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

It's largely a volume increase at this point. Some price increases are already in there, more will show up in the future as well. So we have to separate these out. Number one, from a inventory and supply standpoint, if we -- depending on lead times and where we're seeing constraints and this changes almost weekly, if we need to buy more or try and build a buffer for some chips and so on for future demand, we will just go ahead and do that. So if needed, that number can go up even higher. Number two, the price increases, whether it's commodity items, chips, substrate, labor impact, we've seen price of steel and copper go up, which sort of has an indirect impact on some of our components, we've seen the price of freight go up as well in moving components back and forth and so on. So all of that is certainly starting to happen. I think this is what Ita mentioned during the earnings call as well that all this will flow through our costs and impact margins at some point, but it takes a while for this to flow through. So the real sort of full impact, I think, will come next year as higher-cost components are revenued.

Okay. I was just going to ask on the gross margin side, it was pretty strong in the quarter here despite the supply chain constraints. I guess if we weren't kind of in this time period of dealing with these constraints, I guess, where would margins be? And then as you do think about the costs being layered in or the elevated costs, how should we think about that kind of going, going forward as well?

Yes. There are 2 ways to look at this, right? We've been guiding 63% to 65% gross margin for the company for quite some time. So we believe we'll stay within that largely speaking. The biggest impact there is actually the customer mix and the product mix. And only after that, the cost of these increased components comes in. But then you have supply constraints, and the cloud business is also healthier than previously expected, then you sort of have 2 factors working against your margin, which, if you're just focused on the gross margin item, doesn't sound great, but the reality is that also means that we are seeing growth in our revenue, which is such more important factor at this stage of our company. So I think it's actually better received than people focusing just on margins. But having said that, all of this will play out, and it's extremely hard to forecast this precisely, right? Because you have the customer mix, you have enterprises but enterprises do have higher software content, which improves margins and enterprises also growing. And then you have all of these costs coming up in components, but because of the constraints, it's not like you have too much flexibility in any given quarter. We'll build whatever supply we have and then the quickest products we can build and ship, we'll just do that. So when you have 0 flexibility in those terms, you'll see the impacts. There's a little bit more volatility to margins than before. But that's a shorter-term impact. If you look -- model it for the next 5 years or so, we don't believe this is going to change our business model fundamentally. And we are doing some selective price increases as well for our customers, maybe not as big as what some others are doing, but we are passing on some of these costs to customers as well. So over time, these things balance out, and we come back to our business models, all other things remaining the same.

Okay. You mentioned the enterprise business in there. How are you kind of thinking about the enterprise segment going forward? I know that in the industry, there's been a lot of talk about a second half refresh or kind of into '22 with people not being able to get on-prem and get some of those equipment installed. But how do you kind of think about that and being able to capitalize on that opportunity?

Sure. When you look at the enterprise, sort of almost an implication in that question was enterprise equals campus. But for us, enterprise data center is a pretty strong segment as well and growing very well. We're doing very well over there. Our products are very well received. We are not -- no longer trying to prove ourselves as an alternate vendor in the market. People already recognize that and will accept that. And they know our technology is better, our platforms, our U.S. software as well as CloudVision, which gives you a lot of improvements in automation for the enterprise. And sometimes the same customers, sometimes new customers are also deploying our products in the campus, which has the same EOS software and the same CloudVision. So it's a huge operational benefit to these customers. So I think we are seeing good growth in the enterprise overall. To your second question on enterprise campus refresh post-COVID as an example, we've seen both types of customers. And we're not a bellwether for that market yet as our goal for this year is $200 million for campus. And what we have seen is during COVID, we have had many companies come and say, we want to wait on the upgrade because there's no one coming to the office. And then we've also had many companies come to us and say, we want to upgrade the campus because no one is coming to the office right now. So both parts played through. What is happening, though, is some of the network architectures are changing, and automating segmenting your workloads or your users in the campus, especially with flexible workplaces where you have these pods where people can come and mix, they may not have a dedicated office cube like before, and that actually plays in our favor because our portfolio already provides that from day 1. We don't have to develop something new. This was the advantage we already had with the segmentation solution we have, EVPN, VXLAN like technology and using some data center techniques applied to campus. So I think we're well positioned there, but we'll see how the market plays out.

Sure. I guess as we think about the broader portfolio that you do have in the campus -- around the campus now and you can potentially sell routing, security, some of the Wi-Fi and other software solutions, how is your customer base thinking about expanding just beyond the campus switching solution into some of these other areas?

Sure. Well, our enterprise customers do like to have a broader solution from their vendors, including us. And when they engage with us, it's not a point engagement just for 1 or 2 sites for the campus or 1 or 2 data centers. It continues to expand to go after other use cases, other roles we have as well. We are actively deploying DSL in campus, but also in monitoring with DMF monitoring fabric, which is very useful for enterprises to monitor what's going on. Sometimes that solution is purchased or deployed by the networking teams. Sometimes it's actually deployed by the security team to monitor what's going on in the enterprise. Similarly, we have the NDR solution, the network detection and response solution from Awake, which is more of an AI, machine learning-based technology to detect malware or threats within your enterprise, within your campus without having to deploy agents or other software on every endpoint. So with IoT, with cameras, batch feeders or the technology or even user traffic, the Awake solution goes a long way in detecting these types of breaches or threats and blocking them where necessary. So we are expanding well over there. Same thing, Wi-Fi is already part of our campus solution. We're expanding in some of the routing use cases as well. So I think, overall, that gives us a lot of confidence that our enterprise segment can continue to grow. It's a large TAM. We have a great solution. And we have, I think, many, many years of growth left over there.

You called out a big retail win in the quarter here with selling this kind of broader solution set. How did this deal come together? And what was the real differentiator for Arista that ended up being the winner here?

Right. I think what happens is when you involve with these customers, as I said, they start looking at the entire portfolio. And in each of our offerings, we have a best-in-class solution for their data center or their campus, both wired and Wi-Fi, and that solution, they manage everything from CloudVision. So the automation piece is such a big deal to enterprises because with just a few engineers, they can manage their real estate throughout the United States with a few engineers, which was unheard of for them, right? The number of people they have needed in the past is quite large and maybe we can go address that for them. So when you add all of that along with a Wi-Fi that is not controller-based but is cloud-managed, and they have more opportunities to grow with us even in the security side as well. But the data center campus solution fit very well. And everything that we're maintaining, you as an investment community played out with the value of 1 single EOS, 1 CloudVision automation suite goes a long way in easing the pain for these enterprises.

Okay. I want to ask a little bit about kind of the broader product suite here, and I think you mentioned about some use cases in the cloud around kind of load balancing for some of that. But how do you think about the opportunity in load balancing as maybe a potential product adjacency in the portfolio?

When you look at the offerings we have there and when we say things like load balancing, this is not head on competing against an F5 as an enterprise load balance or even a software load balance that the cloud companies are building. This is much more of traffic load balancing within their large networks. When you have traffic coming in from the enterprise to the edge, it needs to be load balanced to the right data centers, to the right links and so on. And we've built some good, unique technology there to help our cloud customers onboard enterprise workloads and map them on to their virtualized networks internally. So it's sort of mapping from the legacy tunnels that might have from the enterprise or a service provider connection onto a SDN type of sort of virtualized network on the cloud side, and we become that bridge in the middle doing all of that for them.

Okay. Got you. Also on the product side, I want to touch on a little bit what you're saying, but you had 200-gig and 400-gig upgrade cycle in that part of the market. I guess what is the status of those upgrades today? And how do you see that -- when do you see that beginning to inflect going forward?

Yes. The overall 200-, 400-gig cycle is not a complete replacement of the customers' installed base. So these will be more gradual than what we saw with the 100-gig cycle in 2017. But the good news is that this cycle has started. We had mentioned in the start second half of this year, we've been in a lot of pilot sites and deployments, test deployments with these cloud customers, and it's starting already. So you'll see that gradually grow over the next 18 to 24 months. There is also a lot of 100 gig. So you'll have to continue to look at the mixed 100, 200, 400 business for us. But our new product is also extremely sticky, and our customers' feedback has been very positive on the next gen. Most of the issues that were found in the lab and so on are all resolved, ready for prime time production now.

Okay. I think you mentioned on the earnings call that it's a different power architecture shifting from 200G and 400G from the prior cycle. How does that kind of impact the deployments that your customers are beginning to roll out? Does that mean that it's either more gradual? Or does that mean that they need to upgrade a broader footprint to make sure that they're able to upgrade you to those architectures?

Yes. That was in response to what I think has been a misconception in some of the investment community. But before I get to that, the way power is measured is now in picojoules per bit. So the picojoules per bit is indeed coming down as you get to 200 and 400 gig. So on a per bit basis, you'll save power. But there's been this underlying assumption by some in the investment community that when the world went from 40 gig to 100 gig, it was a quick 4 to 6 quarter, everything just switched. And the same thing is about to happen in 200 or 400 gig. That is not true. 40 to 100 gig was the same power per port. 100 gig to 400 gig is not the same. In fact, you're paying 2 to 2.5x on cost and power on a per port basis, on a per bit basis or per gigabit basis, per bandwidth basis, you are coming down. But if you don't need the 400-gig bandwidth just shift, then why pay the premium. So which is why it's not the same adoption curve. 400 gig or 200 gig are getting deployed where customers absolutely need it. But they're still deploying 100-gig where that's more than sufficient.

Okay. Got you. Okay. That makes complete sense there. I don't really have -- a couple of minutes left here. But I guess as we kind of think about the investments that Arista is making, I guess, where do you kind of see the biggest opportunities over the next couple of years, either from newer solutions or newer use cases that are beginning to emerge?

Well, Steve, as you know, we already have $32 billion plus TAM that we're going after. And we are roughly at $2 billion, $2.5 billion plus in revenue and growing. And there's a long road ahead of us. So to some extent, I think our plate is full. We have plenty to execute on data centers in the cloud, in the enterprise, campus for the enterprise, some of the monitoring and security solutions we have as well as multi-cloud connectivity we have with cloud EOS and related offerings. And then you take all of this and take to the next level, as an example, we launched CloudVision as-a-Service, managing infrastructure from the cloud, there are more opportunities there as well. So I think we just have to execute well in these areas that will give us plenty of room to grow for years to come. And new opportunities come up, we'll keep exploring those as well. But oh boy, the opportunity ahead just with what we already have is already good.

Okay. No, that's great to hear. You did mention some investments there in both the software opportunity and also with some of these things being run as a service in the cloud. How does Arista kind of think about those different deployment models and the opportunity there?

When you have a large site to manage, on-prem can work. And for security and compliance reasons, financials, banks, many sort of enterprises worry and like that. But like the retail, when you refer to, when you have a distributed campus, all spread out or a large company with 70, 80, 100 locations in the world to manage, then managing from the cloud with our CloudVision as-a-Service offering, or CVAS, is a huge benefit, right? You don't have to worry about infrastructure to deploy in every site to go manage your network. It's just 1 click and you're ready to go. So I think that will give us growth for many years to come as well in the enterprise.

Yes. Perfect. I think we're running up against time right now. So Anshul, really appreciate you joining us today, and I want to thank everyone else for listening in to the session. Hope this was -- this is great. So thanks again for being here.

Great. Thank you, Steve.