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

Hello. Good afternoon, everybody. Tim Long here at Barclays. Thank you for joining our TMT conference for a fireside chat with Arista Networks. We're happy to have John McCool, Chief Platform Officer, SVP of Engineering and Operations, with us today to talk about the business. So John, thank you for joining. I think maybe we'll just start -- get right into it.

Obviously, one of the hot topics in this space in the data center part has been the transition to 400-gig. It's obviously taking a little bit longer. So maybe if you could just kind of update us on the timeline that you're expecting for that to be more meaningful. And then if you could just touch on, obviously, Arista has done really well in 100-gig, so maybe talk to us a little bit about how you think that success positions the company going into this next cycle.

Sure, absolutely. So yes, 400-gig has taken a little bit longer than probably folks would have projected 3 or 4 years ago. At the beginning of the year, we talked about this year, laying the foundation for the transition. The switching products, our products, were out ready for 400-gig, optics beginning of this year, not very mature, still being qualified as things progress through the year. I think we've seen a full suite of inside data center optics being developed, being able to be deployed. The ZR optics that give the extraordinary long distance being a long pole in that transition, but I think that's moving into market at the beginning of 2021. So we've -- from our platform perspective, we're feeling pretty good about the progress we've made in terms of moving through early field trials, into early deployments with the 7800 Series and also our 7280R3 Series and some of those products as well. And I think we'll see the tier -- the top tier, the data center interconnect, move to 400-gig the most quickly. I think our position with the cloud providers, more generically, that we established 200-gig serves us very well going into the next cycle. Some of that was around our density of 100-gig. But we had a lot to do around the ability to integrate EOS operating system into our customers' management stack, run their software on top of EOS to provide that integration, the visibility capability we had in EOS and probably more fundamentally just the foundation quality that we had in that software stack. So we see the same thing happening with 400-gig. The one thing I would add that's a little bit different about this cycle. When customers were transitioning from 40- to 100-gig, they really had to buy a 100-gig port. And if they wanted to continue to use their 40-gig optics, they use that port at 40% of its capacity. So it was a bit of a waste, right? So you saw a pretty wholesale move of products that were 100-gig-ready. In this transition, the way that the silicon has been built it's been constructed that it can be deployed at 400-gig or with multiple links of 100-gig without wasting any bandwidth. You construct a 400-gig link out of 8 50-gig lanes or you could build a 100-gig port using 2 of those lanes. So there's much more mix and match in terms of capability from the products that we're deploying and bringing to market. A good example is the 7368. That's 128 ports of 100-gig. You can also operate that product with 32 ports of 400-gig. But the sweet spot today in terms of the marketplace is that high-density 100-gig deployment. So we don't expect that all customers will move simultaneously like they did in the 400-gig cycle. And we'll see a diversity in that transition across the customer base. And that will show up in just the port metrics that you see from some of the analysts.

Okay. Great. And maybe just to follow up, you talked about kind of the integration of software and EOS as one of the real things that helped Arista in the 100-gig transition. When you think back to the pain and the process of going through that, do you think that provides challenges for other players that are going to try to recapture or insert themselves in 400-gig if they weren't as strong in 100-gig?

I think it's something that people have to work through. And it's -- when you talk about these transitions of port speeds, we tend to think only about the hardware dimension, right, and my box has more 400-gig ports than yours, I win. There's a software element to this that's very fundamental. It goes down to integrating your software with the ASICs; to the stability of the operating system; and then third, to the integration in the management stack. You have to have all 3 of those elements sorted out to get into high production deployments.

Right. Okay. Great. Staying on the cloud topic, could you talk a little bit about the whole move to white box in parts of the network? How you currently look at that transition for some of the big cloud players are more aggressive and some are less? How do you see that transitioning? And as that happens or if that happens at a customer, how is Arista still relevant in that portion of the network?

Yes, absolutely. So I think white box, on one angle, is competitive. But it's also for us synergistic. If you look at the white box share, it's dominated at the top tier of the cloud providers, 2 of them having developed their networks early on, really not having appropriate commercial alternatives and having no choice but to develop their own software stack and hardware stack built around OEMs. And we believe they'll largely continue with that strategy. As folks came on later in time, they had more of a dual sourcing strategy that has benefited Arista. And we are synergistic and will continue to be with that with the view, Arista and a white box source as alternate suppliers and value actually the redundancy. These networks are too big to rely on a single source. They also have their own operating system capability, which is important if you're going to deploy white box. You need an operating system that runs reliably on it. And they have also redundancy in terms of their supply in a commercial stack from us as well as their own internal source. They can develop their own internal software that runs on the same silicon that we, since we're merchant-based, deploy within our hardware and they're able to run their operating system on our hardware as well. So they have a very nice ecosystem between our commercial EOS, their own operating system as well as the hardware supply chain they're getting from us as well as the OEM. We participate in that business since day 1 and are going to continue to interoperate in that strategy going forward. And as Anshul mentioned on our call, we don't see any shifts in these strategies in the customer base going forward.

Okay, great. Maybe if we switch a little to talk about the campus. Obviously, this is a pretty big growth opportunity for the company, where you kind of took some time to be -- to develop the products internally. Maybe give us a walk-through of where we are. Obviously, you've added some wireless capabilities as well. So walk us through kind of the cadence of product roadmap and kind of what you need to continue to move this product along and then to the extent you can also talk about channel here because it's obviously a little bit different channel dynamic when you're going after, say, a Cisco in the enterprise market.

Sure. Maybe I'll start with the latter and the channel and then back into the products, if you will. I think it might make it a little more clear about our strategy. So I think when people think about campus, they think about a broad range of customer sets, some of them being mid-market-oriented. We're staying true to our Arista focus of the top tier of enterprises, think about it as a Fortune 2000. And that go-to-market strategy has been being built for many, many years. We started out in the financial vertical, in the data center only. We moved to media entertainment as kind of another vertical that was very strong for Arista and continue to build out verticals and go-to-market channels around the data center that supported that growth. We added to that capability organically into routing, so we could address not only in the data center but data center interconnect in the enterprise space. And then you think about the campus coming in as a third product set in some way into that element. And with each successive step, we've built out a broader customer base and channel capability focused on that top tier. So it's very much still Arista led from driving the value proposition. These are customers that deeply care about the network as a revenue-generating service, either producing media entertainment internally or exploiting this for some connection to their customers. And that's our view on how we need the channel to be able to reach and deploy those customers as opposed to more of a mid-market channel-led component. So I think that's important to establish. That lays the product strategy that we're building on campus is really oriented towards the top-tier enterprise customer needs. So we built the wired side organically, extended EOS to run at the edge, if you will. There was a significant amount of leverage from EOS, almost, I would say, 80% of it, 90% of it, some feature set around recognizing PoE management, features around access controls that needed to be added to address the edge, some integration with RADIUS server technology, and we've been through that. We've extended CloudVision to be able, this is our management stack for Arista networking products, to address concerns around the edge, like identifying what devices are on the network. We had very good technology around recognizing flows from devices from the data center. We were able to map that into devices at the edge and report on what flows are there. So I think from a fundamental software investment, we're very well established in what we need. What you're starting to see though, I think, is kind of a second wave of campus products coming out of Arista. We announced the 7050 modular series. There has not been a lot of attention on the modular space at the edge for a long, long time and not a lot of innovation. As you start to see not only users come into campuses but more hardwire IoT devices or even WiFi 6 devices, there's an increase in bandwidth and density that's expected. And this plays, I think, to our strength coming out of the data center in terms of core capability in terms of performance. And we see a strong interest in that product around high-density deployments in top-tier customers. We've also upgraded our acquisition of Mojo to WiFi 6 technology. And as we've gotten into both the campus and the WiFi side, I'd say our field and sales teams have gotten much more comfortable in terms of their approach in selling those products. Fundamentally, we lead with the position that you can manage and maintain your network from the data center to the routing to the edge in a compatible and synergistic manner without having to make specific investments to any one or the other. And it's proving to be a big advantage.

Okay. Great. Maybe if we could just touch on SD-WAN. That is something that comes up often. Curious what your views on that market are and how Arista will participate.

Yes. So we're not in that market with a direct product entry. So I think I would classify it as we play well with others with folks like VMware or a Versa, Silver Peak. And I think they're very much like on the security side, we're viewed as a trusted partner, not being a competitive product in that space. And customers that are making those choices, I think, see the value of working with Arista on the routing and wired side to complement their build-outs on SD-WAN.

Right. Okay, great. And maybe just to some of the verticals that you mentioned some already, but I was just hoping to talk about telco and 5G. And maybe you can touch a little bit on progress you guys have made with the evolution of the router that started off with some more limited use cases and now has grown pretty well with Jericho. So can you kind of walk us through how you're viewing the telecom market? Is 5G a real opportunity for you? And how should we see that through the product lens?

Sure. So when we started out with routing, we had great success in the cloud vertical and came out of the chute with some strong wins that, I guess, we'd look back and classify a little more as greenfield opportunities, so really embracing the change to IP, less about legacy features. And then as we moved forward, found that a little bit more difficult to replicate as more service provider customers wanted to build out in synergy with their existing brownfield investments in the core. So that required more functionality for us to put into EOS. And we continue to do that. And I think we've had some significant software releases at the beginning of the year. And we're starting to see the fruits of that labor. I think we have always had a strong telco cloud offering around the data center but continuing to build out our capability in terms of peering networks and next-generation core. And while we don't have an edge router for a 5G piece of the puzzle, as 5G comes into these networks, it does generate a need to look at your architecture, a need to change. There's certainly a lot of interest in NFV that tends to drive more of a telco cloud architecture. And the investments we've made in the last 1.5 years around software features, I think, will serve us well as customers start to transition to 5G.

Okay. Great. And I did want to touch on government as well. It does seem like there's a pretty healthy pipeline that a lot of your cloud partners might be going after or you might be doing some of it direct as well. So can you talk a little bit about opportunities there? And anything particular that needs to be done to your software or hardware platform to be more compliant with government standards?

Yes. I don't really have anything specific. It's very segmented. And I think we've done well in data center capability and deployments, especially in more of the large government settings. I think there's definitely opportunity for us to extend that into campus. And you do get into compliance or issues that are very specific to those different agencies that you're dealing with that would need to get worked through, but nothing that's fundamental to reach some of those customers.

Okay, great. Yes, I did want to ask one on silicon strategy. We did have Andy last year and he had a slide detailing the cost that it costs to develop your own silicon. Obviously, you guys are on the merchant side. So has anything changed in that dynamic? We do have now Nokia out in the market with their own purpose-built silicon, trying to make an impact as well. Of course, you have both Cisco and Juniper also. So what do you hear from your client base around silicon strategies? Do they want to make sure? Do they not care? What's your views there?

Yes. I think, first of all, I acknowledge that customers always like choice. That kind of comes up in this conversation, right? I wish I had many choices and many options. As a practical matter, it's getting -- it's more difficult to fund an investment in next-generation silicon. 7-nanometer products are out, moving to 5-nanometer. And without the larger silicon portfolio to drive those investments, it's getting extraordinarily hard. You're competing for wafer starts at TSMC or some foundry. You're out in the market buying substrates. You needed to develop IP for these chips, either in terms of memory that goes on the chips or SerDes technology. And a silicon company can amortize that over a large amount of use cases in compute and storage. And often, they're much larger opportunities than the networking piece. So you're subsegmented into not only networking but your market share in networking. So it's extraordinarily difficult. And I think there was some hope that the 400-gig transition, if you hit that, you'd be fine. This is a never-ending cycle. If you look at the merchant folks, they're marching down to new integration and proliferation of SKUs for every use case. And I think that's just a very difficult strategy to execute on.

Right. Okay.

The second piece I would offer is I did talk about the synergistic element of white box and Arista. And part of that is enabled by having a very similar feature set in terms of the silicon that you can amortize your investment across white box as well as the products that we deploy. You add another diversity into that mix, that's more to support and more investment on the side of customers to have that diversity. So it does come at a cost.

Okay, great. Also curious, this -- some of this may be a little bit outside the area. But when you think about your software, as software becomes more important to your customer base, how do you keep pace with that? What kind of new strategies do you have to employ? And the service business is pretty meaningful for Arista and growing well. How does that play into the way you look at these platform solutions that you're bringing to market?

Yes. I think if you look at the cloud segment, a lot of that software capability, management stack, network visibility is part of their software architecture and how they deploy the data centers. Outside of that cloud segment, and as we -- we're looking at driving cloud networks in the enterprise. They want the same thing but don't have the wherewithal or investment to make those software investments. So we started out, I think, with a unique architecture around the device operating system with EOS that provided the capability for a database that would allow a lot of network visibility. We added on CloudVision, which manages our routing and switching estate, if you will, not only in terms of configuration but visibility into those devices. We just had a recent announcement around the Big Switch acquisition that we made around integration with our Arista analyzer and capability. So you have now visibility being able to extract from the devices into the recorder nodes that Big Switch had developed. And I think a uniqueness that we have is not relegating that just to the data center but running against routing in campus, so you have network-wide visibility. And then with the integration of Awake, that provides a data source for NDR to be able to see network events and play them back in time. So you can see us starting to develop a broader network stack that, in some ways, mirrors what the cloud providers have done organically but be able to provide that to enterprise customers.

Okay, great. Yes, it's definitely been noticed as an area of focus and more opportunity. How does monetization work as you start adding these extra tools? Is it largely a retention deal -- market share-type game? Or do you start to see a little bit more of a revenue model as those other businesses start to scale?

I think, yes, to both. I think one piece is just on the core. As we're competing against incumbents, we've gone from a great data center provider to more of an enterprise networking provider that can offer an alternate to the incumbent and more of a broader perspective. So for a procurement, which I invest my time even taking a meeting with this company, the answer now is yes because we can play in multiple roles, right? So there's that. There's the element of just expanding our relevance in the core and the enterprise. And then on top of that, there's service subscription that comes from either our WiFi offering and be able to manage that in the cloud or what we're doing with the network monitoring capability that adds on effectively to the core hardware.

Great. Interesting. Okay, John, I think we are running up against our 25-minute timeframe here, so really appreciate your time, appreciate the insights into the business. I'm sure it's appreciated by investors as well on the call. Thank you, everybody else, for joining. Have a great rest of your day, and stay safe.

Thank you.