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

Great. Good morning, everyone. Thank you for attending the UBS Global Tech and AI Conference. I'm David Vogt. I'm the hardware and networking analyst here at UBS. And we're really excited to have with us today Arista Networks. On stage with me today is Chantelle Breithaupt, the Chief Financial Officer and SVP. In the audience is Liz Stine from her Investor Relations. So if anyone has a questions, follow up with her afterwards. So Chantelle, thank you very much for joining us. Before we get started, I just want to make sure we don't need to read any sort of disclosures or disclaimers. I think we're pretty good.

I think we're covered. Thank you. Good morning.

Perfect. So in the interest of time, I think what we wanted to start with is -- we've been getting a lot of questions since your last earnings report. Help us understand how you think about the framework for 2025. I know you're obviously very involved in that process. How do we think about the puts, the takes, the paths to achieve that number, and why the guidance was set at, call it, 15% to 17% growth next year? And then we can jump off from there.

Yes, sure. Happy to. Good morning to everyone in the room and those attending virtually, if that's happening. I think that -- so we gave guidance for 2025, a growth rate of revenue of 15% to 17%, putting the midpoint at 16%, which Jayshree, the CEO, and I think is a very respectable growth rate going into next year. So we're happy about that. I think the way that Jayshree's style and mind are very similar, in the sense of we prefer to put a guidance out that is something that we see as achievable. And then we see from there if there's other things we can do as we progress through the year. And so with that there, it comes to what we call the term optionality. And so to your question, David, there are different ways we could get to that number next year. As of now, we see it mostly as a cloud AI kind of heavy year, and therefore, the gross margin guidance, which you may ask me about, was set at 60% to 62% from a customer mix perspective. But we're very excited about enterprise, both data center and campus. We're very excited about AI. We're very excited about classic data center for the cloud. And so upside to the guide could be everything hits on all cylinders, right? That's a potential. Or we see moderation and the 15% to 17% was the right guide. And so those are the different things at least from an optionality perspective.

Can we start -- maybe let me take a step back now on '24. So obviously, '24 has been a great year, a very strong year. You recently took the guidance up again in '24. Can you remind maybe investors where AI sits today in '24 and how to think about the glide path to '25, both on the back end of the network and on the front end? And then maybe we can follow up on more detail there.

Yes, happy to. So if you guys -- as I started the year coming in and transitioning as CFO, 2024 was originally guided at 10% to 12% growth, and that was due to -- at that time, it seemed like maybe a moderate cycle for classic data center spend, and not really much about AI was clearly known in the sense of what it could mean in 2024. As we progressed through the year, we saw that the CapEx that was allocated to AI was not cannibalizing the CapEx spent on classic data center. So that allowed us to raise our guide. We saw strength in enterprise, both campus and data center. So that allowed us to strengthen our guide in both the enterprise and Tier 2 networks. In that, as we progressed through the year, we kind of reiterated this goal in 2025 of $700 million of AI revenue. And the glide path is progressing well, and we can talk about the pilots that we talked about in that glide path in your questions later on if they come up. But from the perspective of coming through the year. So the back end, that's just -- the $750 million is just the back-end AI cluster, very easy to measure, very easy to kind of contain and explain. But as we kind of came through the year, we noticed that in the industry, there were so many definitions of AI, we needed to maybe open up the aperture and talk about the potential front-end pull-through. And that front-end pull-through, we see right now as being potentially 30% to 200%. As we did the 2025 guide, we kind of put it down the fairway and said, let's assume 100% and got to the $1.5 billion that you're referring to in 2025. And that front-end pull-through can be seen as either AI or data center, just classic spend, it's a little difficult to pull them apart. So we're not quite sure, but we expect the overall to materialize.

So do you think it makes sense to maybe not classify the spend, to your point, as AI? Because ultimately, you're going to have a significant investment cycle in the front end of the network also directly attributable to AI. So while it's not classic AI in the sense that it's directly connected to an AI cluster, the investment cycle is driven by said AI cluster. So as we move forward long term, this is more of a longer-term question, as you build and stack revenue, is it going to start to basically get blurred with your customers, right? So one of your large hyperscaler customers is going to come ask you for deployments of X, Y and Z, and you're effectively indifferent, whether it's an AI cluster in the back end or in the front end because it's all revenue dollars that are fungible to you.

Yes, I think you're spot on. I honestly think if it works for us as a community, starting 2026, maybe time frame, we stopped delineating AI generally in the conversation, enterprise or cloud. But we'll progress and see how it goes. But I think generally to disband it at some point is a good idea.

And just going back to that ratio front end to back end. So on your call, you and Jayshree mentioned $750 million back end, starting point for '25, $750 million as that front-end tie-in. So that's $1.5 billion. And we did get some questions last night about, in the context of your $8 billion guide, roughly speaking, for next year, the conservative nature of your guide would suggest, if I just take that $1.5 billion and I give you some credit for campus growing to $750 million or so, the rest of the business could be flattish. Maybe can you kind of help us understand, go back to your first point about, there's multiple ways to get to the guide and then upside to the guide. How should investors think about that framework?

Yes. I think back to the point we just mentioned, so the $750 million front-end pull-through, you could classify as AI or classic data center, which would just be growth in that perspective. So it depends where you put that. But generally, again, we picked one path that we see strength in and visibility to. But we expect -- we're very excited about enterprise. We're very excited about kind of the cloud service provider. So again, if everything materializes, we will continue to revisit the guide. But from this point, we put in the visibility we have and therefore, the margin kind of parameters that we could expect, and we'll see how it develops through the year. Yes.

And then sticking with AI for a second. So you've been very clear about the number of trials and pilots with different sort of timetables from each customer. Maybe we can talk about where each of your customers, without getting into specific names, are in their journey effectively with you. I know, I think, Jayshree mentioned one is a little bit delayed or paused maybe a little bit later. Can you maybe help us understand where each of those pilots and trials sit today as we go into 2025?

Yes, happy to. And so I would start just maybe at the beginning, kind of in Q1 when I entered the role, there was a lot of discussion on InfiniBand versus Ethernet. And I think we're very pleased that 4 out of 5 has become 5 out of 5 in the sense of looking at Ethernet. So we're very pleased generally at that construct. Now if we talk about kind of the piece parts and the timing. So it's trials, pilots and production, that's kind of the process. Trials are a couple of thousand GPUs, then you get to tens of thousands and then there's aspirations of 50 to 100. That's kind of the progression. 3 of the 5 of the trials have moved to pilots and expected to move into production over the next kind of '24, '25 time frame. The fourth one is working through basically their funding model. And so they need to think about how they're starting up and funding. And so we're here for them when they're there and ready to do that. And the fifth one, we're very pleased because it's traditionally an InfiniBand shop that now has engaged with us on a trial, and we're super excited about working with this customer in this capacity.

On that InfiniBand customer, does the formation of the Ethernet consortium, did that help? Or was it just something that had been in the works between Arista and this customer for years? They know your technology, they know your EOS operating system. How did that come to be migrating from an InfiniBand-only effective customer to obviously opening the door to an Ethernet solution down the road?

Yes. I don't know for sure if the UEC benefit -- and we can talk about those in a minute. I don't know if that's the pinpoint or inflection point. I do think that there's perhaps within the customer just realization. It is someone we worked with, have 2 to 3 years of road map development together. Maybe that gave them confidence in the sense of do it. And maybe they just wanted to have an experience of dual-sourcing and then understanding how they want to do their back-end AI clusters. I don't know for sure, but that would be my hypothesis.

And then along those lines, so obviously, sticking on data center for a second on the AI side. Obviously, the market is a little bit more competitive today than maybe people would have characterized the hyperscaler community 5 years ago, right? Cisco is a resurgent competitor. Obviously, NVIDIA has an offering. Nokia to a lesser degree, I think, has an offering. Obviously, you guys have a really competitive offering. How do you think about sort of your positioning with your existing and potential new customers going forward? Like what sets your strategy apart today that is durable that you think going forward relative to this increased competition? Or is the rising tide just so large that you're somewhat irrelevant -- it's somewhat irrelevant to you in terms of how you get there because the revenue dollars are going to be so large over the next 2 to 3 years?

Yes. I think generally, Jayshree and I and the leadership team feel, if this market is what we think it is, we added $10 billion to our TAM, raising it from $60 billion to $70 billion, looking at the AI addressable market that we see. We feel there's room for a few competitors. And if we demonstrate our execution, like we usually do with innovation, we feel we can still do what we projected to do. But I think one of the key things, at least what we're proud of for differentiation, so I think we're well known and established in the front end, right? I think that's something we can say with confidence. And I think when it comes to the back end, in June, we released products that we feel give our customers best-of-breed breadth of portfolio. The 3 that we announced offer fixed form factors, virtual output queuing, chassis design and then a distributed Ethernet switch. And we have some co-development with one of our customers' third generation on the 7700 DES. So I think we're showing we've had the innovation on the back-end AI clusters. We have a great use case with AI. And then with best-of-breed choices, we feel confident that we have a good shot at it.

Let's pivot to what could go wrong. So we always hear there's a lot of gating factors on power, cooling. Other companies have talked about the network being sort of a gating factor in terms of the deployment of GPUs or maybe the utilization of GPUs. Where do you see sort of those challenges today? And how do we think about their potential governing of your glide path next year? So is cooling an issue? Is power gen an issue? How do we think about what -- how that might affect your multiple path to hit your targets for next year?

Yes. I think I'll start with maybe the network piece first and then work to other ecosystem parts. So one of the benefits of the UEC, 2 of the things we're working on specifically for back-end AI clusters are congestion management and efficiency load balancing. And as we work through, we expect the timing to be early next year if everything is on track. Arista is one of the founding members, and now it's up to 97 members participating to try to get the spec from end to end to allow for these things to happen. And the products we released are forward- and backward-compatible. So we feel ready for that. I don't know if we would say networking is necessarily a block, but we'll leave that to other's opinion from the perspective of all the other ecosystem parts. So there's a lot that needs to come together. And this is really a year of technical roll up your sleeves, right? These are large installations. And so it's power, it's cooling, it's facilities. It could be in the sense of a new administration coming in, perhaps there's macro things that come into play in the sense of how we source and move piece component parts around the world. So those are some of the things that could become an issue, but we don't see those so far. And in sense of our guide, again, we go back to try to have an optionality, and we take an assumption with all those things we know.

I was going to bring up the new administration later, but since you mentioned it, obviously, it's early days. Is there anything that you take away from Trump 1.0 versus Trump 2.0 that informs your view of how to think about '25? If I'm not mistaken, data center networking is priced in U.S. dollars, so that's not an issue. The supply chain obviously takes into consideration potentially some tariffs. Like how are we thinking about that today? I know it's very preliminary, we won't hold you to it. But is that something that has been an ongoing discussion since early November? And how should we frame it in the context of whether it's revenue, raising prices and/or margin impact from your perspective?

Yes. So it's early, but I think we absolutely have a responsibility for a company of this size to be thinking ahead of time. And so we have, for sure, John McCool, who's the Platform Officer, Supply Chain Leader, and I have definitely spent time. And what we've learned from the first time around is, we actually learned how to manage the tariffs, understand the tariffs, have a COVID kind of agile supply chain, which was during the 1.0, not because of Trump. So from that perspective, we have learned how to be agile from our supply chain and sourcing perspective. So we've been discussing scenarios there, David, to your question. I don't think we're at a point of talking about price increases in any aspect. I think that would be a wait and see exactly what happens. But we're ready in the sense of making sure our customers get what they need when they need it.

Got it. Okay. Pivoting to enterprise. So obviously, enterprise has been an incredibly strong vertical for you for the last couple of years. Obviously, it's a slightly different go-to-market motion for you. We talked a little bit about this last night, and we can touch on margins later. But how are you thinking about sort of the enterprise opportunity '25, '26 and beyond, given the strength that you've managed to demonstrate with a very relatively small footprint in terms of channel partner relationships, et cetera, versus some of the larger players out there? So how should we think about your -- basically your enterprise business as we go into '25 and '26 and given the footprint that you currently have right now?

Yes. So we're very pleased with the growth we've had, the share we've geared -- the gain we've shared -- the share we've gained, from that perspective. So I think that -- but here's where we are. So if you look at from a logo penetration perspective, we're maybe 20%, 25% of Fortune 500, Global 2000. So we still feel there are lots of logos to go get for both data center and campus. And when we talk about campus, our goal of $750 million of revenue next year is -- a TAM, we think is more in the $10 billion range of campus switching large enterprise. So we declared cloud. We went and built for cloud. We declared data center enterprise. We've done well in that. Now we're declaring, "Hey, campus is our next kind of intention." We built out the portfolio. I would say we're getting brand recognition, name recognition. The thing that we're very proud of is now we're winning campus deals before having the data center. So you can imagine we're winning the campus and pulling the data center through, which I think is a great testament to the team, both the sales team and the R&D team in the sense of how making that happen for our customers. So we're excited. We're excited about international growth. Part of what I spend a lot of my time on is enterprise, everything from pipeline to conversion to coverage, international. And channel, we're being very intentional about. With channel, you can lose your destiny a little bit on margin and inventory management. So we're doing that carefully where we need to with some specific elite partner programs. So I think we're very excited about campus, lots of room to grow in data center and enterprise.

So I think you've mentioned this repeatedly that selling networking gear from an account management perspective is very technical. So when you think about your opportunity to grow and expand, how are you thinking about that talent, finding the right people to augment what you currently have without kind of disrupting what Arista has made, what Arista's reputation is in the marketplace, right? So there's a lot of disruption in the market today. There's a pending transaction. Whether it closes or not, we don't know. But how do you think about putting your footprint on what Arista 2.0 looks like now going really aggressively at enterprise, a little bit more aggressively enterprise and campus. What are you looking for to bring in? And how do you intend to use like your OpEx dollars to hit those targets?

Yes, great. So from a sales perspective, there are 2 things we're focused on. One is to continue our double-digit growth in sales head count for coverage. And I think part of what you're mentioning, David, is we're absolutely seeing tailwinds from competitive events that are freeing up great talent, that are having people decide maybe Arista is a company I want to come join. And what we look for is great depth and technical capability. That's what we expect from our sales staff. The other thing we're looking to invest in are the things that are kind of scaling the company, spending time on our CRMs, our ERPs, our channel models. So I spend time on that working, getting the processes because it is a different motion, right? It's higher volume, lower dollars. So you need -- it's a different rhythm than the cloud hyperscaler sales. So we're definitely focused on that. And [ Krishnan ], who's the sales leader and I spend a lot of time on that on a weekly basis.

So let's bring it to margin then. So your initial margin guide for '25 reflects mix on gross, but it also reflects the investments that you just sort of alluded to on the OpEx side go-to-market. Can you kind of walk us through, maybe remind everyone enterprise versus hyperscaler Titan gross margin profile differentials? And then how to think about sort of the investment dollars. You mentioned up double digits just recently, but just maybe kind of help tie together to get a sense for where the leverage is in the model if things come in stronger than anticipated?

Yes, for sure. So from a gross margin perspective, the guide of 60% to 62% is customer segment-driven. As you can anticipate, like most companies, larger deals get better pricing. That's just the nature of that kind of business. So there is a differential. And the other thing, just to make clear, and thank you for reminding me is from an AI perspective, we don't have AI products specifically. We have products that serve many use cases and AI is one of them. So from an AI perspective, AI cloud will follow cloud kind of margins and AI enterprise will follow enterprise-type margins. So there isn't a third category from a margin perspective. When it comes to operating margin, we have the guide that we have. And the leverage there is we have a very, I would say, productive model when it comes to cloud and AI. And I think we're gaining some productivity in the enterprise side. So if everything runs on all cylinders, and we have a different view going into next year, it wouldn't be a one-for-one increase in the sense of we would have some productivity leverage. And so we could talk about that guide as we go through the year.

Well, so that was going to be my follow-up question. So obviously, the guide from an operating and a gross margin perspective was set based on client -- customer mix and your investments. This is going to come off sounding a little bit like Pollyannish. But over the last 3 years, you've done the same thing, right? And every year, there's been incredible operating leverage where you haven't been able to invest at pace with revenue growth. Why should '25 be any different? I know you want to be conservative. I understand...

No, I think it's -- I don't think there's anything to not be proud of in the sense of how we approach things. I think that it does leave us room. So the one thing, for example, we're not clear on yet is, what do we need to do from an R&D perspective for AI looking at '26 and '27? Perhaps there's something we need to do from a small tuck-in, reinforce some R&D bandwidth, et cetera. So it leaves us the room to do the necessary. And if we don't need to, then obviously, we have the conversation about delivering that value back to shareholders. But we'll always leave ourselves room to grow the top line through investment. I think that's just a philosophy...

But it's not a shift in vertical focus from data center -- hyperscaler data center enterprise campus. We're not going to see a pivot to -- because there's market disruption in telco, hypothetically. You're not moving into telco routing by any stretch of the imagination.

Yes. From the tradition of traditional ISPs, we haven't changed our philosophy on that. No. No, it's just to allow us to do what's needed if we need to. And then if not, then we talk about how we move that through the year.

Got it. And then from a -- from a silicon perspective, just so we have an idea of how to think about '25. I know Mr. McCool mentioned on the call and Jayshree mentioned, you mentioned there's some increased purchase decisions to ensure that we don't have a supply chain issue. Is that factored into your sort of OpEx numbers as well?

Absolutely. Absolutely. It's all connected.

Right? So that's where it -- it's connected into the numbers.

Yes. Yes. And I think just to be clear, at least the way I interpret, one is it's assumed in the financial guide, but it's also assumed in the commentary around purchase commitments will be -- purchase commitments will be of different magnitudes in different quarters based on what we need to do from a lead time perspective.

Right. So it's not -- it's just not linear, right? It will be quarter-to-quarter.

That's right.

So that was the other question. I was going to go back to gross margin. I know we don't give quarterly guidance on gross margin, but mix purchase commitments could create some lumpiness in gross margin next year. It's not going to be linear, like we start the year at hypothetically 60% and we have a steady improvement as the year progresses. Is there a degree of variability, do you think depending on how revenue falls next year?

I think that if you kind of trail the dot through, the reason the guide 60% to 62% next year is because of some of the larger cloud AI deployments. Those are by nature lumpy. Therefore, it could be lumpy through the year. Yes, absolutely.

Got it. Okay. And then when we think about '26, you mentioned R&D investments. Just -- this is more of a philosophical question. When you're looking -- when you're having a discussion with 4 of these 5 customers, where do you think your visibility sits today in terms of beyond '25, in terms of road maps? Is it 18 months, 24 months, 6 months, like sometimes Jayshree mentioned? So I'm just trying to get a sense for where you think...

Yes. We talk about -- it's actually -- thank you for asking because it's a great clarification. We have 2 types of visibility we speak to. One is the 6-month time frame that Jayshree and I have, and that's almost like PO-in-hand business that we know is going to be conducted. The other is the sense of the really great engineering type, close relationships we have, and those are 2 to 3 years. You're seeing some of the co-development of the 7700 DS, that's a co-development. That's a byproduct and outcome of these kind of conversations. We like to be seen as an extension of the engineering team at these large hyperscaler customers and some of the large enterprise as well. And so I think that's absolutely still 2 to 3 years. We're getting great questions. There's lots of questions on liquid cooling and optics. And so I think those are really great conversations that the thought leaders and Arista are having, and we appreciate those.

Got it. So we got a question on campus. I know it's a small part of the business, but it's an important part of the growth driver over the next couple of years. What are you doing differently -- I'm going to paraphrase, what are you doing differently today than 3 years ago to attack this market given there's a large 800-pound gorilla in the market? There is some disruption potentially going forward in that market with these 2 competitors. How do you think -- does that change your view of your ability to take share in that market? And if so, how do you go about, obviously, driving that share gains if this disruption is real in the market?

Yes. I think if you look at '24, '25 versus 2 or 3 years ago, I think what Arista has done is definitely developed the portfolio and worked on the name brand recognition in the campus market. So we have been proactive in that regard. What the market has done that has allowed us some tailwind in the sense of competitors is confusion and road maps. We do get customers who we tried to get to a few years ago, saying they had an incumbent and they were good. Now they're actually coming to us saying we'd love to talk to you again because we don't feel confidence, or we don't have confidence in the road map of some of the competitors. So that's a great tailwind coming to us. I think that from the perspective of just having focus within internal processes to have intent is something that we've done different than 2 to 3 years ago. So portfolio intent and setting up the structure. And the market has allowed a little bit of a jump start in the sense of some of the things that you mentioned. So I think the 2 to come -- we're very excited. And I think that this is that $10 billion TAM that we can go after, and we're just getting started.

And so just maybe one follow-up. So in addition to revenue opportunities, you mentioned it briefly, freeing up, potentially, talent. Have you seen that? Or is it too early to really see potential really excellent engineers, product managers looking for new opportunities given the uncertainty that they may be facing in their current...

Well, I think we have definitely interviewed and even recruited some great talent that's opened up in the market, and we're very happy to see it. And so yes, we're very thankful, excited, and we'll continue to go after it.

One other question I got from the audience is on capital allocation. Historically, Arista has done smaller deals, where there's a technology focus, a need and/or when the stock has maybe been a bit weaker than expected, has repurchased shares pretty timely. How do you think about capital allocation as we go into '25? You mentioned earlier some of the R&D initiatives, some of the technology-focus initiatives that you have on your plate, but how should investors think about 2025 from a capital allocation perspective?

Yes. I think that I would say organic investment is one of the key ones in the sense of doing what we can to the talent that we spoke of before. I think that, obviously, share repurchasing is always top of mind, and we continue to watch that for the right conditions and market opportunity. Tuck-in M&A is always an option. We look for that for pockets, whether it's talent or technology. The thing that we're trying to be very careful of is that the -- one of the differentiators of pure-play networking for Arista is the EOS beauty of having the software and the hardware and having that all come together. So we're very cognizant that larger M&A needs to continue that and not disrupt it. We're very clear on the sense of that's our goal is to not disrupt what we have today. And so a larger M&A, as you know, integration can be very disruptive. And so it would have to be a really clear case that, that made sense. So we're always looking for those opportunities, but we don't really see them.

But there's no -- I mean, as we sit here today from our view, there's no glaring hole in your business that would require a larger deal to kind of meet the road map over the next 2 to 3 years with your hyperscaler and enterprise customers.

Right. I think it's irresponsible to do M&A just because you have the ability to. Like that's not a reason to do M&A. So that's -- sometimes we get questions in the sense of, hey, with your cash, why are you -- and that's not a reason to do M&A.

So that brings up the final point. So you generate a lot of cash. You're theoretically going to generate a lot of cash next year. There's only so much, I think, that you want to buy back in terms of shares. So should we just expect cash to build through '25 unless there is a unique tuck-in, bolt-on acquisition that makes sense? Is that kind of what we should expect from a cash perspective?

Well, I think the cash will flow with that capital allocation strategy. And the one thing I've spent time on with some of the community is the sense of, if you take cash as a percentage of revenue for Arista, that's one ratio. But if you take it as a percentage of market cap, we're nowhere near different than the average in the market. So we kind of look at both to make sure that we understand where we are, and we'll continue to monitor from there.

Got it. I have a ton more questions, but I want to give you an opportunity to maybe hit on anything that we didn't cover that you think maybe the market hasn't quite understood or maybe there's some misconception or misperception. So I want to give you an opportunity...

Yes. Thank you. I appreciate that. I appreciate that. The thing I would love to close with is, I go back to -- for the AI market, Arista is known in the front end, and we've done very well, and we have developed product that has a great use case for the back-end AI clusters. We offer breadth of product for our customers in the sense that there is optionality and choice, and we feel that, that's what the market would like to see. And so when we're invited to proposals and have the opportunity to show this product, we feel we have a really good chance to win.

Great. I think with that, I've got 20 more, but they'll take 20 minutes. So why don't we end it there? Thank you, Chantelle.

Thank you so much.

Thank you, everyone, for joining, and have a great day.

Thank you. You too. Thank you.