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

All right. If everybody can go ahead and please take their seats, we're going to go ahead and get started with our next session. We are very pleased to have Arista's VP of Cloud and Platform Product Management, Martin Hull joining us. We also have Liz Stine, who runs IR, here with us as well. And for those of you who don't know me, I'm Matt Niknam, the networking analyst here at Deutsche Bank. So Thanks, Martin, thanks, Liz, for joining us.

Thank you for having us.

So maybe just to start, Martin, maybe for the benefit of everyone here in the room. Can you tell us a bit about your history at Arista and what you're responsible for in your current role?

Absolutely. So I've got an interesting role at Arista. I'll come to that in just a moment. I've been at Arista for a little bit over 12 years. Before that, I spent a similar amount of time at the other place, Cisco. But before that, I got started in networking back when mainframes walked the Earth. So I've been doing networking for a while. At Arista, I currently run the platform product management team. So everything that comes out of Arista in a gray metal box is basically my responsibility. I also have responsibility for the Cloud Titan sales quota and the systems engineering team that works for those same customers. So I've got a broad role.

So why don't we jump right into that. Let's start with Cloud Titans. Obviously very, very topical. Maybe just to start, if you could talk about what you're seeing from your Cloud Titan customers right now in terms of underlying demand and whether maybe a sharpened focus -- we hear so much about AI-related spend, has that largely been neutralized by deferred spend elsewhere?

There's a lot to unpack there. So we have multiyear relationships with the Cloud Titans, two of whom are our biggest customers, but we sell to the cloud sector. Not everybody is in the Cloud Titan Group, but we have a broad set of customers in that cloud market. So is there a sharpened focus on AI? Absolutely. We would be sitting in this room asking the AI questions if there wasn't, but we have been working with these customers for, I say, for many years. So the deployments that we're going through at the moment are largely in what we call the front end network, the IP and Ethernet space. But it also extends out to the data center interconnect, the wide area networks and many other parts there, colo, their transport networks. Those deployments have to continue. They can't hit the brakes. They can potentially choose to increase CapEx on the AI back end, and that might have a change in focus. But over the long run, they have to continue spending on all the other parts of the IP network, right? The IP network is how they deliver the services to their customers, whoever they are, whether it's a Software as a Service cloud, whether it's a media and entertainment cloud, whether it's a productivity cloud, everybody has to deliver that service. The AI part of it is how they're going to underpin their next-generation solutions. They still have to have delivery, transport, customer connections and that's all the part that we're interested in. Was that a part I didn't answer there?

You did not. There's a lot of subparts to that question, but I think I got that. But I started with the worst that is the Cloud Titans. Is -- maybe just to stay with Cloud Titans for a second, are you seeing any impacts from a lower macro backdrop, elevated interest rates? Is that weighing on sales cycles or customer decision-making processes?

In the Cloud Titans?

Yes.

I don't think that they are necessarily exposed to increased interest rates, parts of their business that we get involved in, where networking is a relatively small part of the company's CapEx is. So I'm not seeing any change based on the macro there. We saw -- gosh, we're in August now nearly September -- earlier in the year, we clearly saw a couple of those large companies doing cutbacks in terms of headcount, right? That's in response to maybe then making overcorrections earlier in the pandemic, those sorts of things, we've got no control over. Absolutely, right. But when they want network infrastructure, when they want to build out next-generation solutions, then they're going to be having conversations with us about that. You know, macro, I'm sure that plays a lot into their thinking long term, but we don't see that short term in us.

Just maybe while we're on the topic of macro, as it relates to enterprise, obviously, it's been a pretty meaningful share gainer. Is the maybe a tougher macro broadly or even the topic of rates, maybe as you think about enterprise based, has that impacted the pace at which maybe you're able to take share from incumbents?

Yes. So there's possibly 2 or 3 aspects to that. And that is, as you say, we're in a net share gainer. So it will always be difficult to look in the rearview mirror and say, could we have done more or less? And was that shaped by an economic backdrop? Difficult to say. We continue to be that net share gainer. If you then go into those conversations with the customers, if the customer on the left is saying not this year, well the customer on the right is opening the door to you. So we've still got those opportunities ahead of us. When we get deeper into the engagements, you do then tend to find the -- to make sure that you can move that deal to completion, you've got to make sure you have all your signatures signed off all the way up. And as long as we're doing our piece to make sure that we can get all the technical sign-offs, all the business sign offs and then finance on legal or getting a way. Contracts perhaps taking a bit longer to get closed out, but we're not seeing a slowdown in our ability to grow our business there. It's difficult to say, could it be more? You can't ever tell.

And just maybe to close out on the topic of Cloud Titans. And by the way, just for everybody in the room, I will pause maybe with 15 minutes left for any audience Q&A as well. You obviously have 2 very large cloud type customers: M&M; Microsoft and Meta for those who don't know.

Microsoft and Meta.

I was going by market cap, but yes, maybe -- is there any discernible difference in terms of their spending patterns, the priorities that you see from them?

Yes. So if you even walk outside the M&Ms to our other markets, what we find interesting because everyone of them is trying to solve the same problem, and they all come up with a completely different way of solving that problem. Okay, we've got a product portfolio that can fit into whatever use case they have for it. So yes, they definitely deploy the technology in different ways. You've also got to think about what it is that they are as a fundamental business, right? I think we all understand what Meta is, or what Meta does. Is Microsoft their competitor? Not really. It's a completely different set of business services. So they have fundamental different businesses. They just happen to be very, very large deployers of cloud infrastructure. So that's -- we sell similar products to them and they use them in different ways.

Maybe let's hit supply chain. That's obviously been very topical over the better part of the last 4, 6, 8 quarters. What's the latest you're seeing on the supply chain front now?

So it's been going on for more than 8 quarters. I still live supply chain challenges effectively day in, day out. That's not to say that most of the problems are actually in the rearview mirror, but we are by no means back to normal. The standard lead time for components that used to be easy to get hold of are still extended. So we have to plan in longer cycles. We then ask our customers to work with us to plan in those longer cycles. We do expect by the end of this calendar year that we'll get back to more of a normal period. But in some ways, I don't think we're ever getting back to normal. Some of these lead times are just going to stay long for I don't know how long. We've never seen this in the 30 years I've been doing it and Jayshree has been doing it even longer. None of us in this industry have seen this kind of scenario play out. So I think it'll take certainly a few more years to get back to normal if we ever get there. But most of the challenges are behind us, and we understand where we are now. We just have to have better business planning, short term, medium term and long term.

Is there maybe easing or moderation relative to peak supply chain headwinds we may have seen not too long ago? Is that impacting customer decision-making behavior across the base?

It has to impact the customer decision when they -- we have these multiyear relationships. So we're talking to our customers about not what they want this week or next week or next month. We're talking to them about their decisions that they're going to make in a year's time. If the components that they need to have inside the systems we're going to deliver in the next generation product or on a 52-week lead time, you better tell us tomorrow. And realistically, some of these components are. Now we work with them. Are those orders? No, these are strategic directions that we're all walking together in hand in hand. So those relationships allow us to partner. So that has to change how they think about working with us. They can't keep their decisions hidden behind the table. They have to help us so that we can help them.

Just relative to -- if we think about lead times, I know in the past, I've heard Jayshree and Ita maybe quantify some of the lead times across different customers. That's where they peaked, where they expect to normalize, if we can maybe refresh us on that as well.

For the products?

Yes.

So we have managed to reduce the lead times on a subset of products so that the typical enterprise customer is more able to get hold of what they need because they can't plan in those year and multiyear cycles. And their build outs don't last as long. So we reduced the lead times on a number of those products that are key to them and then we continue to work with the larger customers on trying to get visibility into longer-term planning. So not committed orders necessarily, but just trends and directions. So we're still working on the business as a 1-year plus horizon.

Going to go back to AI. Obviously, I can't let you go -- I ask one question, but I got to go back to it because obviously, it's a huge focus. From a very, very high level, how big of an opportunity can this be for Arista, maybe from an addressable market or addressable opportunity perspective?

So it's not really for us to try and sum all of that up. I know a lot of the industry analysts have looked at it, and they come up a number of 2 to 3 yes -- 2 to 7, 2 to 3 -- it depends on what horizon you look at, it depends on what you want to group into there and what you don't want to group into there. A lot of the Ethernet technology that we'll sell into a cloud could be deployed in an AI environment and not an AI environment. You try and work out where it got put. So it's challenging at best. It's definitely an incremental TAM in that there's more CapEx going into that area. And then there's the debate of how much of that is going to be an Ethernet-based technology, a multi-vendor and how much of it is going to be a proprietary InfiniBand technology, what that ratio is, how fast it changes. And then you're still going to come back to the underlying, "How fast will this go?". Jayshree said it, we're still in some very early phases here. I know that the hype cycle is pretty high at the moment. Hence, all these questions on AI and totally understand that. But we're still waiting for next-generation silicon on the GPU side, waiting for next-generation silicon on the networking side. And as each one of those iterations happen, I think there's going to be a functional step in terms of how much you can deploy, and we've barely scratched the surface. Right. AI getting deployed in some of these high-profile environments and the stories we see about what they can do. If you're trying to an SAT, you're going to get a perfect score from now on. But if you think about how, you can apply AI to health care, Fintech, pharmaceutical research, all these areas where AI can be deployed for the good, not saying that getting a perfect score and SAT isn't good, it's not maybe quite as helpful. So there's many, many areas where AI can be deployed to all of our lives. And I think that's the thing that we haven't started yet, right? And I see plenty of people out there saying, this is the start of the Internet boom like it was in the '90s. Sure. Okay. We'll see. 20 years from now, we will look back and I know what the answer is. But AI, we're very, very early into this.

Got it. So I'm going to ask, this is going to be a very high-level simplistic question, but at what point do you think -- obviously, you mentioned significant investment being made in GPUs, obviously, broader internal corporate focus. Let's pivot to AI. At what point, if we're trying to sort of track that trail, when did that start translating into more material needle-moving contribution for Arista?

And -- so we're -- keep saying it, early. It's Ethernet. So it's challenging for us to say this is AI, and this isn't because the front end network is kind of still part of their AI infrastructure, it's just well we're going to buy that anyway. So it's difficult to know when it starts and when it finishes, it's difficult to quantify it. But here we are back end of 2023, we're still in the early trials. '24, I think we'll get into meaningful deployments and then '25, I think, is when you'll start to see that kick through into what you choose to call needle moving.

The revenue base is bigger. So it's got to be needle moving. Okay. Ethernet versus InfiniBand. Interesting topic. I know it's becoming more and more topical. What are the relative advantages of Ethernet relative to InfiniBand? And I say this as a maybe less technical person asking more of an expert in the space. What are the relative advantages of Ethernet relative to InfiniBand as it relates to AI networking infrastructure?

So there's probably 2 or 3 points and not all technical. But Ethernet is a multi-vendor technology. You can buy it from a number of vendors. I'd always like the customers to buy it from Arista, but they have choices, a multi-vendor market out there. Ethernet is the primary technology deployed in data centers today for all network functions. InfiniBand, the single supplier technology that gets deployed in the back-end network, effectively a niche, a corner case. Because you want to try and expand your AI scenarios to be ubiquitous in every location around the world, are you going to repurpose every single data center from Ethernet to InfiniBand? No. So how much does InfiniBand grow in those ubiquitous deployments? And how much does it get contained within a niche? That's going to be a decision each customer chooses to make. But when you look at the collective, you look at the broader picture, Ethernet inevitably becomes the significant technology for AI deployments. That's kind of the business side of it. If you want to go out to the market, you want to hire a networking expert, you challenge them on all that questions. Those are going to be Ethernet questions. You're not going to be asking InfiniBand question. So where do you find the spots? Customers don't necessarily want to go to InfiniBand, but that's where they are now. Then you look at the technology, scale out, Ethernet, multi-tenant segmentation, multi-clients, hosting giving you a slice of an AI and a different customer, a different slice of an AI, Ethernet has the tools and the technologies. Ethernet has the troubleshooting, Ethernet has a monitoring. And then by the way, that I mentioned, multi-vendor. So I think it's inevitable. But I'm not going to call the score in the first innings.

Okay. Okay. Great. And just -- you mentioned customer like effectively just using sort of multi-vendor and maybe customer propensity or desire to do that. Any initial reactions here from customers just in relation to the Ultra Ethernet Consortium that you're part of alongside some other partners?

So the Ultra Ethernet Consortium. The key is in the third word consortium. It's multivendor. So it's systems vendors, it's silicon vendors and it's some very large-scale customers. Recognizing jointly that there are some areas within the existing technology that can be improved and enhanced to be more suited to the challenges that AI is bringing. You form a consortium, you work collectively on it, bring the best brains industry together and you enhance the existing Ethernet protocols. If you do that so that you now have a solution that has the best of breed from multiple suppliers, silicon, systems and customers. Customers look at things like flow control, monitoring, troubleshooting and saying it'd be nice if we had this technology, this capability, this new header in a field, and if we could all just agree what the header looks like and what we're going to call it and how we're going to use it. Let's implement that in silicon, then in systems so that the end customers get that advantage. So that's what the Ultimate Ethernet Consortium is trying to do. There's some initial work. There's a draft going to come out soon on some of the first works of that. It's now been opened up that other people can apply to join the UEC, it's an open consortium. And I look forward to seeing what comes out of that in the next few years that helps customers take Ethernet to the next level.

You obviously mentioned responsibilities tied to Arista's broader platform. Maybe from a high level, if you could talk about some of the key differentiators that have enabled Arista to win the type of market share that it has. And we can maybe spread this both across Cloud Titans as well as enterprise. Largely getting to what's the secret source?

Well I'm the hardware guy. The secret source is software. It pains to say it. I joined Arista say, a little bit over 12 years ago. And I used to do -- I was joined as the systems engineer. So bottom of the rung. And we used to go out on customer calls -- but I would talk about latency and packets a second and how bright the LEDs were and the customer's eyes would roll in the back of their head. So we pivoted and started talking about software, open, programmable, scalable, resilient, reliable -- and by the way, you can buy the software to run on this really cool hardware, if you want. So software is a secret sauce, software differentiates everything we do from an architecture perspective, up sent into a high-quality product. And again, it pains me to say it, software is the differentiator in this industry. And the EOS software quality speaks for itself. Right? If we didn't have a quality product, we wouldn't have got our first customer, we wouldn't have got our 7,000th?

9,000th.

9,000th customers, right? And that's how we have those conversations with customers. Give us a chance. [ What have you got] to lose? But we get them into what we call a proof of concept, they test it, they try it. They put it into a production environment. And the proof is there.

Are there other -- obviously, we think about sort of adjacencies, maybe broadening the platform. Are there areas, opportunities you see where Arista may be able to bolster the platform either organically or inorganically? And are there things that customers may be asking you or want to see more of in terms of platform enhancements over time?

Yes. So no one knows the future, but if we look to the past, we've done both, internal development and we've done bolt-on acquisitions. So if you look at if you look at a typical campus network and here we are in a campus, we're using WiFi on our phones and laptops, WiFi was clearly an area where we needed something to be able to talk to the enterprises, we did an acquisition a few years ago. Then you start thinking about how would enterprise runs a network in has to be skewered from the inside out. So we have Awake acquisition. We have also an internal development of an identity solution. And we recently announced our first foray into enterprise wide area networks. So you kind of circle that core enterprise and you look at the technologies that are layered on top of it. I think those are the obvious areas for identity, security, connectivity, WiFi, and then the XR connectivity. I think those are the areas where you're going to see us doing that. Big bold moves, I don't think that's in our DNA.

I'm going to pause and just open it up to anybody. A question. I see there's one here.

Martin, just to follow up on the enterprise side. Can you talk about -- has the pace at which you guys can gain share, has it seen a step change? Is it something different than the past? And the reason why I'm saying this is that I understand the point you made about the portfolio being fuller, about what you can offer the enterprise now. But also part is just replacing existing equipment from the incumbent. And I always thought there was only a certain pace at which you can gain share. I'm trying to understand, is this recent accelerated growth something that's really changed in your mind? Or will it be steady share gains as we've seen in the past? Is there any difference?

There's probably a couple of dynamics playing out in that one. Clearly, a lot of people hit pause 3 years ago. And so there's a tax they had to be paid down some infrastructure that's aging and it didn't get refreshed for a while. So it's due, it's coming up for renewal. There's also the 800-pound gorilla in there, had their first generation of their licensing subscription model. That's also coming up for renewal on their first rounds of re-ups. So both of those create opportunities, at bats, if you will. As you commented, our portfolio has expanded. So the conversations with the customers now are more full. If the customer isn't necessarily interested in an infrastructure refresh maybe there's a security conversation. Maybe there's a WAN conversation. But we're able to have more conversations with the same customers. So you can keep engaged and that then leads to being in the right place at the right time. So multiples of those factors kind of come together, but you still have to execute, right? The 800-pound gorilla is not going anywhere and why would they. So we still have to be on top of our game, right? We can't assume anything. We don't walk into these opportunities. But you do development in more over years, you develop the relationships. And the other thing we're starting to see now is key customers advocating for us and turning around to their peers in the industry and going, why wouldn't you use Arista? So you start to get that name recognition, the brand recognition, the referrals. And again, sometimes that's an accelerator to our business model there, all the verticals. But within -- if you think about the financial services in New York, right? It's a fairly close shop of there. There's a whole bunch of technical guys who talk to each other. They change firms over years. They know each other. You can say the same thing about where from the Bay Area NorCal, Customers know each other. There's an undercurrent and a great line of have you used it? What do you think? How was your experience? So that peer review, that peer recognition, nobody necessarily wants to be first, but nobody wants to be last either.

So let's pivot a little bit and talk about 2024. And Liz, I apologize in advance. You can jump in and stop me whenever you like. So on the last call, there was talk of some optimism around maintaining double-digit growth next year. And again, as I sort of look at your 2 largest customers, at least, they intend on accelerating their CapEx spend just based on what's out there in consensus for next year. How does what you see from them from these larger customers and broader customer base align with sort of your current visibility and align with that view around maintaining double-digit growth?

Definitely going to put that one to Liz.

We'll definitely get into kind of a little bit more of the details around 2024 at our Analyst Day in November. So shameless plug for November, right? But I think as we go through the year, obviously, we upped our guidance for 2023 to 30%. When we get into November, we'll have a little bit more of that visibility and we'll kind of break out what the business looks like going into 2024. But this internal goal of kind of that double-digit growth, I mean that low double-digit growth, this has been something that's been around for a while. Jayshree, even mentioned it kind of last November when we were talking around that kind of 5-year CAGR that we put out there from 2020 to 2025 of 20%, and that implies some deceleration. And so the internal goal has always been, can we grow the business double digits even if cloud could be somewhat cyclical. And we've always said that could be somewhat cyclical, what that cycle actually looks like is kind of the hard part to put together and the timing of that cycle. But that's always been the internal goal.

Okay. We'll stay tuned.

Yes, stay tuned.

Just one -- we talked a lot about cloud sites. And obviously, we focused on the 2 bigger ones who are certain names. What about the use cases and opportunity you have with other potential? I think there's about 5% of that 46% last year. There's about 41% from M&M. There's about another 5% that are the other. Where do you see potential for Arista to maybe address opportunities with maybe other cloud or scaled customers that are not being met today?

So as you call out that clearly, we have some business with them already. Would we like it to be larger? Absolutely. We have a role to play in their network infrastructure today, but it's not at what you like to think is at the heart, not at the center. Opportunities within those areas are not necessarily technically based. These are strategic organizational decisions as they've taken many years ago and their selection to go in a different strategic direction or kind of the triggers that it will take for us to be able to move in that direction, we engage with them deeply, same engineering teams that report for the other hyperscalers, same engineering teams, their criteria, their requirements aren't fundamentally different, they just have a different sensitivity in terms of build versus buy. So we remain ready. What will it take? I don't think that's in our control.

And is the change in supply chain dynamic, the worsening dynamics last couple of years, has that changed the thought process from some of these other customers?

No. They've got deep supply chain organizations. We're not still networking, we're talking about building facilities and compute and stories, they effectively built everything themselves they have deep supply chain organizations and they have got just as many professionals in those organizations as we do in ours, sometimes [ come to get it ] more. So they're just as able to address the problems. I don't think anybody solved it but addressed the problems when we did in the same way that we did. We have conversations in supply chain to supply chain. What are we seeing collectively? What do we think is going to happen a year from now? So those conversations not just engineering. We're talking supply chain to supply chain to make sure that we're all -- everybody leans left at the same time. Everybody leans right at the same time. We all act as humans, right? We all kind of go in the same way. So sometimes, it's good to make sure that we are all leading in the same direction. We're seeing the same thing from supply chain. So we have a deep engagement with them, engineering level, supply chain, software to software and what will it take? Difficult to say.

Okay. we've woven in a couple of questions on enterprise. But if we can maybe talk a little bit more around how demand within Enterprise has been trending, what the growth profile looks like within enterprise? And -- maybe I'll pause there and then throw in a question about backlog afterwards, but maybe we'll start with you.

Yes. I think when you look at our enterprise business, it's all about new logos. It's all about expanding the portfolio or the presence and footprint within our current customers, right? And we are still the share gainers. If you look at financial is our oldest vertical, we'd still probably say that we're underpenetrated there and they're still a long runway of opportunity in that. And now we've broadened out our verticals. You hear Jayshree talk on all the earnings calls, try to give some examples of some of these enterprise customers that we're dealing. And then obviously, we've expanded our portfolio. As Martin was saying earlier, that gives us more opportunities at the table. We're no longer sitting around waiting for a data center opportunity, which could be 5 or 7 years in the making. We have opportunities at campus, routing, network monitoring, detection, like all of those are opportunities to prove out the quality of the EOS to prove out the operational efficiencies of CloudVision. And so we're always looking for that corner case that use case that we can kind of prove out those because really experiencing the difference of EOS is what makes the decision, right? Everybody is selling a quality product. You want them to experience it. So looking for that corner use case, having more opportunities at the table to prove that out and then doing a natural land and expand. So it's really around new logos, and then expanding those levels.

Is there still -- I mean, maybe what is the growth profile, if you can help us just ballpark, think about how fast the enterprise is growing and how sizable if it still is, in terms of backlog burn and contribution to that growth?

When you look at kind of the data center side and you look at the campus side, I think the campus is a good metric to watch. Obviously, we've been making traction in the campus. We had the $400 million goal for 2022. We talked about meeting that from a demand perspective and obviously, from a shipment perspective, maybe fell a little short. We've put out kind of that longer-term 2025 goal of $750 million, and we continue to execute well against that goal in campus. And obviously, that's still a very small portion of that overall market. We're really targeting those large enterprise customers. If you look at the campus market overall, maybe $20 billion, and half of that is those large enterprise customers, and those are the ones that we engage with in the direct sales model that we have at Arista.

Okay. And then just maybe one final question at the time that we have to maybe tie this all together. And maybe I'll throw it out to both of you. What do you think is the most underappreciated aspect of the Arista story from your seat internally. And then maybe Liz from your seat as well.

I'm still going to say software and I hate to say it. So seriously, software, EOS is the biggest difference, right. I came into Arista 12 years ago excited about the silicon and the low latency switches in the high performance, but at some level, anybody can build those. The software is the biggest differentiator. And I think it lasted us this long and lots of conversations with customers over the years about our quality is better. Everybody says that. We have a single source tree. Everybody says that. Everybody says it, nobody means it. There's been a number of Aha moments over the last 10 years where we've been able to deliver a feature to a customer that was game-changing for them, right? And the features are in software. So the ability to use software to leverage our capabilities. We are a software company who happened to run it on top of merchant silicon on dedicated hardware appliances basically. But that's the biggest thing that we keep saying it, but we still get underappreciated for the fact that software, EOS and CloudVision are what makes us unique.

Yes. I mean I think that I would always say kind of the enterprise piece of the business just because I joined 11 years ago, also as an SE kind of targeting those enterprise customers, right? And seeing our product expansion, obviously, being able to offer kind of that end-to-end enterprise network, I think that's somewhat underappreciated. And everybody wants to talk about cloud, everybody wants to talk about AI. They forget that the enterprise is a good portion of our business, both in the data center and the campus and the routing and the network monitoring and detection and we continue to build new products and features and functionalities that are really targeting these enterprise customers and helping them operate much like the cloud.

Great. Great place to end it. Martin, Liz, thank you both.

Thank you.

Thank you, Matt.