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

Good morning and welcome to Needham's Technology, Media, & Consumer Conference. I'm Ryan Koontz. I cover the networking and optical sectors as well as the space sector here at Needham. Really excited to have Arista Networks with me today for a fireside chat. We've got CFO, Chantelle Breithaupt. Welcome, Chantelle. How are you?

I'm good. Thank you. Nice to see you, Ryan. Happy Friday.

Yes. Happy Friday. Well, let's dig right in. Arista just had results, feels like to me a couple of weeks ago because it's been so crazy. But that was just a couple of days ago. Let's start with a review of your first quarter and what some of your thoughts were. What were the main takeaways you want investors to highlight from your results?

Yes. Thank you. Jayshree and I were pleased with the Q1 results. We thought it was a very solid print to the quarter, 27.6% revenue growth, solid margin and I think very robust operating margin. So I think, pleased with the momentum in the first quarter. And I think you could see it through some of the customer wins that Jayshree spoke about, some of the things we're seeing with the AI customers and cloud. So super excited, and that turned into a pretty robust Q2 guide as well.

Yes. And the margins there, nice uptick or steady and strong on margins. Is that related to customer mix, product mix? What were some of the mix results there?

Yes. No, we entered the year with a 60% to 62% gross margin guide, so coming in 64%. It was customer mix-related. Our products generally don't have mix between them, but the customer mix, we were pleased to see some of the upside coming from enterprise and our cloud specialty providers. So across the board, pleased with the growth that we saw.

For sure, for sure. And relative to your guide, you had a very healthy second quarter guide as well. Are you feeling good about that outlook, at least in the next 90 days before the tariff uncertainty hits us?

Yes. Generally, we see momentum for the year coming through the top line, the second quarter. We chose a point versus a range this time just to be sure that the outside world understood our confidence of that $2.1 billion. And so that was very intentional by Jayshree and I. So yes, we're excited. And again, we're excited across the various segments that supports the 63% gross margin guide. So yes, we're having -- you saw the reiteration of CapEx from some of our largest customers. And so generally, the conversations are going very well.

That's great. And then you maintained your year, which implies maybe a flattish second half from first half with maybe the uncertainties through the 90-day tariff passes. I mean, what was your thinking on kind of maintaining the year at this point? And investors may be a little disappointed with that outlook, I mean, considering all the strength you guys have been delivering for many years.

Yes. I think there's some context here. We're very aware of what the math says in the second half, but we wanted to give context. So we entered the year at 15% to 17%. We raised it to 17%. We had a very robust Q1, guided a strong Q2. We're not changing the year yet, which isn't that different from Arista's general philosophy on guidance, until we have some of the clarity on the cost side. So the tariffs to us aren't necessarily a top line conversation. It's how do we bring the whole P&L together for one connected updated guide. So we see momentum continuing through the year. There's no scenario where we see a drop-off in the second half. So I think the full year update will come once we get the pause, the fentanyl, the immigration, the reciprocal tariffs sorted out. And then we'll come with one guide. So that was a choice. We could change our guide, drop our guide or hold the guide. So we tried to give the wording through our language on the first half. We see the momentum continuing through. So we look forward to a very good year, me and Jayshree.

That's great. I mean in the big scheme of things, that's a real win. I mean a lot of companies withdrew guidance or really withheld comments and said, "We think it's going to be okay, but we really don't know". To maintain the guide, I think certainly is a win there. Relative to the high proposed tariffs, I think, that have a lot of the entire tech industry a little bit on edge here, you're not alone. Do you anticipate that you might have to raise price in exchange for that and that raised price could impact demand volume? I mean, what are the different scenarios that could play out for you here? Walk us through the different options as -- investors should think about in the second half.

Yes. And thank you for that question because we put a lot of work to be as clear as we can on how we feel the tariffs will unfold. So the first point I'd like to go is we've kind of bookended the impact if all the reciprocal tariffs came into effect, if USMCA held as it was. We had said on the year, it's 1 point to 1.5 points of gross margin if we did 0 mitigation. So that's, if you want to call it, worst-case scenario. And if you're running at 64%, 63% and the guide 60%, that's why we kind of held the guide because it encapsulates that. It doesn't mean that's where we're going to end. So we have 3 things we're working on. We're working on the fungibility of our operations, like everyone else is. We've had great conversations between Malaysia, Vietnam and Mexico. The second one is how much can we absorb to have the right commercial conversations with our customers. And third is, we will have a tariff increase for prices if we need to, but it's kind of the third on the 3 of the mitigation strategy points...

Yes. Of your outputs, yes.

Yes, yes.

You work your way down from the top-down there?

That's right. That's right. So we're pretty clear. We've done a lot of really granular work. So I feel confident coming in, we have a handle around it, assuming nothing changes with the Section 232 ones. Those are the only ones we don't have information on.

Got you. And any thoughts about accelerating production ahead of July 9 at all in terms of like getting product here to kind of ease that transition ahead?

Yes, absolutely. So the other thing you did -- our inventory went from $1.8 billion to $2 billion quarter-over-quarter. Most of that, a lot of that was to do with buffering, to your point, Ryan, in the sense of bringing components or finished goods in, yes.

Yes, yes. Perfect. Exactly. And then relative to customer behavior, so what are you broadly hearing from customers about the situation? And are you seeing evidence of customers that want to pull in or might be pulling in a little? What's been kind of the tone relative to that [ metric ] in July?

Yes. I would say, generally, the statement is not a lot of conversation to try to get things earlier. It's probably less than one hand of customer account, discrete accounts where we've had conversations, so very bespoke. So not a lot of customers asking or suggesting or feeling that change in their pattern for this year. So a few, but not material. Now I don't know if that changes as we get closer to July. But I think generally, people understand their spending for the next couple of quarters. So...

Yes. Great. And you mentioned you've got a lot of flexibility in terms of working with your CM partners about location and setting up in -- Mexico and Southeast Asia seem to be the the favorite flavors of this decade. And you've got flexibility on both sides there?

Yes, we absolutely do. A couple of things is we've had great relationships with all of them as we've come up in the 10 years since the IPO. So we've established great relationships. We're also of a size now where those contract manufacturers would like to do business with us. So lots of great capacity-exploring scenarios in all 3 areas. We're close to having most things USMCA-certified. There's a few things we're just working on. So generally -- yes, generally, we're aware -- now no matter what that fungibility is, it still takes 6 months to reorient the supply chain. So we're trying to balance bringing inventory in and not making changes too early that we wish 6 months from now we hadn't made.

Yes. And what are some of the activities there? You mentioned that Mexico -- certification for Mexico production. What sort of things are those related to?

It's just a few -- there's just a few products that need a couple of subcomponents to get 100% that's -- it's just around the edges, yes.

Content-related stuff?

Yes, that's right. That's right.

Got you. That's great. That's a great update, And I'm glad you guys are feeling good going into this. A lot of companies aren't as lucky as you to have the purchasing power and flexibility and the nimbleness, so that's terrific. Maybe shifting gears here and talking about AI, which every investor, I get peppered weekly. As for the #1 topic of what's happened in the AI spend and infrastructure, and we could probably talk for hours about it, but you've talked about your 4 really important pilots that can -- that have potential to scale to 100,000 GPUs and your $750 million target for back end for this year. Can you update us on those projects, and from your perspective, how they're progressing and how investors should think about them over the balance of this year?

Yes, happy to. So let me start with the punch line: we reiterate and are confident in the $750 million back end and the $750 million front end, so the $1.5 billion combined. Now talking about those pilots specifically. So we have 4 pilots and then a bunch of different various enterprise CSP kind of AI things going on. So I think the 3 out of the 4 -- so 2 out of the 4, very confident, 100,000 GPU range to close out in the fiscal year to get to production. The 3 out of 4, it's going to be Q4, Q1 just balancing on the edge, but again, 100,000 GPU. The fourth one is an InfiniBand shop that just started. So they'll be more of a '26 going through the stages. But generally, what I would say is great partnerships. They're all doing it a different way trying to get to the same kind of goal. So we're learning a lot of partners to them as vendors, suppliers, extensions of their engineering team. But we're very excited about what they're learning. But it takes some time. But once they have these up and running, they themselves will be able to rinse, repeat and hopefully working with us as they continue to build out their CapEx plans.

That's great. I mean it sounds like you're really positioned to gain some share in the AI back end here as spending continues to shift and seen -- there was a lot of investor skepticism, I'm sure you heard, coming -- in the first quarter about spending, where it was headed, are we -- have we overspent on AI infrastructure. And every company came out, including yourselves, and said, "No, it's -- demand has remained strong." You feel good about the overall market opportunity for you in AI and the budgets out there?

Yes, we do. I think we do. Yes, it's always nice when some of our largest customers go first before us in earnings so we can be sure what we're -- but yes, we do. We feel just as much momentum as we saw last year in the sense of the types of conversations. I think -- this is a hypothesis, but I think we're starting to realize AI might be secular, not cyclical, because I think there was a little bit of that going on. So I think -- that's not for me to determine, but that's what it seems like. And so there's this really great kind of outlook by the 650 Group, and they talk about we're kind of in year 2 of 3 of content creation and models. And then we're going to go into this Agentic AI, which is supposed to be just south of $1 trillion of spend, and then you get into autonomous. So there's almost like an 8-year road map of AI-related topics and scenarios and use cases. So very excited on that. And I think for the back end, because we have a lot of choice, products for our customers, I think that's -- when the option is best of breed, we have a very good chance of winning.

Yes, for sure. I mean, especially upmarket with the biggest cloud builders, they -- I'm sure they tend not to take a turnkey NVIDIA solution. They want to go best of breed and they want to do it their way. They're sophisticated enough to do that. So yes, great presence.

Yes. Thank you. Yes, we agree.

That's fantastic. And then on the front end, you mentioned briefly there the $750 million target. I'm sure it's a little hard to always parse out exactly what is in the front end versus what's in a traditional cloud model. How do you measure that up? And what kind of trends are you seeing relative to the front end on these AI-focused programs?

Yes. So it is a bit difficult to parse it out, to your point. But generally, I would say, the fact we are reiterating $750 million on the front end, we're very excited about that. We see customers realize that it's going to take both to fulfill the needs of the end-user case that they have. I'm pleased that on the industry data, Arista shows up as the leading vendor, not just branded but leading vendor when it comes to front-end Ethernet AI revenue. And so we very much look to keep that spot. I think the choice of our products, the learnings we get from some of our customers, we can pass through the industry. We're very excited about that. And at the end, as you kind of move into Agentic AI versus the model creation, it might open up this aperture to be -- you know how Jayshree talks about 30% to 200% kind of additional front ends needed for every back-end dollar?

Yes, yes.

I think that the 1:1 right now works well, but I think it could get north of the 1:1.

That's great. Super interesting. Yes. And in terms of your big customers, Oracle has really emerged as a big player, not just from Arista, but a lot of the other players -- I was just [ with ] their management team talking about how Oracle is doing. And they've really come on in the scene strong here. How are you -- how have you been successful at Oracle there? Any ingredients to your success at Oracle? And I'm sure each of these big customers has different kind of architectures and requirements. How do you align yourselves so tightly with these huge players in the world of cloud?

Well, I appreciate that. So the thing I would start with, having been at Arista just over a year now and been at some of the other companies that are much bigger in the world, we're just 100% pure-play networking-focused. And so you feel it, all 4,500 people wake up every day just to think about networking. So if you have that in mind, no matter what we're working on, it's to solve those things for customers like Oracle. That's all we're focused on. And so if you look at the breadth of product that we have, we have a lot of choice for the back-end Ethernet AI. We have a lot of choice for the front end. And so I think we have the products, and then it's just how do we adapt it to the use case, Ryan, versus not having that fungibility or flexibility across our product line. So that's the first thing as a product. The second is the software, the OS. Customers, if they're familiar, it's easier to grow with knowing what there is. And the third thing is I think we just have really focused engineers in our team that partner with them. And so they feel it's an extension of what they're trying to do, and they really appreciate it.

I mean, I think EOS is -- every company thinks their tech knowledge is the best. But from my own work in this space, I mean, EOS really seems unmatched in the industry in terms of its uptime and modularity and all these super-important factors that I think other companies just cannot match it.

Well, we fight hard to keep that spot. Thank you. And the fact you can upgrade as you're operating is a big deal for a lot of our customers [ who happen to ] have a lot of planned downtime. So all those things, but we continue to not sit on our laurels. We fight every day to make sure we're staying in that engineering kind of mind space with them.

And is Oracle a relatively new large account to you coinciding with it being added to the cloud titans that it just got -- you were so successful there that you wanted to move them into that group?

Well, there's a definition for titan, right? So it's the amount of connectivity they have. So they get to distinguish that based on how they operate as a customer company. But for us, I think that we're thrilled in the sense of the amount of business we have with them, it suits this kind of category. So it makes sense to us, yes.

Awesome. Yes. And the other big question besides AI that I get, and I'm sure you get constantly too, is competition from white box. It's always top of mind from investors. Can you kind of walk us through your view of where white box sees the most success within the big customers -- the big cloud customers, the titans and where Arista sees the most success and you have your barriers to entry that you think are formidable?

Yes. No, I think it's an important question, I think, right here and right now. Although I'll say the topic ebbs and flows depending on what's kind of going on in the new cycles. So I think that white box, if anything, I think that our customers are even more certain now having worked 10, 15 years at Arista, 10, 15 years coinciding with white box, we've always coexisted. I would say that the great thing is some of our largest customers are super clear and we work with them. This makes sense for white box, this makes sense for Arista or a branded vendor. And so I think we're pretty clear when we're having customer conversations. We're not surprised. They're multi-vendor strategies. There's use cases. And you've heard Jayshree mention on the earnings call, there's use cases for white box, and you can imagine what they are. It's low margin, high output, not complicated. If you want to do the more complicated things, the more robust things, the higher outcome, premium value and engineering, then it's a branded vendor like Arista. So for us, nothing has really changed. This market is growing at such pace, everyone is going to have room to grow. And so we don't see it as a share shift. We see white box kind of share shifting with each other. And so that's for them to have. But as far as barriers to entry, the barrier to entry to us is we don't want to be a 10% margin business. That's not what we do. And so we could choose to play there, but we don't, obviously. And that's a choice as a business model. And so I think that -- to me, it's very clear on the distinction.

That's great. Yes. In the big picture of AI, kind of going back to kind of where we are in the cycle and we start the conversation there, there's been some fears that we were maybe at a peak, and there was going to be some consumption issues around overbuild of capacity. Where do you feel -- based on your demand signals, where do you feel like we are relative to demand and supply of AI infrastructure?

Yes. I would -- if I had to bet, I would say we're nascent. And I do not see overbuilding. I see that the pace is quickening in the sense of a lot of these players. And it's not just the few we talk about. There's a lot in between the neoclouds. There's some of the larger enterprise getting in, some of the providers who provide the AI networks. And so for me -- for us, we don't see it as a -- at the edge and maybe we're ending. We see it as kind of just the beginning of the next few years at a minimum. Now how the CapEx moves, I don't know because we don't know '26 CapEx yet. But generally, the conversations on the road maps, it feels like the beginning.

Yes. And the neocloud types, the Tier 2 types, it seems like that's been a solid segment for you historically, maybe not quite as fast growing as the titans. But relative to AI, it feels like very early days. And are you in the early days, you think, of penetrating that emerging segment? Do you have a similar position, you feel like, in -- with the neoclouds?

So I do think that generally, this market takes longer. There are a few that stand out, of course. But generally, this is a space that's going to take longer, either through their funding, their business model, their talent. So I think this is just starting to go. We have great conversations. When the customer is looking for choice of design, we have very good conversations. If they're looking for turnkey or are asked to look at a turnkey kind of bundled solution, sometimes we're not invited to those conversations. But we're seeing a lot of them wanting choice across all the parts of their neocloud venture. And so it takes a little longer, but we are very much in the conversations, very excited. And I like the fact that these are more international conversations, which is great. It's not -- it's definitely not just in the Americas, as you can imagine. So yes, very excited. And hopefully, they get their teams and their designs and things up together, and that's what will dictate, I think, how fast they grow.

Yes. It feels like the market really is pushing the bounds of how fast it can grow, right? There's only so many -- so much talent, so much available equipment, so many leaders in this space that can only install so fast. So it feels like we're pushing the boundaries of that growth limit, including [ maybe ] like power and data centers, all...

Yes. That's what I said and mentioned, yes.

So lots of limits, and we're growing at this. And hopefully, that limit is enough to sustain a relatively steady growth rate and not have too big of an oversupply at some point. But...

Yes. Totally agree, totally agree. But the other thing I would say, too, is that part of what we're also excited about is that just regular data center spends happening. Campus is growing for us. We're just getting started in campus. So even with the diversification, hopefully, they all cycle through in a great way, we can keep the growth rates we hope to.

It's pretty amazing. Up until a couple of years ago, all we've -- all I've focused on was traditional cloud, and that was the shiny object. And everyone forgot about the shiny object.

Yes, yes. Exactly, exactly. It's fascinating how quickly we can pivot with recency bias, yes.

And that old business is pretty solid and it continues. It's a profit engine for you guys, I'm sure. Just keep cranking, you got customers, you got share. You're not having to prove yourself in that market, evolve, but not disrupt necessarily.

Yes. So that's absolutely a share gain conversation, which we don't take for granted, we very much focus on. But they're great conversations. And so we look at all those revenue drivers, Ryan? Just so -- everyone thinks we're not just focused on AI, we are, but not just that.

Totally. Back to neoclouds here. It would seem that maybe they're less sophisticated in general. And their early deployments maybe would be more of the turnkey NVIDIA approach because they don't have the expertise to go best of breed and they kind of would need to make Phase 1 -- kind of simplify before they go best of breed. Is that -- have you seen that behavior somewhere where NVIDIA is effectively competing in some of these early nascent players at early stages of their evolution?

I would say yes and no. I've seen -- we've seen both. I definitely have seen both. And not surprising, it's basically usually us and NVIDIA in the conversations that we see. So it makes sense. There's a bit of both. There are absolutely ones that are looking to diversify away and some that -- for them, that seems like the best option here and now. We always remind them, "You have more choice if you go with Arista in your design." So -- but yes and no. But the market is growing. Like we talked about, it's just starting and should grow pretty well. So we'll go in and look for our fair share of it.

For sure. That's great. So the -- we've -- I've taken also a lot of questions about some of your balance sheet items around purchase commitments and deferred and is -- does the AI business cycle for Arista look differently on that balance sheet compared to your run rate regular cloud business? Can you walk us through that, maybe changes in deferred, which seems to be growing quite a bit?

Yes. So deferred is a combination of new products, new use cases, new customers. And so we have all 3. And so when it comes to new use cases, new product, that's the AI Venn diagram. And that's some of the biggest, most complicated things that our customers are doing. And those things take 12 to 18 months. That behaves very differently than new products, existing customer. That's probably more 12, 14 kind of months. So yes, bigger acts differently. And I've tried to be clear in the earnings that there's going to be some swings because there are larger things that come through. Volatility, the amount growing, that makes sense to us in the sense of some of these use cases. So all of those things are in there. And I think it just represents Arista getting to a new kind of step-function with a new set of markets. So -- and this is the same on the inventory purchase commitment side. That's absolutely -- we have a long lead time cycle on our chips. So that -- if we're going to be ready for some of these things, we need to be ahead of that. And that's always been the case there. But very comfortable with what's happening there.

Got it. So it sounds like the composition within deferred might -- maybe changing a little bit because there's so many AI going on -- so much AI going on? So...

Yes, yes, yes.

That makes sense. That's great. And then as you think about your '25 growth here, 17% you reiterated, how much of that is AI-driven? How should we think about the growth rate of traditional cloud and the balance of your business? From a top-down perspective, how would you kind of parse out the growth?

Yes. I think that -- so we haven't changed the year, but sticking to what we know now, but see the momentum in the first half. I would say the AI, we've said, is at least 1.5. It's a little difficult to know more at the front end, but at least that 1.5 front-end/back-end AI. So let's call it the 1.5. Now we're seeing some great growth in campus. We're excited about the data center refreshes. So we'll see what we come out with as our revised guide. But generally, I think those are all working well. But the AI specifically is that 1.5 target that we can kind of measure.

Got it. And -- that's great. The -- and then in terms of competitors, we haven't talked a lot about Cisco or Juniper too much. Do you bump into them much in the new AI opportunities? Or are they more competing in traditional cloud?

Yes. That's more of a -- that's actually -- so for Cisco, it's more of the enterprise conversation, same with HP-Juniper. And I would say we've -- I'd say the only thing that we're seeing, and I think you've heard it mentioned before, is Juniper seems very much to be in a price situation in the sense of where they'll go from a pricing perspective. So that behaves a little differently but not surprising, I suppose. So I think that's -- nothing really changed there. When it comes to the AI back end, that's the NVIDIA and Arista conversation. So those haven't changed a lot. So I feel that we're pretty aware and know what we'd like to do in those scenarios. And I think it comes down again to -- for the back-end AI, it's would you like a bunch of choice, best of breed or would you like one bundled conversation. And so the customer needs to decide at that point. We obviously have our view. And we continue to introduce great products into the enterprise and the front-end AI. So we'll continue to do that.

Exciting. And you mentioned campus there as important. Can you maybe unpack that a little bit in terms of your progress to date? How do you feel like you're doing relative to your share you aim to gain? How much work still needs to be done, either on products or channels to really execute there? It feels like a little heavier of a channel sale model there. What can you tell us about kind of the campus opportunity for you? It feels like it's taking a little bit of a back seat to all the AI excitement.

Well, I think it's -- so I will say internally, I spend as much time on campus as I do AI because I understand the long term of what this means as a foundational revenue diversification. And so a couple of things on campus. One is of our $70 billion TAM, it's roughly $16 billion. We're looking at $750 million. I would say -- Jayshree and I talked about this. This is kind of the first year where all cylinders are firing for campus. We kind of introduced it through COVID, then we had to do some portfolio adjustments. Now we feel the hardware, software, the portfolio, we feel like we understand from a who's working on what engineering-wise inside the company. We're clear on how we're going to go to market. So I was very pleased with the Q1 growth of campus. We don't disclose that quarterly, but you can just know I was pleased with it. So we feel we have all those. And the way that's vindicated or validated for me is that we're winning campus first and not small campus deals without even being in the data center. So for me, that's a great signal that Arista is in campus now. And so we're -- it's a longer sales cycle, to your point, Ryan. We are working with the channel and partners more intently for this particular segment. So we've been putting money there, investing in salespeople. So I hope everyone's excited that Arista has declared we're getting into campus because we're ready to.

Yes. That's great to hear. I mean Cisco is the gorilla there. And they seem pretty beatable these days.

Well, we're going to definitely do our bit to get that share that we want.

Yes. Juniper has been doing pretty well, too. And that's in campus with Mist.

Yes.

Great. In terms of wrapping up, anything you wanted to close out with, a message for investors here and finishing our conversation?

Yes. So just to wrap it up and remind, as you guys exit to your weekend and many other calls today, remember the $70 billion TAM Arista has ahead of us. Recall that Jayshree mentioned on the earnings call, we see $10 billion earlier than she originally thought at the last Analyst Day. We're excited about AI, traditional data center and campus. We have clarity on tariffs. So we've tried to be as clear as possible to give you bookends. And we're very excited to get to our next earnings call, talk about more innovation and give you what our new guide is.

Great. Well, thanks so much for joining. Great conversation, Chantelle. Thank you.

Thank you. Take care. Bye.