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

Good morning, everyone. I'm Jason Ader, and welcome to our virtual fireside chat with Arista Networks' President, Anshul Sadana. Anshul, thanks for being with us today.

Thanks, Jason.

Before we begin, I'm required to inform you that a complete list of research disclosures or potential conflicts of interest is detailed on our website at williamblair.com. Also, I think there should be a question box in the Zoom app, please send questions during the fireside chat because there's a delay, and I want to be able to get to your questions immediately after my questions to Anshul. With that out of the way, let's dive into questions.

Anshul, Arista has achieved a lot of success over the last decade with the business now north of $2 billion in revenue. What should give investors confidence that you guys can continue to achieve the same level of success in the current decade?

Sure. Jason, we've come a long way in the last decade in building a product, a company, a team and a customer base in what is indeed a large TAM. And in the history of networking, no other company has managed to do this against the primary competitors who have existed much longer than us. When you look forward, the growth factors for us at Arista don't change that much. It is still a very large TAM. We haven't yet penetrated it in any very significant number. There's a lot of growth left and a lot of opportunity left. This includes data center switching, this includes routing, this includes campus. When you add the 3 up, depending on how you look at it, this is somewhere between a $25 billion to $32 billion, $33 billion TAM. And every year, we keep on increasing our product offering and go-to-market coverage to try and gain more and more access to that TAM. The cloud journey for us has been tremendous. 10 years ago, I don't think any of us expected this is how large the cloud will become. But at the same time, I don't think the cloud growth is done. The cloud companies will continue to grow bigger and bigger, and there are various estimates out there in the analyst community. And some people think the cloud is only 10% or 20% built out compared to what it needs to be to take on all the workloads that could move to the cloud. There's tremendous growth left for the next 10 years just in that space. On -- In enterprise, where we have been growing as well, we haven't reached saturation. It's very early days of penetration, proving the product, proving our quality and execution and continuing to win over there as well. And now we're in a position where customers are coming to us, where they're no longer worried about Arista being the start-up and the small company, will it work. The technology works, they just know it. They've heard enough good things from their peers as well, but we have to keep executing. So I think the growth factors are still there. It's a tremendous opportunity. I'm very proud of what we've achieved so far. And I'm also actually very anxious and excited about what we can expect in the next decade.

On the cloud side of your business, it's clear that widespread work-from-home is driving up consumption. I think a lot of investors feel like Arista should be a direct beneficiary of this. Is this the right line of thinking? And if so, how come we haven't seen more positive impact on your results and guidance in the first quarter?

Sure. So when you look at the work-from-home trends, there's audio, there's video involved. It's generally communication that has moved or grown tremendously in the last 3 months. And when you look at network capacity needed for video or any of these conferences, the edge for the cloud needs to grow. I mentioned this over the earnings call as well. And there's lots of statistics on how traffic quadrupled in a region or traffic tripled in a region, and how everyone needs to keep growing and so on. You're measuring bandwidth in terabits per second. And when someone says traffic has tripled, maybe for a certain region, the total video traffic on video conferencing and so on went from 1 terabit to 3 terabits. That means that some cloud companies need to add 2 more switches to each of the regions because each switch is a terabit or more in bandwidth. So the growth at the edge is good for us. No use case is bad. But the edge or the Internet sort of traffic workloads are much smaller in size compared to what happens inside the data center. We've all heard about East-West traffic and how 95% to 99% of the traffic is inside the data center, only one person goes to the Internet. We're seeing that tremendous growth in that one-person dimension. And yes, we see the benefit, but it's not that significant to move the needle the way people would expect. It doesn't mean cloud growth is not good for us. If workloads are moving to the cloud, that actually does help us because the cloud grows. But I think that's a long-term trend that many companies haven't yet decided on really comes back to how the lockdowns end and things like that.

Okay. Your largest supplier is Broadcom. I think everyone knows that. They reported last week, they highlighted strong demand for some of their main chips like Tomahawk, Trident and Jericho. They also talked about cloud providers not only seeing a current demand lift, but also building for next-generation services. How should we interpret this with respect to Arista's cloud business over the next couple of years?

Yes. So there are 2 dimensions there. One is the overall business. The other is the growth percentage on our cloud business. And as a business by itself, our cloud business is tremendous, in my view. We're doing very well. The execution is rock solid, customers love us. And we are only winning more and more designs or use cases with them every year. But we also have a fairly large share and a very sizable base there. So now it comes down to, is the base growing significantly? If it is, I think we'll do well. But when the numbers become so large, the growth rate comes down. And this is largely what we have seen in the last year or so as well as in our guidance. So I'm not at all concerned about us doing well with the cloud. But just given the sheer size of that number, in certain segments like 100-gig or 100-gig modular switches, which are the spine switches, we do very well, exceptionally well, I would say, in the cloud. When those use cases are growing, I think this will be fine for us as well. To a great extent, this part is already modeled into our forecast. So last year, we talked about Facebook as a customer and how things may be slowing down and so on. And it took a while for the market to catch up to what we were seeing and then things added up, and it all makes sense. So we do believe on a project standpoint, we have that close connection with customers, and I don't think we're as worried about a significant surprise here. Now if the cloud increased the number of servers significantly, I think that will benefit us. But it's still to be seen whether the current cycle is a pull forward of demand because of supply and so on or is it simply an actual increase in their build out.

Okay. You talked about customers loving your products and the success you've had over the last decade. One of the questions I get a lot on Arista is really what is the differentiation? Isn't a switch a switch -- switch is a switch? What are you guys doing different? And how durable is that differentiation?

Yes. So if all switches weren't equal, we wouldn't exist. There were many choices in the market before Arista came here. A switch from the outside, looks like a simple sort of device that has 32 ports of 100-gig. It has some CPU in there. It's still running a Linux-based software. And the chips are built by Broadcom, the CPU is built by Intel or AMD, what's the big deal? What you have to realize is that to build this switch, you need some extremely complex signal integrity and PCP design and so on. And that design and the strength of it, the advantages that come with it, on the cleanness of the signal, the length of the cables you can connect to the server. Does the link actually work at every possible speed and so on, on the hardware side in a power-efficient manner, does matter. But then on the software side, the EOS software is over 15 million lines of code that ships inside our product. It's -- sometimes a software, there's a lot of work that goes in, not just to write the software, but to actually test it and sustain it through the life of the product. And that's where the advantages show. Customers do give us feedback from a quality standpoint. They do believe that our quality is 3 to 5x better than any alternative that they have seen, whether it's other large companies or a white box ecosystem and so on. So these cloud companies are always looking for cost savings, but they also have another goal in mind, which is their primary goal that the network needs to work. And it needs to integrate in their automation system and we do very well over there. So the advantages are quite sustainable in terms of the complexity of the code that we have, the design and the quality. And by the way, going forward, as the speed increases when we go from 100-gig devices to 400-gig and then 800-gig in the future, that complexity is only increasing. It's not going to go down and cloud companies are actually coming more and more to us to help integrate or test a full solution with them. So which is why I'm not as worried as people think -- seem to think. These are equal devices that can be replaced one for the other.

When you say quality, what would be examples of, I guess, low-quality from a...

An outage?

Just outages.

So outages, number of security vulnerabilities, number of times you have to upgrade your switch, so to patch some bug fix or some security vulnerability and so on. All of that results in downtime to the customer. The SLA the cloud companies have to achieve are very different than the enterprise. So to give you an example, the total downtime budget of any cloud system is about 0.5% per year. That's roughly about 4 hours worth of downtime in 1 year. And you have to -- that's not just the switch, by the way, that's the server, the applications, the net, the load balance for the security systems, the data center networks, the WAN networks, the Internet providers like Equinix and so on, everything combined has to work really well. So total budget available for a switch inside the data center to go down, on average, is less than 1 minute per year. It cannot have more than that down -- more than 1 minute of downtime in 1 year. And typically, you need about 3 minutes to reboot, which means you can only touch the device once every 3 years, sometimes once every 5 years. If you miss that SLA, the cloud companies will lose out on their uptime or they have to pay penalties on customer contracts and so on. And the Arista products, the hardware and the software help them a lot in achieving those goals. Again, this is where I think people think about competition being just because someone has a similar looking hardware piece. It's not the same. You have to achieve the same quality and integration as well.

Okay. On the competitive front, what do you think your market share, I guess, is in the cloud vertical today? And then I know there's a fear that some of your larger customers like Microsoft, Facebook, will do more white box and that guys like Cisco and Juniper are kind of touting some share gains in the cloud vertical. And I was hoping you could comment on that.

Sure. So when you look at the cloud overall, let's just focus on the U.S. cloud companies for this discussion, which is where our primary business comes from in the cloud space. The cloud companies have been multi-vendor for a long time. And they will always stay that way and change allocations a little bit here and there to keep everyone in check on prices and things like that. But the feedback we get is that they just love our execution and our products and the engineering team, and the way we can customize the product for them. So we are not hearing or sensing what some of our competitors are claiming. We don't believe we're simply going to get designed out. What is happening is they might have been a third or fourth vendor whose share is now being given to some other company. And these other companies may not have gone on any business from these customers for the last few years. And they somehow think it's a major win for them and so on, but we actually don't see any change on our side. So that, I think, is just normal business. That is how life has been for many years, and it will continue this way in the multi-vendor ecosystem. On the white box comment, the work we did with Facebook is a perfect example of how we view the opportunity vis-à-vis white boxes. The problem Facebook was trying to solve was that, number one, they do like to build things on their own, to integrate better, to get that learning curve and so on. It's easier for them to do that near the server for some of the use cases, which are sort of easy for them to build and test for and so on. But it's more complex when you try to do this at the router layers or when you're doing for Internet peering and so on because there's a lot more development that's needed. But as they start that journey, they realize that if they start to customize too much, they end up being locked into their own stack. And they become single vendor. Yes, that vendor happens to be inside the company as far as their customer base is concerned, but they are -- they have the same risks that everyone else has being single vendor. So they came to us and said, let's solve this problem together, and we worked on this co-development effort. Facebook calls it the Minipack, we call it the Arista 7368. And the products are quite interchangeable, and that co-development has been a good success. The outcomes are very much in line with what we expected, and Facebook was extremely happy with the outcome as well. So now going forward, I think there will be many more of these that will be happening. And this co-development and an ecosystem of build and buy, not build versus buy, will actually continue with these cloud companies. So if you look back from 2013, 2014 onwards, this hasn't changed as much, by the way. They have been doing it for quite some time, and they do view us as one of their most significant partners in this journey even in the future. So yes, there's a lot of talk about these companies trying to go to white boxes, doing things on their own. But I have to remind everyone, companies that are talking about white boxes for many years now are our largest customers. And often, the problems they're referring to of not being happy with a certain vendor or not being happy with the results is largely referring to the state of the industry for the 2 decades before Arista was deeply involved. In the context of Arista, the discussion is quite different. It is one of co-development and integration, not one versus the other.

I guess does that mean that white box -- if you think about the whole pie for a Facebook or a Microsoft, does that mean white box takes more of that pie over time?

In my view, it remains stable. It doesn't change. It is at a place which is healthy for them and for us. And I think it just stays at that level.

Okay. Let's turn to the 400-gig transition. I know that's a hot topic. What are the big customers saying to you in terms of timing now? I think you mentioned on the earnings call that the time line might have gotten delayed a bit. Can you just talk us through that? And then where you see your market share landing with 400-gig relative to 100-gig?

Yes. So before I answer that question, I do want to mention there's almost this misconception in the investor community that this 400-gig market and very similar to 100-gig, on one day, sometime in the near future, the market will flip. And there's this race to grab market share in 400-gig. What happened in 100-gig was quite different, where the entire market switched from 40-gig or 10-gig to 100-gig in a very quick, compact window and compressed window. And as a result of that, there was a race and whoever could execute better and so on won. In 400-gig, the technology is a little bit more complex. And there isn't a big driver for every company out there to go from 100-gig to 400-gig. Some companies are, in fact, increasing their footprint in calendar 2020 and 2021 and growing 100-gig, not switching to 400. 400-gig will have several variations. One is built on the 8 x 50-gig SerDes 8 x 50-gig PAM4, which is essentially the technology shipping now. This will quickly evolve also into 4 x 100-gig. At the same time, the 100-gig ecosystem is going from 4/25 to 2/15 in some cases to match the breakouts and sort of going end-to-end from the 400-gig port. And at the same time the servers also have to move to a 50-gig PAM4 to avoid gearboxes and extra power and cost inside these data centers. So there's a whole ecosystem transition that needs to happen. And you ask why? The assumption a lot of people make is everyone needs that bandwidth. That's not true. 100-gig to the server is actually at quite a bit as well for most use cases. So which is why there is no single 400-gig wave coming. It will come in layers once -- well, first for the DCI or the data center interconnect layer, then maybe for some customers for a specific cluster or 2, some certain architectures inside the data center and so on. So having said all of that, 400-gig for certain cloud companies will happen but it's happening in a staggered manner. It is largely constrained by availability of optics of every type. For data center interconnect, you need the ZR optics that can go 100 kilometers. For inside the data center, you need DR4 and FR4, which is still in its early stages, it is ramping on yields and availability and so on. So this will happen through 2020 and 2021, in my view. The actual transition to 400-gig, I think, starts to happen in 2021, maybe even mid-2021 by many of these cloud companies. The pilots are starting to happen either now or sometime this year, just to test out the technology and develop their tooling and automation and monitoring and so on. So it is a very slow-moving market compared to what's happening in 100-gig. But remember, 100-gig is not done. It's actually growing.

Okay. Let's switch gears to the campus business. Can you talk about how COVID potentially affecting the ramp here? And what is the, I guess, value proposition of your products versus the incumbents in the space?

Sure. Well, as for the near term, every company has been focused on just keeping their business running, how do they get the lights on despite the lockdowns. And can the IT staff come on site, can they come on-site and so on. We are just now entering a phase where people are relaxing all of the rules. And as long as companies can implement their own policies on social distancing, the IT staff is starting to come back on site. Last 90 days has largely been a lot of virtual events for us with our customers, whether it's virtual meetings, virtual webinars, virtual training sessions and so on, and they have been very well received. Now in the short term -- very short term, next 6 months or so, I think the campus market has slowed down simply because there are very few employees coming to the office, and many people are cautious. No one wants to rush back to the office until everyone is sure the virus has gone and there's a vaccine and so on. Has the long-term trend changed? I don't know. I believe the opportunity still exists. It just got right shifted by 6 months or so. And will the companies be able to run their network and never be able to refresh and so on? Not true. They will still have to do the upgrades they were planning on, just that they will take a little bit longer.

And on differentiation?

I think differentiation is already very well received. People often look at the products as the hardware because that's what you can touch and feel. But CloudVision has a way to automate and monitor your campus networks and manage it all from one place. Campus and data center has been a great asset for us. We continue to invest in it. And the feedback from customers is in -- on the enterprise side is unanimous, that this is years ahead of what our competition has built so far. EOS is the same EOS operationally on both places. There are a thousand reasons EOS is a good operating system and each one of them applies to the campus as well, whether it's the openness, the Linux part of it, the monitoring part of it, in SNMP you can do real-time streaming. The integration into customer security systems and NAC and so on. So all that already applies. But on top of that CloudVision is helping tremendously to differentiate and gain advantage there as well. So we're very happy with the product side of things. Actually, that -- I don't think that is a gap or an issue. We'll keep on expanding our portfolio over time, that happens naturally. But for the near term, things slow down a little bit because enterprises were mostly working from home. But I think now as that trend starts to change, things come back to where what we expected in the past.

Okay. Great. And then just on the kind of current environment, we've -- we're 2 months, almost 2.5 months into the quarter. Have you seen any changes in the demand or supply situation that you can talk about since earnings?

Sure. The -- I won't comment much on demand. I'm not here to guide on Q2 at this time. But on the supply side, things are stable. The way to look at this is many of the countries and suppliers in Asia are -- have already come out of their lockdowns. They have restrictions in place, social distancing. Some companies are operating at 50% capacity, some operating at 70% or 80% staffing and so on. But we did have -- we missed -- these companies missed roughly 4 to 6 weeks of production in calendar Q2 or starting late March, and that needs to be caught up with in addition to keeping up with normal demand. So when you spread that 6-week, 8-week gap over the rest of the year, I think things are stable. But they just have to catch up to the pent-up demand they have. My expectation is the supply -- commodity supply will remain somewhat tight through this calendar year as they continue with their build-outs and so on. And then by the end of the year or early next year is when things start to improve and recover back to normal. Obviously, this assumes there's no other second wave and things like that, things just stay where they are today.

Okay. A question from the audience here. It's on Big Switch. Does Big Switch change the one OS advantage you've counted over the years? If not, does this imply that having a unified OS is not as important as we might have believed?

So we've mentioned this in several places now, but even on the Big Switch side over some period of time, it's the BMF, which we call DMF, the DANZ Monitoring Fabric, is indeed getting integrated with Arista EOS as well. So customers will be able to come back to the single OS. It's only a matter of integration and time and that happens whenever you integrate 2 companies together. In fact, there's advantages because Arista DANZ, which allows you to do traffic steering and filtering and so on, on the network layer, coupled with the monitoring nodes that Big Switch has built like the packet recorder and the [ dedup ] engine and so on, acts -- integrate very well together and provide a great monitoring solution to customers from a security standpoint. So that is already happening. And this was all kept in mind when we acquired Big Switch, and I think the single OS does come back on the switching layer. On the monitoring controller, what they have built is unique IP, we never had that. That, in fact, should continue to stay as it is today stand-alone, and we'll keep on continuing to improve it. And integrated with CloudVision from ease of management standpoint, but I think the Big Switch IP and the controller and the monitoring nodes, the packet recorders will continue to evolve. -

Okay. Great. And one more question from the audience. Does Arista plan on jumping into product lines other than routing and switching down the road?

Don't you think our plate is full? Yes, we have a lot to do. As I mentioned, it is a very large TAM, right? From one standpoint, yes, we've come from very small numbers to $2 billion in the last decade. But there's a long path ahead. So I think our plate is full. We need to continue to sort of execute in this space and not get distracted by taking on a lot more right now.

Okay. And looking out longer term, which areas of the business are you most excited about? And where is the most upside potential?

Yes. So I'm actually excited about every aspect of our business. I still believe that we have tremendous growth left over here. The cloud journey is not done. It will keep on going on, both with Tier 1 cloud and even the specialty cloud, as these companies continue to change the business and evolve. Service providers and routing, there's still an opportunity ahead of us. And it probably didn't happen or materialize the way we thought it would, but it doesn't mean the opportunity is behind us completely. I think it's still ahead. And then enterprise and financials, we still have tremendous growth, both on data center as well as campus. So when you add these together, all of these dimensions of growth are important. And when 1 or 2 engines are kicking in parallel, you'll see really good results.

Okay. See if there's any more questions from the audience. It doesn't look like there's more questions from the audience. Let me ask you to leave us with any final thoughts on Arista, on the business and on the outlook?

Sure. Well, great teams that work together and execute well are of tremendous value, and that doesn't disappear in a short moment. We've done very well last decade or so. I believe we'll continue to execute. The execution is only getting better inside the company and with our customers. And the opportunity is still there very much in front of us. So I'm very, very excited with the position we are at and kind of to grow and sort of succeed from here on. Yes, there's always competition, but that doesn't worry us.

Okay. All right. With that, we'll wrap up. And thanks, Anshul, for spending some time with us. And good luck to Arista going forward and stay safe.

Great. Thank you. You too.

Bye-bye.

Bye.