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

Perfect. Good morning, everyone; good afternoon, depending on where you are. My name is Amit Daryanani. I'm the IT, hardware and networking analyst here at Evercore. I'm delighted to have with us Arista Networks here for our next fireside chat presentation. And from Arista, we have Anshul Sadana, the Chief Operating Officer, to spend some time with us and hopefully make us a whole lot more smarter on Arista. We will keep this fireside chat around 30, 35 minutes. Irvin and I will spend about close to 20 minutes or so going through some of the questions we have prepared. We'll open it up to questions from the audience post that. If you have any questions, you can -- please feel free to e-mail them to me. Or better yet, if you want to use the chat function below the broadcast, you can do that, and ultimately, we should incorporate the questions as we go along with the discussion with Anshul. So with that, Anshul, thank you very much for your time. I really appreciate this.

Before we get started, I guess, I think everyone knows Arista fairly well, but maybe it's worthwhile spending a couple of minutes just touching on how has Arista evolved over the last few years. And maybe the [ topic ] to focus on is some of the recently improved demand visibility that you're seeing, what is sort of driving that? What is enabling that? That would be really helpful.

Sure. Amit, first of all, thank you for having me over for today's chat. Many of our investors do know Arista for a long time. We started with low-latency trading and then the cloud. But since then, we've actually expanded in the enterprise as well and Tier 2 cloud and service providers. We've gone way beyond just building leaf-spine switches to now routers as well or routed switches. We have more automation in our software with CloudVision. We've done acquisitions like Big Switch and Awake. So overall, the portfolio has expanded quite a bit, and we continue to grow our footprint in the overall market. Now in terms of visibility, what's happening is there are 2 factors. One, last year, especially focused on the cloud titans, everyone was worried or asking about, "Hey, how you are doing with your next gen?" But it was just too early because the products weren't fully ready, they weren't fully qualified. But this year, the products are now fully qualified and in their pilot sites and getting deployed. So as a result, we do have a bit of visibility on the demand from customers. And at the same time, as many of you know, the lead times and components went up. So that also is forcing us and our customers to plan better and talk through things on how will supply look like in 6 months, 12 months time frame. And as a result, the customers have to actually give us that guidance, too. So with all of that in mind, I think we have higher confidence in the overall demand looking out 6 months or even longer.

Perfect. And then one of the big topics, I guess, really overall has been what's happened with supply chain broadly. And I know Arista reported actually about a month back at this point, and I think Arista performed fairly well from a supply chain perspective and being able to procure components. But I think since, we've heard from Cisco, for example, a few weeks back, who talked about having procurement issues impeding both the top and bottom line, and they quantified the gross margin headwind, too. So let me just get an update on how do you see the supply chain dynamics impacting Arista. And how are you reacting to it?

Well, if you look at our commentary since February of 2020, when the impact of COVID started to happen around the world, we've actually been talking about supply chain constraints since then. And every quarter, something new has come up or something more unpredictable has come up, and we've had to buffer up or work around those issues. And I think now the entire industry is seeing it in a very acute fashion, especially on semiconductor-related sort of supply chain of substrate materials, things like that as well. Look, all of us are equally impacted by these issues. Supply chains are highly interlinked. And when there's a global pandemic, it's hard to escape this, right? It's not as if you build only in 1 country. You're impacted everywhere. Now we did invest quite a bit in our inventory and components over the last few quarters and managed to increase our levels relative to our historical averages. So I would say we are -- in the near term, we're able to manage through these issues. It's not as if we are living in a comfortable world. It is very much a tight situation. It is fluid, but we're able to manage through it, and hence, you were able to see that in the results as well. When it comes to impact on the bottom line, because it takes time for components to impact flow-through revenue and then impact the margin, you'll see what happens in a few quarters. But look, if the impact is small or something very temporary, we'll absorb the hit as far as we can. But if the impact is longer, then we'll have to have discussions with customers and passing on some of the cost increases to them. We're not at that point yet. So as our suppliers finalize their sort of terms and pricing and delivery commitments and so on over the next couple of months, I think we'll have to reach those conclusions as well. But it's a bit too early for us to say anything right now.

Understood. And then I think on cloud titans -- I'll maybe focus a little bit our discussion on cloud titans. Can you just -- like what is driving the growth vector with cloud titans at Arista very specifically? Because if I think about the last earnings call we had, you were all very positive on this vector but I don't think you're seeing the big 400 gig ramp happen quite yet in -- with the cloud titans. So it appears -- what is it that's driving the positivity with that vertical right now?

Yes. The cloud in 2020 was somewhat soft for us and for various reasons. Customers decided to pause. Customers decided to skip a compute cycle. Customers decided to wait for 400 gig. Customers decided to wait for ZR optics, for DCI. Lots of factors, nothing individual but everything added up. In 2021, we're now at a point where the next-gen product cycle is starting to happen. And these products are qualified, very positive feedback from customers as we've shared in the past. A lot of the work we do with our cloud titans is essentially core development. So these products are designed together. A lot of effort goes into that. And now you're seeing the fruits of that labor, where you can see the ramp happening. We have mentioned that we're expecting the next-gen product cycle, which is not just 400, by the way, it's 100 gig, 200 gig, 400 gig. But next-gen products start to happen second half of this year. And then that was in our commentary as well and hence, gives us the positive outlook on cloud titans relative to what we said in 2020.

If I sort of stick to this theme, right? When I think of the next-gen cycle and 200 gig, 400 gig and so on, you said it starts to happen in the back of the year, I think. But maybe you could contrast this versus the 100 gig cycle that we saw. What are the differences? Does the ramp look a little bit more staggered? How does that look for you? Maybe just contrast the 200, 400 gig today versus the 100 gig from a few years back.

This is actually a very good question and often a misconception on Wall Street, that just like 100 gig, 400 gig is coming and the entire world will just switch. 100 gig happened in the 2016, 2017 time frame. At that time, the previous technology was 40 gigabits, and 100 gig was the same cost, the same power consumption and fully backwards compatible to 40 gig. It was a no-brainer. There was no reason to buy older technology when you can buy 100 gig that works pretty much the same. Even if you don't want to use 100 gig, you don't waste any money. 400 gig is not the same. 400 gig on average costs about 2.5x the cost of 100 gig on a per port basis. It consumes at least 2x the power, and backwards compatibility is not as straightforward. You have different types of optics to worry about, different types of breakouts to worry about and so on. So as a result of that, our customers are not deploying 400 gig just for the sake of it. They're actually very thoughtful that they're deploying it only where they absolutely need it. Now you have to think of the customer's architecture. You have compute storage, other clusters underneath. And then you have lots of interconnects for these clusters. Then you have spine leaf for data centers. Then you have data center interconnect. Then you have backbones and edge and the Internet. So when you look at all of that, these customers are going to deploy 400 gig only where they absolutely need the bandwidth, and they choose to apply 100 or 200 gig at other layers, so -- which is why the next gen is a mix of that. It's not a big 400 gig cycle. In fact, in 2021, 100 gig is still growing in market share in number of ports. And it's only next year that it starts to decline as 400 gig picks up, but 400 gig will never replace 100 gig anytime soon. Before that happens, there will be the next gen and the next gen, and there's already talk in the industry about 800 gig and 1.6 terabit Ethernet that comes in the next couple of cycles. But 100 gig just was very unique, rare to see that again.

I see a question in the queue asking about the enterprise opportunity. Arista's comments on this end market have become increasingly positive, I would say. Can you perhaps just touch on what are the drivers behind your recent success here? And related to that, how should we think about the campus opportunity from your perspective?

Sure. Well, the enterprise is a very broad market, as you know, and we've been investing in this space for several years. And not just now but the last 2-plus years, we've had good growth in the enterprise. And every quarter, we've reported that, and it has been the fastest growing vertical for us consistently. Now that's largely been on the enterprise data center. And on top of that, we have been growing in enterprise campus as well. The investments we made in go to market, especially our sales force and better coverage of both sort of Fortune 500, over 2,000 customers as well as other large enterprises, more technical resources in that space to invest with customers help them operationalize technology or some dimensions of it. On other dimensions, we have software products like CloudVision that really allow customers to automate their infrastructure as if they were a cloud player but without needing the 100 software engineers that each of our cloud titans have. So the products are very well received by the enterprise. They love the high quality that we bring to market and are not afraid to use our products. But in the past, there's almost this fear of not using the industry standard products and they would get fired if they tried something else. That's not the case anymore. These customers are able to take our products and use them because they know they are better and in fact, are able to implement their automation strategies or DevOps strategies even in the enterprise. So that is going well. Campus, as you mentioned, is going through several transitions of its own, including as everyone comes back to the office with flexible workspaces. And we fit very well in this next-gen architecture with WiFi 6, with mGig, with 60-watt POE, with MACsec encryption, with zero trust framework for security for these customers as well. And then on top of that, you add the complexity of coming back from COVID and saying, "I want flexible workspaces. Change my architecture completely. Give me segmentation despite everyone working next to each other." And our solutions are able to do that. So I think we are doing okay in campus. It's a small number. Obviously, it's growing fast because it's a small number, but it's a tremendous opportunity for us for several years ahead.

Got it, got it. And while we're remaining on the campus topic, this is an area where Cisco has long dominated. Can you just touch on why you think Arista can gain meaningful traction in this market given, historically, Cisco has been able to fend off competitors pretty well?

Yes. I don't know how many competitors like Arista have shown up in the past. The reason we're able to win customers is when you're able to offer them a differentiated solution -- and differentiation can be on different dimensions. And the cloud speeds was a big deal. In the high-frequency trading space, latency was a big deal. For the enterprise, they care less about speeds and feeds, but they do care about high quality. They care about automation. They care about a solution that doesn't hold them back with something proprietary. And today, the enterprise customers are very open to that, and they know for a fact that Arista solutions are better. They're not going to let them down on quality or technology. So it's not as if they're taking a risk. And in many cases, they're already a customer using our products in something they consider even more critical, which is their data center. So if they're successful in the data center with Arista products and EOS and CloudVision, then campus is an easy adoption for them as well. So we are certainly seeing all of that. And I think the market is large. The market can easily support multiple players. And yes, Cisco might have had a dominating share -- dominant share for many years in the past, but there's plenty of room for us to grow and maintain that balance in this ecosystem as well.

Anshul, if I just keep up with this campus discussion a little bit. I think the commentary Arista has made is it's about a $200 million business this year in calendar '21. I get asked this a fair amount, which is how big can this really get over time for Arista. And maybe there's 2 ways to talk about this. If you could get the right penetration with the existing customers, what does that mean? And beyond that, if you can just scale, how big could this business get for you?

Yes. Amit, this is often asked question. We don't have a direct answer on exactly what that number will be. A lot of people are just looking for that number, what can they put in their model to gauge the growth rate and so on. But a different way of looking at this is campuses -- campus switching and then campus wireless, these are large markets. Campus switching alone is a $12 billion to $13 billion TAM. And even if you make it sort of ex China or ex some of the Asia countries that are hard to penetrate from here for various reasons, there's still roughly at least a $10 billion TAM for us. And we are only talking about $200 million today. We have a great technology, great product. So I think there's ample room for us to grow for many years to come. Now today, it's largely led by a direct go to market, direct sales force, but we are increasing our investments in the channel as well, and you get a different set of customers through those routes. Gradually, that increases too. So I don't know the number yet because it's very, very hard to forecast the model and COVID changed the course quite a bit for everyone. But I think for the next 5 to 10 years, this does seem to be an area of growth for us. And then plenty of room to increase our footprint and advantages there as well.

That's absolutely fair. I have a question from one of the folks in the audience, sort of going back to, I think, a comment you had earlier, which is 400 gig is -- I think you said 2.5x or 2x more expensive than 100 gig. What is the big cost driver that's driving that uplift? And as this ramps up, what does that mean for ASP of 400 gig versus 100?

Yes. I wish I could tell you we want to make more money, hence the prices went up, but it's a very competitive market. So you have to look at technology from its very basic building blocks, and in this case, we have to go back to how many SerDes or I/O connections do you need for 400 gig. And for 100 gig, you needed 4 lanes of 25 gigabits to get 4x25 or 100 gig. For 400 gig, you need 8 lanes of 50 gig. So right there, the I/O doubles. When the I/O doubles, the size of the chip that you want doubles as well. You need more power, more heat sinks, more PCB, more traces. And the entire product ends up costing more to build relative to 100 gig, which is why you end up roughly at a 2.5x cost structure for 400 gig. But the market is not buying just 400 gig. The market is buying 100 gig, 200 gig and 400 gig. And you can take a 400-gig port and break it down, break it out into 4x100 gig. So used in that mode, the ASP of 100 gig actually comes down. So while the ASP of 400 gig is higher than 100 gig, because 100 gig goes down when you blend the 2 together, it roughly comes back to where the market is today or a very little bit of increase, which is modeled in the roughly 9% or 10% CAGR that the industry analysts have for the data center switching market. So it's already factored into that.

Got it. That is really helpful. And then another one from one of the folks in the audience is really around -- going back a little bit to the cloud titans. And I guess the way I'd phrase this question is -- we spent the last 2 years, '19 and '20, probably talking and worrying about the risk of things going more and more to white box, if you may. And I would love to get your perspective on, as technology gets more complicated, it sounds like 400 gets there, 800 may get much more, what's the potential for some of these cloud titans that rely more on web scale -- or white box to come back to Arista? Could that be a real opportunity for you as you think about it? And could we have a cloud titan that's 10% of sales maybe over time?

I wish.

Jayshree would love that, I'm sure.

Yes. Often, these questions result in a new goal for me, given to me by Jayshree. So the way to think about the whole white box scenario is you have to go back and understand why do companies build on their own. And you'll get various answers. In most cases, it's because of flexibility. They want to control their own destiny. They want to have their own schedules. They want to have their own secret sauce and integration internally and so on. 15 years ago, the answer would have been cost. But for the last 10 years, the answer has not been cost. The industry is already competitive. So the real question then to ask is if someone had to start over today, would they still build their own white boxes? And sometimes, the answer you get is surprisingly either a no or maybe. It's not as straightforward as it used to be. So there are companies that talk to us and are considering whether it makes sense for them to buy from the industry rather than build on their own. Now this may not be for every use case. But for certain use cases where they need more functionality than what their internal stack can support and they have SLAs with external customers that they need to honor, they feel they're actually exposed and taking too much risk building everything on their own. However, there's one other factor that you need in order to have a true decision on build versus buy, and that is being objective. Sometimes companies have religion. They have [ a bias ]. So until the management teams are not objective, you can't have an open discussion. In at least one of these cases, we are having a very objective discussion with 1 cloud titan customer, and we'll have to go through the phases with them. It's a discussion on build versus buy. There will be certain places where, if this is successful, they may use our products instead of white boxes or even products from other vendors instead of white boxes. But doing the co-development, doing the pilots, doing the POCs, deploying in their networks, it's going to take several years. So I want to set everyone's expectation. This is not something that's slam dunk and going to happen easily in the next few months. But over the next couple of years, this is very much possible. The reason it's important to note that is because there's so much hype about white boxes. And as an example, last year, the cloud was slow for us. And people assume -- not just assume, white box companies started making comments that that's because they're gaining share. And people thought, "Hey, Arista must be losing share to white boxes," which was not true at all. What was happening was some companies were gaining share from a different white box company, but they were sort of vocal about it and people get confused. So we started to talk about this since last year to remove that confusion in the market. And as we make progress and as we are ready to talk about it publicly, we will. And I think that can be a good outcome for the overall ecosystem.

That's extremely helpful. I think to your point, everyone has thought about the headwinds from white box really, really well, but there's a narrative of how this -- the other way in your favor that is less discussed or less focused on and the probability might be purely attractive there. So I do know it's a good discussion to have. Maybe if I shift away from cloud titans for a bit. On the service provider side, right, in a way I would have always thought very naively that if you can solve the most complex problems with the most demanding customers, the cloud titans, you should be able to scale really well with service providers because I think their bottlenecks are somewhat comparable at the end of the day. But I think the ramp-up in service provider has been a little bit more gradual versus what I would have thought. I would love to know why do you think that's been the case. And how do you see this growth rate stacking up as we go forward?

Yes. That is a good question. We do love as a company to solve the most complex problems in the cloud or other spaces. With cloud titans, the decision-making process is either straightforward or fairly short, meaning 12 months or so. You make some commitments, you make some bets, you build the right technology and in 12 months, you'll start seeing results from the customer, maybe not in revenue but at least you know whether you're getting design wins or not. But service providers, those cycles are longer. And as a company, we are investing in the service provider space. It's not as if we're ignoring them. But the question is can we do a lot more. And it's a measured investment simply because they still need a lot of the legacy features in addition to new technology. Whereas with the cloud, you can have the discussion about, "Hey, what if I just skipped the legacy. Here's a new architecture. Can you go deploy it?" They might have to make some compromises, but they will and try to gain an advantage. With service providers, they have a business model where they're interconnected to everything in the past so they can't just get up and move to a new greenfield architecture, which is why it takes longer to meet all of the requirements. Having said that, we are deployed in the telco cloud, which are the NFV data centers. We are deployed in a lot of the sort of cable-style networks as routers for their backbones. And we're also deployed in some of the telcos for more peering or edge use cases and getting more now to ready for backhaul-style networks for them, too. But it's a gradual pace. So the feature list that they need is too long, and there's not a single requirement they'll ever drop. They'll only add more to it. So we'll get there.

Great. One of the other topics that has been more sort of top of mind is really around silicon diversity, right? And maybe it's Cisco that's talking about it. But yes, I do think vendors or OEMs like an Amazon and Microsoft, some of the cloud titan that really talked about a need for more silicon diversity. How does Arista think about silicon diversity and its implication to you?

Sure. Well, ever since the company got started, we've championed merchant silicon. We don't like these closed stacks where you don't have openness of how the hardware is actually working. And fast-forward 10-plus years, here we are with merchant silicon having 80%, 85% market share, and the others are now crying wolf and saying there should be diversity when they are the ones who create the problem in the first place with having proprietary APIs and locked-in ecosystems. Now customers would love diversity and so will we. And we have had choices from Broadcom, from Intel. In the past, we've carried Xpliant, which became Marvell. We had Falcon, which was also acquired by Intel. So we've actually done quite a few things in this space to keep a broad open ecosystem. And while there is a need for diversity simply to have competition in this space and so on, you also need options that actually are better from a competitive angle. That's where people often struggle. And can you take on the entire merchant silicon space and build something better with the new start-up? That's rare. But we keep our eyes out. And if any good technology shows up, we absolutely invest in that. And if good choices show up or our customer leaders, they will build those products, too, we're actually fairly happy overall with the ecosystem.

I see a question here on one of Arista's key adjacencies, which is security. How do you think about security being an integral offering for customers? I understand Arista provides network detection and response security capabilities via the Awake acquisition while, at the same time, you are partnered with leading security vendors, including Palo Alto Networks, VMware and Zscaler, among others. So which areas of the security market will you have a presence in? And when does it make sense for customers to select a security partner?

Yes. Security is a very, very broad space, right? While it's all clubbed under security, security means firewalls, it means Internet edge it means endpoint security, it means encryption, it means connecting to the cloud and includes the SASE solutions like Zscaler. So you have so many choices out there that you cannot get a good complete solution from any one vendor. For Arista, we participate in the best-of-breed stack for our customers. So it makes sense for us to partner with best-of-breed companies and bring those solutions to the market for our customers. We did this in the data center space with macro segmentation. We are doing this for our campus customers with MSSG, which is a way of doing group segmentation. As well, again, we integrate and partner with many companies here. In addition to that, we recently launched our zero trust framework that is critical not just for customers to think about for their security teams but also for the networking teams because many of the attacks are going through the network, and you've seen so much happening. Every day or every week, it's headline news on a new breach and a new ransomware attack and a new lockdown that you have to go fight. So we are working very hard actually to mitigate as much as possible from the network and -- like the Awake solution you mentioned, I'll touch on that as well, but then partner with the ecosystem for the rest. Now some of this space is evolving, and NDR or network detection and response is a relatively new space. This didn't exist 10 years ago. But now there's so much telemetry, so much information inside the packets that are going to the network that you have to leverage that as well to give you one more angle to protect your enterprise from attacks. And especially because a lot of the attacks are using encrypted traffic, so the legacy tools won't actually work to detect or prevent against these attacks, whereas with Awake Nucleus, which is the AI engine we have, we can already do that. So we'll continue to increase our footprint here and partner simply because of the complexity of this ecosystem.

Got it. I know we're coming up on our time. We got about maybe 5 minutes left. So maybe a financial question on your margin structure. As Arista is presented with new growth vectors, especially in the enterprise end market, how do you think about the investments for growth required to drive scale of the faster-growing opportunities? And the potential impact to your operating margins?

Yes. Look, we are quite focused on growth as a company, and we continue to invest. As I mentioned earlier as well, we're increasing our sales and marketing footprint. We've added significant head count consistently for the last 3 years, and we'll continue to do that. We're increasing our investments in channels. We're increasing our R&D investments as well, both for new products as well as existing as you get to next gen. So as a result of that, if we can get to higher or better growth, we will just go ahead and make those investments. We are less tied directly to our operating model. The operating model is the result of everything else that we are doing. And it's a healthy space, roughly operating at the 37% operating margin today. But it's a good balance of the cloud is probably lower gross margin but lower cost of sale, whereas the enterprise is higher gross margin but also higher cost of sales. When you blend those 2 with roughly the same type of R&D, you end up at the operating margin we are at. So we're not expecting that to change significantly. But as we look at more adjacencies, if such opportunities did come along, then we'll absolutely have an open discussion with all of Q2.

Okay. I think before we sign off, I do want to sneak one more in for you. I think Arista has been really selective when it comes to acquisitions, but you're certainly seeing an uptick in the number of deals you're doing like Big Switch or Awake that you talked about. Maybe help us understand how do you think about your M&A strategy. And how does that fit into the Arista framework as you go forward?

Yes. M&A is a very sort of heavily debated topic inside the company for the last several years because you want to do it but you want to wait for the right reasons, right? So we're not doing M&A for our core business. We know how to build switches. We know how to build EOS. We know how to build our hardware. So that's not what we're trying to get from the outside. M&A is for adjacencies where we don't have the skill set or the go-to-market presence that we need. We want to acquire companies that have good talent, that have good teams and will culturally blend into Arista. If you have a cultural mismatch, then it's a wasted effort and a lot of time that goes in an integration that will not see the results. So we are very careful about that part. And the last part is valuations. And if you told me last year or 2 years ago that the valuations would be where they are today, I wouldn't have believed you. The valuations have only been going in one direction in the last 5, 10 years. So we watch for that, right? We don't want to make a mistake and just throw and waste our money. We'll be opportunistic about it. And if the right thing comes at the right valuation, we'll absolutely look at it. So that's why we've been selective. But we continue to grow on our own with the core business we have and the adjacencies, and the rest is opportunistic.

Perfect. I have about 90 seconds left here. So maybe I'll pause here, Anshul, turn this back to you. If there's something I haven't touched on and we haven't covered that you think I should, investors should be aware about, please just flag that for us. I'll turn the mic back to you.

No, Amit. Thank you for all of the discussion. You've covered all of the key topics, everything that's been on investors' mind. So thank you for the opportunity and the time.

Perfect. Thanks a lot for this. This was really informative and helpful for me. So thanks a lot.

Great.