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

Yes. Thank you. Thank you, everybody, for joining us this afternoon. Extremely excited to host a quick discussion with Arista's COO, Anshul Sadana. Anshul, thank you so much for letting us host this discussion here at the Wells Fargo Tech Summit.

Absolutely. Thanks, Aaron, for having me.

Well, great. Well, we'll just -- 30 minutes is going to go by really quick. And so I want to just dive right in. We often talk about Arista in the context of the Cloud Titans, right? The Cloud Titans being 35 to [ 40 ]. But I think what started to really resonate with investors here in the last quarter or 2 has been just the enterprise traction is becoming a very notable part of the growth story for the company. So can you talk a little bit about what Arista is seeing as opening up those opportunities for you in the enterprise market, and any kind of things you can share around where you exactly are seeing some of the success in the enterprise space?

Right. Well, Aaron, as you all know that we now have approximately 7,000 customers and all. And majority of them are actually from the enterprise space. And we're continuing to do very well there. Enterprise has been our fastest-growing segment now for several quarters in a row. And the customer reception has been overwhelmingly positive as well. When you look at the enterprise market, the needs are actually not that different from the larger customers we have, whether it's the Cloud Titans or any other group because the enterprise also needs infrastructure. They also need infrastructure that stays up. And they also need to automate and do more with less stuff. And we're able to fulfill all of those needs and provide a better alternative than any other option in the industry.

Right. Right. And can you talk about if it's architecturally, what's driving some of that engagement? The move from, obviously, towards 100-gig, would you guys, obviously, were very successful in taking a leadership position, we talk a lot about, and I'll talk a little bit later and ask about 400-gig, but are there architectural shifts that are happening at the enterprise a bigger enterprises, like Wells Fargo, for example, we just heard from our IT guys talking about going and completing their leaf-spine architectural approach. Is that inertia just starting to kind of resonate more broadly from an architectural perspective?

Right. Well, we've been bringing this way to the enterprise for quite a few years now. And the enterprise does need a more resilient architecture, maybe not at the same scale as the large cloud. But at a smaller scale, you need the same benefits. So we do implement our leaf/spine designs. But in addition to having simply sort of 4-way or 8-way equal cost routing, you have 8-way redundancy, we also offer segmentations to our solutions with EVPN and VXLAN, which helps with security needs within the enterprise, which is not easy, right? Because in the cloud titan space, you control the entire stack. In the enterprise space, you don't control the apps directly from a networking standpoint. So we're able to fulfill that gap. This -- there are other aspects of having a higher quality software stack in this network where you can do easier upgrades. You can -- number one, you have very clear security vulnerabilities. And when you do have them, it's very, very easy to patch quickly and not have exposure or downtime like you do with the incumbent vendors. We also have automation offering here with things like CloudVision. CloudVision is our automation suite that allows enterprises to deploy and manage all the way through decommissioning of the product 5 to 7 to 10 years later and not miss a beat and stay fully automated. As an example with CloudVision, you get an update every day about new changes of bugs or vulnerabilities or attacks that you should be aware of and whether there are patches available for you to adopt and make your network more secure. We do all of that for the customer. It's right there on the dashboard. And as a result of that, enterprise customers really appreciate that, right? There's no lock in on the network from a protocol standpoint. It's all open protocols. But in addition to that, you get the benefits of a cloud-like architecture without requiring a large team. And that's where the adoption has been good.

Right. Right. Kind of same question, similar question, but on a service provider and Tier 2 cloud vertical. Again, that's mid-20%, high 20% contribution to total revenue. What do you see happening in that vertical of business for Arista?

Sure. Yes. Both of these segments are back to growth mode, which is good for us and for the investor community. And on the service provider side, we went through [ dip ], but we've been adding more and more features in EOS, and our platforms, including the new [ 7200R ] 7280E and 7800R3 series, which is the Jericho2 platforms, now have enough capability to take on more and more of the router roles. So we're doing well over there and growing. More focused, I would say, some of the recent wins have been Tier 2 and Tier 3 service providers, but we are starting to win more and more across a broader customer base, a more diverse customer base, including international regions. And I think that is certainly helping. It will continue for many years, in my view. We will keep on adding these features in every 6 months or so, the product gets better and better for these customers.

Right. And kind of completing the vertical context. Again, we've seen the company executes through some slowdown in the cloud titan side over the past 12, 18 months. But it sounds like we're going to get back to degrees of growth in the cloud. So how do I think about the cloud titan vertical as kind of the growth profile of that as we start to move forward?

Sure. Well, we've done very well with the Cloud Titans. And I believe we continue to do well. We're executing well with them. They did slow down a bit last year. Sorry, this year, in 2020. And going forward, we believe growth is back with many of the next-gen products with the Cloud Titans. And in addition to that, the other verticals are all growing as well, which is what allowed us to reinforce the consensus growth view for our projections for 2021. And I believe, because it is a large market, and we are performing very well on product, on quality, on supporting these customers, we have this growth opportunity ahead of us for several years to come.

And I think Jayshree said, given what you're involved with, I mean, server cycle -- you got cloud demand, you've got server cycles kind of set up into next year. Just remind us again how important server refresh cycles are before that cloud titan vertical. And how do I think about that pull-through or drag effect, more pull-through effect for deployment of network infrastructure?

Right. Absolutely. Some of the slowdown was because the server cycle got delayed with a few customers. But as these customers migrate to next-gen CPUs with more advanced mix and so on, you are going to have a constant need for more IO, which means next-gen products on the network side and forces the refresh cycle. So we are starting to see some of this in the plans and the roadmap from our customers. And we believe that will certainly help with the growth in 2021.

That's great. On the technology side, kind of shifting gears here a little bit. I mean, every call you get on, every conversation you probably have with an investor today is going to bring up 400-gig, if not #1, probably #2 or 3 on the list questions. So I'll have to ask that question. So 400-gig, I think it's important to note that it's just not all about 400-gig, right? You deploy switches that are 400-gig-ready today, right, in these infrastructure. So just give us your current updated thoughts on 400-gig. What we should be looking at is really starting to see that become a growth driver as we move into '21 and certainly beyond?

Yes. So there's this view out there that because 400-gig technology is available, everyone will rush to adopt it. And that was certainly the case in the 100-gig cycle. But that's not the case in the 400-gig cycle. And the reason these are different is because 400-gig is more expensive on a per port basis than 100-gig. So it consumes more power. So only customers that absolutely need 400-gig are going to go there first. And this will happen in maybe the DCI layer, the long-haul optics and so on. But the qualification and testing for next-gen architectures is coming along well, including with the Tier 1 Cloud Titans, including some of the Tier 2 cloud specialty providers as well. And I think this will be a gradual progression through 2021, even 2022 as the world keeps on adopting more 400-gig. Having said that, we are quite ready for it. We have more than 20 variance of products with 400-gig already shipping now in production, all driven by our larger customers, whether it's in the cloud space or the financials or service providers and so on. And I think are well poised to capture or benefit from that transition as well.

And just as -- I mean the competitive landscape that some of your competitors would argue that they've kind of closed the gap. If it's density, the metrics that they might look at that says, hey, we've caught up with Arista on the 400-gig. How do you respond to that? How do you respond to the competitive setup of 400 versus maybe comparing that to what was clearly a very successful competitive positioning, leadership position for, I think it was a couple of quarters if not more, at the 100-gig cycle?

Yes, there's a lot more than simply having the right number of ports. But let me remind you that our competition said the exact same thing about having caught up with us on 10-gig in 2011, 40-gig in 2013, 100-gig in 2016 and now 400-gig in 2019, 2020. So I don't think anything has changed. It's not like our competition forgot to build switches in the past. If you know this industry, they've been ahead of building products for 2 decades. I would be surprised if they forget to build a product. But it's more about the full solution, including the software stack, including the features, including stability, including the customization that some of these customers need. And I think that's where we shine a lot more. We'll continue to do well. Customers are giving us very positive feedback. So while they will, they might be rendered, I think Arista will retain the largest share here, and I think that's already part of our projections and expectations going forward.

That's great. Shifting gears to the campus strategy, expected to see another doubling of the revenue contribution by the end of calendar '21 as outlined in this most recent call. You recently introduced the 750 Series modular Power over Ethernet series switches. Can you talk about where we stand on deepening the campus strategy? How much of the addressable market do you think you play into today? And is there more we should be thinking about as far as deepening that strategy going forward?

Right. Well, many years ago, around 2008, we started our data center journey with just 3 SKUs. We had 124-port switch and 248-port switches. And since then, we've expanded our portfolio, we have over 75 variants of the data center products now for switching and routing. In campus, we don't need that many variants because we leverage our existing products as well for campus, spine and so on. And with the addition of the 720 Series, the 750 Series, the modular switch and the WiFi 6, we are -- we now have basic coverage for almost all of the campus use cases, especially in the enterprise space. But every 2 or 3 years, we'll keep on working on the next-gen, refreshing, expanding the portfolio. So that journey is a natural course we have to be on, and you'll see natural progression over there as well. But for large enterprises, I think we can already address a big part of the use cases, maybe some of the far end sides of that bell curve, we still have to tackle in the future as we expand our portfolio.

And can you talk a little bit about how the drivers of campus refresh, what you're seeing, where you think we stand today? Do you think that, that starts to accelerate as we move into '21? Just there's -- WiFi 6, multi-gig, et cetera, that's happening. Talk about some of the technology drivers on the campus side.

Yes. So this is actually very interesting, where people thought that campus might slow down. But campus is actually doing okay because you have not just users, but lots of other devices, sensors, cameras, badge readers, elevator controllers, whatnot, inside your infrastructure. And we recently had worked with our customer where for 3,000 employees in their building, in the total buildings, they needed 12,000 ports. Not because they have that many phones and laptops to worry about, but they have that much sort of IoT infrastructure to worry about in the new build-outs. So there's a lot more automation and so on people are doing. When you look at cameras, when you look at new WiFi 6 technology and you aggregate a lot of it, you start running out of bandwidth fairly quickly. WiFi 6 absolutely saturates 1-gig, you do need more than that, which is why you need 10-gig. Many of the automation devices out there, lighting control infrastructure, some of the outdoor cameras and so on required 60 watts or even more of power. And hence, you need mGig and more PoE power. And as a result of that, you are forced into a refresh cycle, if you haven't already started thinking about that as a company, and many companies have started on that journey. Lastly, with all of the video conferencing now we are used to, if you want to do HD video conferencing from several users or all the users in your environment all the time, then the legacy or subscriber infrastructure is not going to work, and you need to add more bandwidth and capabilities there as well. And in addition to that, you have to enhance it with more security functions. So all of that added together is what's enabling the campus refresh and allowing us to participate very, very well. EOS and CloudVision shine in that environment too, not just data centers, and in fact, many companies go to Arista because of CloudVision, because of the benefits of EOS, resiliency, upgrade, simplicity that they get in the campus.

The campus insertion, you get engaged with a customer, an enterprise customer, have you seen campus engagements result itself backwards? Or I shouldn't say backwards, but result in a pull-through effect for the data center business. Are you seeing those enterprise customers? I know, again, we're still early innings of ramping the campus. But as that ramps, is there a pull-through effect that you see just more engagement on the data center switch layer side because of CloudVision, because of that ability to manage from core to the campus and edge with Arista?

Right. Right. We are certainly seeing engagements with customers. These are brand-new customers to Arista. They've been friendly to us, but for the first time, started buying our products in the campus. And their feedback to us was what has been. But, look, our data centers were small. So trying to engage with Arista wasn't strategic enough because the bigger spend was in the campus, and the rest is in the cloud. And now with our campus offering, they are very happy to engage with us and essentially say, okay, you have the entire state, please go for it and make it work. So we are certainly seeing that. It's not backwards at all for companies where the bigger investment is in campus infrastructure, we are now much more strategic to them.

That's interesting. That's helpful. In the 10 or so minutes I got left with you, I've got here a couple of other quick topics. I can't help but to kind of think about Arista evolving to a point where just software becomes a bigger component of the narrative. You've touched several times on CloudVision. Recently, the Awake acquisition, You've done Big Switch. So how could we or investors think about the evolution of a software-only -- software contribution element of the Arista model?

Sure. The software pieces have to be natural. They have to be actual software offering, not trying to spin your business as software when it's not. As you know, some companies have been trying to do that. That never succeeds in the long run. So we have our switching business. But in addition to that, whether it's big switch, whether it's Awake, CloudVision, CloudEOS are all starting to add up. And we are starting to see more stickiness with customers on this recurring side as well. And as a result of that, the value we are providing, the complete solutions that we are providing is being embraced by many, many companies. So you will see us invest more in this space. But again, in value-add, Big Switch is a perfect example where the monitoring software provides constant value, allowing businesses to monitor and observe their networks and steer traffic to the right security teams, to the right tools for monitoring or to detect hacks and malware and so on, steer it not the right tools, and some of those tools might be things like Awake, which is an NDR product. And Awake as a software-only offering. You don't need any hardware so that you run the software and some servers and complement the story very, very well. So we will continue to invest in this space. Obviously, it's a good business model for us as well. And as long as adjacent to us, you will see the software footprint continue to expand. Mind you, there's also more in the routing software, which might not be subscription to date, but that also adds significant value to our business.

Yes, yes. And a final couple of real quick questions. Kind of similar to the cloud question, one always has to ask Arista, is -- and I think I know the answer to this, but I'm just going to ask. Is the insertion at all change or your mindset at all changed about the competitive dynamics of white box? I'm certain to keep a close eye on that. Is there anything that you kind of -- you keep your eye on that, we say, okay, this is something I got to pay attention to that could drive more white box insertion or competition versus Arista? Just any updated thoughts there?

Well, we touched on this during the last earnings call as well, but I want to repeat that over here, which is, in my view, the white box market has been maintaining status quo for the last several years, especially in the cloud titan space. The companies that have been building their own white boxes are continuing to build them. Continuing -- the companies that are buying from Arista are continuing to buy from Arista. And there's some in the middle like Facebook where they build and buy from us, we do a co-development with them. And I think this ecosystem is quite balanced. Our products are highly competitive. They provide the right value and for customers to start out today saying, okay, let's start building white boxes today, may not make as much sense as it did prior to Arista, which is why some of those models exist. So I think that ecosystem will stay as it is. Having said that, companies that build on their own often analyze the other way around, which is, is building on their own becoming a disadvantage? What if they miss a product cycle? And do they need more features than they can build on their own? And some of those situations, they are actually exploring buying from the industry in the future. And I think -- which is why this is not just a one-way train. It could also be some white boxes getting converted to industry OEM switches, not just the other way around.

Yes. And then the final quick question for you from an architectural perspective, I know this came up on the last call, too. But I watch the semi space as well. And one of the things that's evolving is this next layer of compute, right? I think data processing units, SmartNICs, not to say SmartNICs haven't been around and talked about for some time. But we've seen, for example, NVIDIA bought Cumulus. There's some networking aspirations that appears there. So how does Arista envision the data center? How do you see the data center evolving around the role of DPUs and the network layer, some of the network functionality, maybe sitting in that compute tier on the server side of the world? Just any thoughts around that?

Yes. So some of these shifts, I think, are driven by more by marketing than real needs. But let me address this head on. First of all, I think NVIDIA and Cumulus is more often [indiscernible], right? Cumulus is no longer supporting any of the open white boxes, they only support the Mellanox infrastructure. That's really an engineering team built for their own internal stack. But the DPU market is simply another type of offload. It's different instruction sets, sometimes focused on storage, sometimes focused on AI or NVMe type of offloads. And as a result of that, it's another socket being added on the server where the NIC couldn't keep up. Whether it's part of a SmartNIC or a separate DPU chip, I don't think it matters as much to the network because you're trying to save cycles on the CPU and the server. The network, I think will actually continue to stay very critical and if COVID has taught us anything at all, the work-from-home patterns have shown that the infrastructure is completely -- is very, very essential and critical for us to do our work. And as a result of that, people are investing more in this space, not less. And coming up with distributed architectures, like the ones we already offer to have the facility as well, that it never goes down. So I don't think the stack is changing us as quickly as or at all. You might even see more advanced functions in the network in the future as well.

Right. I will slip in one final question, in the, I think the 5 minutes or so we got left is, back on the service provider side of the market. You said earlier that you got a lot of traction or starting to see more traction in the Tier 2 cloud guys, but this -- the traditional service provider market as we think about Open RAN, we think about 5G and just the architectural shifts, the cloudification, if you will, the telecom service provider networks. How does Arista see that market evolving over the next 3, 4, 5 years as being -- as they start to invest in the core network side of the equation outside of just the RAN? Has that open up opportunities for you to kind of really get your footprint in that vertical that hasn't existed over the last couple of years?

It does. And in fact, at the last earnings call, we mentioned that we are starting to see growth again even in the service provider segment and had some recent wins in Tier 2, Tier 3 service providers, including international regions, not just the U.S. So our customer base in the SP space, the classic telco space, is becoming more diverse. We're participating more and more, not just in the telco cloud, but also in their wired metro networks and thus trying to win there. It is still a journey. There are more features we constantly have to do to catch up to the legacy stack, but there are some service providers who are embracing the new way of doing things, including segment routing or other ways of [ ring ] traffic engineering and [ POS ]. And over there, we participate very heavily. But even on the legacy side, we are starting to do well for things that are in between like a label switch router or an LSR role, which is not at the edge of the core, but in between because these larger networks are very, very large. We're able to compete for some of these roles now. So I see good growth for service provider market as well for us in the future. And to some extent, our growth predictions already have that built in because we do need all of our verticals to grow simultaneously, and those investments are paying off.

That's fantastic. I think I ran through the list of questions, and we're getting up on time. I really appreciate you taking the time joining us today. Thank you, Anshul.

Thanks so much. Thanks, Aaron. Bye.

Thanks. Bye.