we think about it is -- there's all this, obviously, focused on the early adoption, if you will, of the cloud providers because they are the ones that clearly are, I think, of most need of this bandwidth upgrade. But when we think about the life cycle of that product, to your point, having the ability to then move 200-gig lengths or 100-gig lengths even with that faster chipset, is that enough of a reason for an enterprise data center to make that upgrade for an -- is that type of flexibility? Does that give them advantages or do they wait until they're absolutely say, tapped out at 100 and then that's just sort of the natural then evolution to move forward?

Yes. I think even with the cloud folks, right, the -- it starts at the aggregation point. So the server is up to the top tier, and that's where the pinch point is where 400-gig will enter and the interconnection of data centers or regional data centers into kind of a single virtual data center using high-performance interconnect, and that's why the ZR optics is so interesting because it helps them get involve in those connections at the regional point, it's a presence back into the data center. There are 100 -- there are 400-gig applications in the enterprise. They tend to be smaller just given the size of the enterprise and more maybe mission-critical machine learning kind of environments, high-density routing environments and those type of applications.

Got it. And you mentioned ZR, is that a precursor to 400-gig switching? Did we need that to be available to start to drive some more of the demand of it? Or is it more complementary, I guess, in terms of, to your point, it allows for better aggregation, it allows you to use -- to truly utilize the capacity of the switch to its fullest, if you will?

Yes. I think it was an important ingredient into the mix because it enables more of a holistic view of whether your regional data center, your data centers close by or not doesn't matter with the ZR optics, you can interconnect, just fantastic distances in with that technology and then integrate the optics right into the switch.

Got it. Got it. And as that technology road map continues to evolve, are there things that you have to do as you're working on the next generation to then incorporate, say, what may be coming with ZR+ and 800 and things of the -- does the switch itself have to change a little bit to be able to power those types of pluggables, to be able to -- or really, you can do the work you need to do from a development perspective, independent of that ZR sort of road map and community?

I think we did a lot of that work over the last 3 years. We did have to anticipate the higher power consumption of the ZR optics like 20 watts per port. So we had to design these products so they could contain the ZR optics. On another thread, you talked about 800-gig, there's going to be a transition in the future from the signaling technology that's deployed now, 50-gig PAM4 to 100-gig PAM4 native. And especially in our modular designs, we've anticipated that and planned for those type of upgrade cycles absolutely. Our 7500, that had over a decade run, it was a 10-gig modular switch we took through 40-, 100-gig and now 400 gig. We started the 7,800 product line to begin with a 400-gig cycle and have an ambition to take that forward in time as well.

With a lot of the things that you just talked about there, 2 questions sort of come to mind. The first is on the incumbency side. You mentioned as you're starting this 400-gig cycle, you're now the incumbent. And I think, particularly with cloud titans, there's this bias to sort of think they're quick to sort of rip and replace and always looking for either the best cost or the highest performance and are very quick to sort of swap out. So I guess, is that the right behavior to think of with the cloud titans or having that incumbency now at 100, does that give you somewhat of an advantage as you start to look at for?

I wouldn't say the incumbency gives us a natural advantage that, "Hey, you're here," and people are just happy to have you around. I mean we've got to earn it. But there's a learning, especially in the operating system and software, the 100-gig cycle was also met by Arista with EOS and an operating system that had the stability and quality level to scale out to millions of servers interconnected. And going from 2 to 4-way to 8-way, up to 128-way parallel devices that are operating at the highest tier of the network. That scale is not simple. And we've been able to achieve it with our current platform and then migrating to 400-gig using similar merchant technology is just an easier path for us to meet the needs of the cloud titans. So I think that's where the incumbency helps us. We've done it and can build on top rather than starting back with new silicon, building a new operating system and trying to meet their requirements for the first time.

Right. Okay. And then the other topic that I think comes up quite a bit, obviously, is on the white box side. And I would think as we continue to move forward to your point, what you've done in the switch now to incorporate what's going on in the ZR world to do a lot of those things, I would think would help mitigate some of that concern around the white box threat. And that is still much more of a commodity side of the network. Is that sort of the correct line of thinking? Are there other issues that investors should be focused on when it comes to that competitive white box threat, if you will?

I think that's a fair characterization. I go a little bit further, too. If you look at sort of where white box has gotten the share is largely with 2 of the largest cloud titans, who started before Arista even existed on that threat with white box and homegrown, and they continue to dominate the share of white box. If you look at where we've played with white box, an example could be Facebook, we announced 7368, they had a white box variant of that technology could run their operating system on Arista's hardware platform or EOS and they were very complementary, and that was part of their sourcing strategy to begin with. As you go down to customers that have less servers, less interconnect, less network need, it gets harder because there's a fixed cost for engaging in the white box direction, independent of how big your network is, and it's just more just difficult to justify that make by decision as we go down into smaller networks.

Okay. Okay. And then maybe just a last question here on the cloud market for you. When you're thinking about, say, the next 12, 18 months or so, is the bigger driver of that business new footprint now getting -- being installed again after what was largely put on hold because of COVID last year or is it this move toward 400-gig in that switching environment for them? I guess, when you think about that, the demand environment for that cloud market, which of those 2, I guess, is the bigger driver of more near-term demand for the cloud market?

It's tough to break it out. I think specific to 400-gig, there's just a growth driver in terms of that bandwidth that people are looking forward to as it hits a maturity and price point that's acceptable for the transition. And then I think there's still more regional build-outs of cloud and cloud capacity that's going on as well.

Got it. Okay. Maybe shifting gears a little bit then, I would love to touch on the enterprise and on the campus. Obviously, a big push for Arista here looking forward in terms of where you see a lot of opportunity. When I think about more of the historic version, if you will, of your enterprise segmentation, which was very sort of enterprise data center focused, if you will, that market, can you just talk a little bit, coming out of COVID now? Obviously, to your point, you've got more shifting to the cloud. We've seen the cloud benefit from that. How does that, I guess, impact that enterprise data center market as you're looking out over the next 18 or 24 months?

Yes. I think, certainly, I think enterprises have rationalized what they run in the cloud and what they run on-prem over a long period of time, right? And the people that we talk to are still committed to workload that's in the data center. And during COVID, much like the cloud providers, they had to think about am I delivering the services to my employees and partners not just on-prem, but off-prem and in different locations. So I think there were some rationalization of the interconnect points on those data center networks and maybe a little more focus on the routing and security aspects that were required to deal with people working from home. So that's a component of it. I think one thing that's important for people to recognize about our campus strategy is it's part of our enterprise strategy. So when we take that data center core, which tended to always be focused on kind of Fortune 2000 customers and verticals that really valued the network, financial vertical, media, entertainment, we've expanded that now by having a campus offering, but still focused on those type of accounts that are really focused on their network as a revenue-generating component and something that's mission critical to their operation.

Got it. Okay. Okay. And then when you start to, I guess, move closer to that campus edge, is it tying back to the data center that helps you really differentiate? Or as you come into that market, I guess, where are the points of differentiation with the product portfolio or with the solutions side that allow you to really compete against an incumbent like a Cisco, who's largely been dominating that market now for 3 decades?

Right. Ironically, tying it in is maybe not the exact word, it's the consistency. So you're a data center customer. We entered routing. You can interconnect your data centers with our products, so multiple data centers. And we have CloudVision. And oh, by the way, you can connect your users and your edge through CloudVision in the same operating system, same EOS binary. So in some ways, it's easier to extend your data center Arista footprint to campus than it would be to buy another architecture from my incumbent that uses a different operating system, different management stack, different set of protocols than they use in their routers or their data center. So it's really about the simplification of your operations is the value proposition from Arista.

So I guess to that end then, is it a data center-led sale that leads to -- or I mean, or is there opportunities, I guess, to go right to the campus type of customer who may not be -- and if they do, do you then work it the other way and try to sell up into that data center, I guess?

We have done that, and it becomes an Arista enterprise sale. And we look -- our sales teams look for points of entry, and it might not even be those 2. It could be a management network. As we're selling the value of simplicity of operation and quality, anybody can say that. You don't really believe it viscerally until you use it. So we look for an insertion point in any piece of our portfolio. And then after the network has been running for a while wasn't that easy, wasn't easy to upgrade. And we're ready for the next opportunity that could be a campus opportunity or data center. That's how we're operating.

Okay. And along those lines with the campus, Arista historically was always the performance leader, the price point -- things of that nature. When you start to get into the campus environment, I guess, how do you protect against sort of application sprawl, if you will, for lack of that you have to provide all these other services or needs. And one that I think often comes up is on the security side, right, where security, obviously, particularly at those -- that edge of the network is always such a focus point. Should we expect that that's then addressed through M&A, these different needs or is it more partnership driven? I'm just curious how you sort of go about addressing maybe that sprawl of needs, if you will, other than just a straight networking aspect of things?

No, that's a great question. I mean, first, we've been -- we were very deliberate about entering at a point where we felt like we had done enough in the data center, and we're able to look at that -- the edge more holistically. In the data center, we've also partnered. So partnership is extremely important. We partnered like with Palo Alto Networks development of our macro segmentation technology that ties in with the firewall that prevents movement of East-West traffic. So we decided not to be a firewall vendor, but to partner and partner with our API. Similar thing we've done in campus. If you look at what we've done with group segmentation, we have a similar -- a concept where a security partner can put policy on top of this and create groups effectively that are enforced through the network. We have done acquisitions also where we felt that we didn't have the technology organically. In the campus, we didn't have radio technology. So we made an acquisition of Mojo for the WiFi component. And the other piece, the visibility across the enterprise network is critical. So it's not just data center, it's data center routing campus. What we've done to take EOS and our native visibility format and combine that with the acquisition of Big Switch to pull -- not only pull the data out, but be able to log and categorize it and then put Awake on top, builds out a security stack along the network focused on the network and what you can observe in the network itself.

Is there a way to, I guess, attack the market with what you just described more on the software side of things, whether it is creating a layer where you can almost drop EOS and connect disparate elements so that you get that visit and offer that as a way into the enterprise as well?

Absolutely. I mean, if you look at these 3 elements I mentioned, they're very much a software appliance kind of approach. Our WiFi stack is not an on-premise offering. It's running in the cloud. So it's a cloud-based service. If you look at what we've done with Big Switch, that's effectively running on servers. It's on-prem, but it's a software-based approach and licensing as well as what we're doing with the Awake absolutely.

Got it. Okay. Okay. And I guess as we're looking out over the next, again, let's call it, 18, 24 months, Jayshree has talked a few times about the company is shifting more of a focus on to that, that enterprise and campus market as it's sort of long-term growth driver, how do we think about, I guess, that market opportunity relative to cloud? This cloud kind of come in and out and provide a little upside to revenue numbers. And really, it is this enterprise piece that investors should focus more on for the sustained sort of more forward-looking growth for the company over the next several years?

Yes. I mean, there's no doubt we're focused on diversification, and we're counting on all these segments to grow and pushing hard to make them all grow. Cloud build-outs tend to be lumpy with any cloud provider. They tend to go fast and want to build out their networks. And enterprise, there's lots of them. So it's more puts and takes in different places. So it all depends. I think on the enterprise, the piece that's exciting is we're so underpenetrated today, both in data center and even within some of the large accounts we're in, and we kind of view it as mission-critical, There's still upside in a lot of those places given the incumbency of our competitor.

And given the -- I guess, the incumbency that you have in a number of those Fortune 2000 accounts that you mentioned on the data center side, is it still more of a direct sale for you in the enterprise? Or as you're, I guess, targeting more and more of that edge, do you need to rely a little bit more on the channel and building out that capability in those relationships?

If we were focused on mid-market or SMB, it would probably be different with the channel. It's definitely a direct value proposition led from Arista and often a channel fulfilled or build out. And as we've done more of those, I think the channel has become more comfortable and familiar with our technology and prepared to deliver effectively to those enterprises. But we still need to go in and provide the value proposition and why should you have another supplier? What can we bring to [indiscernible] and really do some -- the heavy lifting on the value proposition side.

Got it. Got it. And then, John, as we think about the service provider market and what that might be over time for Arista. Is it -- I guess, where do you guys really focus on where the opportunity set lies within service provider? Is it in this evolving mobile edge? Is it -- or is it more on some of the more core routing elements, I guess, that traditionally have been an area of focus for your network equipment manufacturers within that environment.

Yes. We're counting on some transformation of the service provider networks. They made no sense for us to build out kind of the legacy protocols and turn around the -- and early out of the chute, we actually found some opportunities that were really well aligned to that kind of approach. Unfortunately, it wasn't as sustained as we would like because providers are moving albeit a little bit slower than we would, and we also needed some more technology features that would tie into their existing legacy networks. So Ken and our software team, I think, have done a good job over the last 24 months, building out a significant set of functionality to allow us to tie in better. And some of that actually is a per provider-based features that tie into specific operators. So that's good. I think there's opportunities, transformation of their cores, when they're ready to go from their legacy core to next generation, I think we're well prepared and have seen that in a couple of situations. The telco data center, if you will, as we call it, putting functionality in servers, we're extremely well positioned for that trend and do well when things move in that direction. And that often fits well with the 5G opportunity because they're doing transformation of functions back into the data center. Content delivery network is another one where our footprint and routing capability and the switches has been proven to do well.

Right. And I guess in that sense as well, does EOS have a role to play in the service provider network in the way that it does, particularly with enterprise and even in cloud? So I'm just curious if EOS has or CloudVision even have that same sort of advantage there or that same sort of value prop there that we see when we look at a couple of your other markets?

Yes. I mean in the enterprise is where CloudVision has really taken off because of the simplification of the operating methodology probably less so with the service providers, but the quality and capability and uptime definitely plays well with the service provider market. Just the pure stability of the operating system, ease of upgrade, et cetera.

Okay. Okay. And I guess, we've only got a couple of minutes left, so I do want to be mindful of the time here. When you think about that cloud market, and you mentioned that it's actually a fairly limited number of customers and there's a few of the larger ones, obviously, even really white box focused. As those customers -- I guess, the growth with those customers, is it more dependent on them growing geographically so as those cloud customers continue to expand into others or is it still just the amount of traffic on the network regardless of where geographically it might reside?

It's both. There's certainly the traffic in sort of the core networks is growing, but there's also not just geographically, but just regional build-outs of the edge that are interesting in terms of their expansion of footprint.

Got it. And I would assume, obviously, as 5G starts to ramp up a little bit, that adds both on the service provider side and on the cloud side, just gets to be hopefully a larger source of demand as well.

Sure. Yes.

Last question for you, just a little bit on kind of what's going on in the supply chain side a little bit. Really 2 questions there. You've managed, obviously, to mitigate some of the near-term challenges here, I guess, where you see it now? Do you see it impacting your customers' demand, meaning, are they struggling to get everything they need to do full deployments and that's been rippling down to maybe their demand on you? And then a second question that's a little bit different there. Given the reliance, I guess, on merchant silicon, do you worry at all about not only those supply constraints, but just having, I guess, you're making sure that those silicon providers keep pace with what you want to accomplish as you're looking out over a longer-term period?

Yes. I guess I'd start -- it's been 15 months we're into this. So this is a continuation of a lot what we've seen, and it's shifted certainly with more focus on the semiconductor segment more recently, right. But it's still the same inherent issue, continued good demand for IT and electronic components, not just in the networking segment or IT, but automotive is being blamed substantially for the current shortages and the demand they're seeing there. As we look down in our supply chain, we're not just focused on the merchant piece. It's just every piece of silicon that goes on to a particular SKU is just been -- the lead times have extended. Upstream, demand hasn't decreased, and our customers also are shifting their focus to really appreciate that our lead times have extended commensurate with that and planning accordingly, if you will, and having a plan in longer durations than they typically would have had to do in the past.

Right. Okay. Well, perfect. Fortunately, that's all the time we have, but I really appreciate you taking some time for us, John. And best of luck, and I'm sure we'll catch-up soon.

Okay, John, thanks a lot. Appreciate the time.

Thank you.

Bye-bye.