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

All right. Let's get started. So I'm Aaron Rakers. I'm the IT hardware and semi analyst here at Wells Fargo. Thank you all for joining us this afternoon. Pleased to have Ita Brennan, the CFO of Arista on stage. I should mention you announced earlier this year the plans to retire. So hopefully, you're doing much more enjoyable things and seeing places, not coming to conferences next year. But congrats on a fantastic tenure at Arista.

Thank you.

So why don't I kind of just start the discussion? You guys just recently had an Analyst Day. You unpacked the longer-term kind of growth story for the company. I think you talked about a 2022 to 2027 CAGR of around 15%. Maybe just help us understand how you frame that. Do we think about double-digit growth out in the latter years of that framework? Just kind of unpack that kind of thought process in that growth profile?

Yes. I mean I think we started with just kind of stepping back and looking at the business, and we've grown so fast over the last couple of years that we grew almost 50% in 2022. We're on track for like a 33% growth rate this year. We've essentially doubled the size of the business in 2 years from 2021 to 2023. So even internally, we were thinking about kind of, what does this business look like over time? Because it is becoming larger and you have -- you almost have 2 different growth drivers inside of the company with the cloud, which tends to be more of a kind of cyclical grower. And then you have enterprise, which has been a much more steady kind of -- we're taking share. We're executing on a more kind of consistent gradual basis. So trying to take all of those inputs and think about what would be a reasonable growth rate. And again, not trying to pick the slope of the curve, right, because I think we just don't have the knowledge to be able to do that yet. But kind of looking at kind of that time period. And we have a very kind of serious focus internally around this idea of double-digit growth because we think that makes the model very powerful from an earnings perspective, et cetera. So looking at that and thinking through some of the initiatives that we have, looking at kind of how the market thinks about growth in those 2 pieces of the business. If you look at the market growth for kind of some of the cloud pieces of the business that kind of lines up with that and then thinking about kind of what we could continue to do in enterprise and campus over time. And kind of taking all of that and saying, okay, is it reasonable to think that you could achieve kind of that double-digit growth over that time period? Again, without trying to pick the slope of the curve, because I don't think we have the fidelity to try to do that yet at this point.

Yes. That's a good overview. And just to kind of think about that overview, and I'll touch on campus here in a minute and some of the other adjacent opportunities that you've talked about, but clearly, and I will talk deeper about the AI opportunity as you'd imagine. But you talked about $750 million target, 2025. That growth rate would embed that expectation. So I guess the question is like, is there conservatism embedded in the non-AI piece of the business to hit that double digit? Because it starts to become a big number as you look out to '25.

Yes. I think there's this idea of AI, non-AI. It's really important right now because it's kind of a way to see how the technologies kind of prove in, and it will show us kind of how -- but over time, I don't know that we're going to care that much, honestly, what's AI, what's not AI. It's going to be a lot of the same products going to the same customers. But in the near term, we kind of understand that it's important to everybody to see kind of the progress that we're making, right, as we start to make some progress, particularly on that back end kind of opportunity. And that's what we're trying to do. It's a tops-down number. It's not that we have perfect knowledge of what that number is going to be, but it is something that we're going to attempt to measure and provide you with some kind of milestones on just at least the best time frame to give some understanding of what's happening. But eventually, I think we're not going to care, right? So right now, is there a shift towards more of the AI-related spending on the part of customers? I think part of our framework for a while now, we've been talking about it for a while, is that we thought '24 would be a cyclically kind of moderated cloud year. I think that's still the case, right? There are definitely dollars that are going towards GPUs and AI. They have spent a lot on kind of, call it, classic, et cetera, over the last couple of years. Do we expect that we'll start to see some investment come back on this classic stuff as we go through the year? You hear Microsoft certainly talk about investing in both sides of that infrastructure. So -- but yes, we had kind of built in this idea that there is probably a slowing in, call it, classic, if you want, and that some of these dollars maybe get redirected to AI and then you start to have some new products that are addressing both in some ways in the longer time frame.

Perfect. Perfect. Campus. Maybe talk a little bit about the success you've seen there. You talked about a $400 million target. How do you think about that progression on the campus opportunity for Arista?

Yes. I think we're making good progress against our $750 million target. We were kind of somewhat light on revenue last year, mainly because of supply. I think we've helped solve for that as we've gone through this year. You saw that kind of in the overall kind of enterprise growth numbers for the year. I mean, there is a piece of that, that is kind of addressing that supply issue and then some continued kind of share gains and progress as well this year. The campus opportunity is big, right? It's large and there's about 50% of it that is large enterprise, which even from a go-to-market is right kind of in our sweet spot. And that's obviously the target. The first target for us, right, is to continue to execute against large enterprise campus opportunities. Sometimes they'll have data centers, sometimes don't. But gaining kind of that an opportunity and then expanding within those accounts is definitely a part of the progress that we're making there. Over time, we are focused on expanding the channel, but that has to be a very methodical kind of incremental thing, right, where we'll find new partners, we'll develop business models with new partners, et cetera. But it's not something that you can accelerate overnight. It does have to be methodical. And the more successful we are, broadly, in the campus or in the enterprise, then the easier that channel will become, right, because there'll be more interest in kind of carrying Arista as part of the channel as well.

Is it fair to say like, I don't know, you might look at the numbers. You have any thought around you win on campus, it pulls data center in the enterprise? Have you seen that? Or is it vice versa, like, hey, we've got a footprint in data center, and we're really starting to see the campus piece of the business get pulled from that?

Yes. I mean we had come into this thinking it would be mostly the latter, right? And that we would see -- because that's how we got there in the first place, right, was we see kind of data center customers kind of pull us to the campus. But we've been surprised at how many kind of campus-first or campus-only opportunities there's been, right? So there is kind of a -- there's an interest in -- from customers in kind of improving those campus footprints. And part of it is like a campus is no longer just an office, right? Although offices are a part of that, right? But you've got all sorts of other campuses as well, right? I mean we hadn't really played in kind of education or health care or manufacturing, those types of mini verticals prior to COVID. But with some of the changes and the more reliance on IT that kind of came out of that period, I think we see more of that now, right? And your campus network operator has a lot of pressure to maintain that IT infrastructure and optimize that IT infrastructure now. And that's kind of -- you saw Ken's presentation at the Analyst Day, but Ken's passion more and more as we get more engaged with these enterprise customers is how do you make the running of these enterprise networks, whether it's campus or data center, more efficient, right? How do you automate more? How do you give them tools that have been deployed in the cloud for a long time now? How do you make that available? How do you improve kind of the service that they can provide? And so he spent quite a bit of time and we'll continue to spend quite a bit of time looking at how we kind of tweak some of the software offerings, et cetera, to optimize that for them.

And that kind of intertwining of that strategy is CloudVision, how that portfolio fits together. And then kind of segueing from there into some of the other opportunities, routed WAN, security, observability, how do I think about success in those other adjacencies as we look forward?

Yes. I think the goal is to kind of take the -- if you have a data center, then data center all the way through to campus, but even within the campus universe, how do we kind of encapsulate as much of that as we can on an Arista platform so that we can manage it with CloudVision and provide visibility, et cetera, leveraging kind of that end-to-end solution, right? So what you're seeing us do is kind of fill in some of the pieces where we didn't have a solution. You can always look at that and say, "Are you better at kind of partnering? Are you better buying? Or are you better making?" And there's core kind of networking things that obviously we believe we should do, and SD-WAN is one of those, right? We feel like doing it as part of EOS and part of that solution, you can deliver something that's differentiated. WiFi was a good example of where our starting point was to buy and then integrate kind of the management plan of that into CloudVision. So we'll continue to do that, right? The NAC solution, AGNI, that we just announced, it was one other piece that was a pain point for customers to be handing off to somebody else, so you solution that. So I think we'll continue to -- the security piece is really interesting just because we have this data, trove of data from EOS, and then what can you do with that? And we're still figuring some of that out. But we are starting to use that in conjunction with some of the visibility and AVA and the solution that we acquired with Awake, right? So you'll see us continue to kind of tweak and add to that end-to-end solution, trying to always trying to stay true to leveraging something that we believe is really differentiated, which is the EOS and the data that EOS is collecting.

Right. So when I roll that all the way back and go back to kind of the 15% CAGR, what you outlined not this Analyst Day but 2 Analyst Days ago was really this Arista 2.0 kind of the platform story. A lot of those other things, routed WAN, security, observability, et cetera, are they kind of embedded in the expectation as you look out at that company? Or like but -- or is it more like if we're successful here, that's kind of additive to the expectations?

Yes. I mean I think all of this is what's going to allow us to be successful and continue to kind of -- I mean why -- because it's not trivial, what we're trying to do, right? We just go into a large kind of established market and drive share gains and target share gains. So you're going to have to be delivering something that's incrementally better to be able to do that, right? So I think these are all contributors to kind of why you win customers, why you win new footprint, why you win ports, if you like. Part of it, you'll get paid for by winning, right? Winning those ports. And these kind of management tools, et cetera, are a big part of the discussion that starts with the customer, and just making -- and driving kind of cost out of just operating those footprints, right? And then the other piece of -- I mean we are driving some software-only monetization out of this. Some of these are value-add features that we can sell as a subscription license, and we are doing that. And we'll continue to do that. But it's both, right? I mean both of those are valuable and kind of allowing us to continue to gain share in that campus enterprise market.

Perfect. So I'm going to kind of fine-tune on kind of the growth expectation. 10% to 12% was the growth guidance for 2024. I guess, obviously, always a lot of people have a hyper focus on the hyperscale cloud guys, your contributions from Microsoft and Meta. I think you talked about kind of large cloud plus AI now, that vertical being 40% to 45%. Does that embed -- does your 10% to 12% embed an expectation that Microsoft and Meta are 10%-plus customers? Or the total aggregate is 40% to 45% in '24?

Yes. We haven't, at this point, tried to break down 2024 between the verticals. I mean the greater than 40% was the 2023 comment. We haven't talked about whether those customers individually yet probably won't do that for a while, right? I think we typically kind of start to get to that level of granularity as we're moving through the year. But they are important, very important customers. We're assuming they'll continue to be important customers, but we haven't kind of broken that out at this point yet.

I have to try. I want to talk about the AI -- I'm going to bounce around here a little bit. I want to talk about the AI opportunity. I think you've done a good job a quarter or 2 ago kind of just helping level set people the cadence of how we should think about these larger scale deployments and the debate of Ethernet and InfiniBand and how that's going to play out. This Analyst Day, if anybody hasn't seen it, I'd definitely go back and look at it. I mean, you talked extensively about the portfolio and what's going to happen in next year. Just remind us again the cadence pilot trial, large-scale deployments, how Arista sees that playing out at this point?

Yes. I mean I think what we've talked about is kind of some trials and technology trials this year, some pilots next year and then production kind of deployments in 2025, right? And I think that there is a lot of technology work that's been done. I think we tried to expose some of that with Hugh Holbrook and Andy's presentation at the Analyst Day. So we believe that as these technologies get delivered, get established, that's kind of what's going to drive some of these chips in terms of deployments of networking. I think you're starting to see customers publish the white papers, et cetera, to talk about Ethernet and how they see Ethernet kind of as a networking solution for AI as we move through these product transitions, and it's performing well. And I think our expectation, obviously, is there's work to do, but we do believe that once you get to large-scale multi-tenanted footprints, then Ethernet brings advantages there. And it's just native to everything that's already been done in these large footprints as well. So there's just a natural kind of -- you have the expertise, you have the capabilities, et cetera. So I think that's -- but there's work to do, right? So that's why we think trials, like lots of discussions around what's the right way to evolve these architectures. Andy likes to say whatever we think today probably won't be exactly how it's going to play out, right? You will see shifts in the architecture. And again, the network is going to underpin that and building the best, most flexible network you can is really what's important for us. And then as long as there is traffic being generated, we should be able to transport that traffic, and that's kind of how we play in all of this.

So when I try and conceptualize what trials, piloted trials, I mean am I thinking about trials into next year being more like 1,000, 10,000 GPU kind of clusters? And then I know at the Analyst Day, one of the things that stood out to me was like you were talking like not just 30,000, but 100,000-plus GPU clusters. Even -- I think there was a slide that showed like 512,000 GPUs in a single Ethernet network. I mean, is that the size and scale we should be thinking about once these things really start full ramp deployments in '25?

Yes. I mean if everything progresses with AI in the way that we think it will and it becomes so impactful across all of these services for these customers, then yes, I mean that's what it's going to be, right? If you think about the scale of their footprints today at their massive scale, and this is only going to kind of compound that. So I think you will see these very large clusters and very large deployments. And what the timing of that is, we'll have to see, but that's -- assuming that this is really transforming business models across the board as we think it will, right, then it's going to cause those customers to need to have those deployments, right? Again, timing. It's difficult to tell exactly what that timing will be. But over time, there is this progression. And it's been part of their thinking for some time that this is a further evolution of kind of their footprints and their services. Some of the generative AI stuff is more prevalent now, but there's always been this move towards AI kind of embedded in some of those infrastructure deployments anyway.

So would it be fair to assume like characterizing your visibility into these opportunities, if we're talking that size and scale of footprint deployments that you would start to -- I mean, what -- are you messaging today like you're getting line of sight into some of that already because power envelope, getting the GPUs. It's not like this stuff happens in a quarter, right? You'd start to see that visibility certainly ramp up as we move through the next year?

I think the technology comes first always, right? I mean if we were sitting here today just talking about 800-gig, it would be essentially the same discussion, right? We have to align on technology and then deliver technology. Customers have to evaluate this. And then you'll start to see kind of what you can expect from dollars, et cetera, right? That execution on product always comes first, right? I mean it's not different, if you think back to the last cycle or the cycle before that. You have to earn, right, by delivery of products. And I think the alignment on technologies, et cetera, is happening now, right? Then there'll be more product delivery. There will be trials, and then it will go from there. So I don't know that we -- I mean, we don't have perfect line of sight to kind of what's going to happen in 2025 at this point, but we're executing on the building blocks.

Yes. I appreciate that. And I think one of the things that's been somewhat eye-opening to me is when you're talking about these big GPU clusters that you're talking about, particularly on the training side, having to work in conjunction, kind of have cohesive performance across all of these GPUs, that just the natural network and spend intensity is higher relative to your attrition. How do you characterize the spend? Like I used to think in the past like high single-digit percent spend on network versus compute. How does that look in AI?

Yes. And it may be a little bit more intensity, but I don't know that it's a huge shift, right? I mean, Jayshree had talked about 10% to 15%, but that was kind of including some of the optic spend. So that's -- so I think there is some increase in intensity, but I don't know that there's a major shift as we go through this.

Okay. And then I'll wrap up on maybe the AI stuff. What this question is, one of the things that I arguably still need to unpack and maybe appreciate a little bit more is, you mentioned 2 things at the Analyst Day. Distributed Etherlink Spine and linear pluggable optics. To me, it sounded like those were core building blocks of differentiation for Arista as we get into next year. I know you're the CFO, but I'm kind of putting you on a spot. Why are those important? What are -- what's differentiated about those?

Yes. I think the DES is going to be a very efficient fabric for AI, large-scale AI deployments, right? And obviously, the goal is lossless transport and using -- kind of creating a fabric that can handle very large AI clusters. On the optic side, I think it's not necessarily a revenue driver for us per se. It's a solution to the power envelope, and it's a contributor to solving the power envelope, right? So we've long kind of played a role in optics where we don't resell a ton of optics, but we're very vested in making sure that kind of the optics ecosystem is evolving at pace with everything else and that the overall kind of solution can be deployed successfully. And that -- a lot of focus on power in that. And I think it is potentially a big advantage from a power perspective, and it would be deployed in conjunction with Avago and their solutions. But really, the key there for us is that it helps to kind of solve for the overall power envelope of what a deployment would look like.

I guess I was thinking about it like if you can have a platform that supports that and you can reduce power consumption by -- I forget if it was 30% or even 50% relative to traditional optics. And I think importantly, it's backwards compatible, right? Like you will be able to deploy and say, hey, Mr. Customer, once these linear pluggable optics come in -- that is a differentiator. Would you argue that Arista's taking a leadership position on ?

I mean, I think it certainly enables us to deploy the overall solution, right, and make the overall solution better. I mean, again, it's not that we're necessarily looking to resell that in volume, right? It's more, again, create an overall architectural design that solves a problem for the customer and you'll get deployed more, right? I think that's always been kind of our approach to that.

All right. So I'm going to bounce maybe a little bit more to the financial model and some of the puts and takes on that front. So one of the things that, obviously, I think a lot of investors have asked you about is just this normalization process of purchase commitments, supply chain. I guess how are you seeing that right now? And ultimately, what I'm going to ask you is what's the right days of inventory for Arista? How do I think about how much you've had to lean in on the supply chain and how much inventory you carry? What's the right normal for you now as we move forward?

Yes. There's 2 pieces. Obviously, there's purchase commitments, and then there's kind of the inventory that we're carrying ourselves on the books, right? Stage 1 was kind of solve and push on purchase commitments as supplier lead times start to improve, right? And I think we've been doing that. We've been doing a good job of that. A big piece of that, of what's remaining, is going to be related to some of the key components that still have extended lead times, right? And we still have significant suppliers out at 52 weeks, plus or minus, lead time. That's going to cause that number to -- we still have some work to do. I think we can still bring it down as we go through '24, but it's going to hit a point where if those key components maintain that long lead time that it's going to be more elevated than where it was before. So we're continuing to drive that. I mean, we want to step out of the contract manufacturing supply chain because they should run that. We're making good progress to being at that point. And then obviously, the key components will continue to be engaged in. On the inventory side, it's something of the same. If you look at kind of -- you can see in the filings, the split between raw materials and finished goods. The raw materials is obviously much higher than what it used to be. Again, that's linked to kind of just how much inventory we need to carry because of lead times as lead times improve. We'll have the opportunity to improve that. On the finished goods side, I mean, we're carrying probably 30 more than we were more recently just because that was very tight. I think we still have some room where we can improve that as well. But I don't know that it changes that much if you look at it on a turns basis. So I think we'll improve turns slightly as we go through the year, and then it probably bottoms out there unless we get some improvement on lead times on key components, and that will help us to kind of improve that further. But that's going to become kind of the bottleneck, if you like, to driving it back to anything like where it used to be before. Because if you think back, those key components were 13 to 26 weeks in the old days, right? And we thought that was long at that stage. But if it stays at 52 weeks, then it's going to get harder to kind of drive that back to anything like what it used to be.

And you're still in that 52-week range in some of the...

We're still at 52 weeks.

So it's hard to put a defined number out there, I guess, is what I'm hearing for you. Like you were 335 days of inventory last quarter, if my math's right.

Right, which is not something, from a financial perspective, that I like. But then when you start to look at the pieces, I think you can improve it from there, but I don't think we'd get to your -- I think you had suggested 200 or something in one of the questions. I don't think it goes back to that, right? I think you get kind of held up by the key components and until that changes, then if that improves, we'll have an opportunity to kind of drive it further.

So one of the other questions I was going to ask you about is the 10-Q, and you mentioned the inventory breakdown is interesting, the purchase commitment color and stuff like that. But one of the other things -- I think I've asked you this offline a couple of times, but I'm going to ask you again here is that one of the disclosures is this idea of binding contractual agreements related to future product shipments. And I think that balance was $507 million exiting this last quarter and exiting '22 is like $960 million. What is that? I mean is that -- that's obviously future deliveries of products.

Yes. So we had -- largely our customer orders are not contractually noncancelable, right? I mean we've -- as we went through the supply chain, we agreed with customers, we vetted kind of their -- what they were trying to buy, et cetera, et cetera, and we got to a pretty good place. But contract -- pure literally contractually binding, we didn't go and reformat all the contracts to try to do that. There were, though, some new opportunities, new projects, et cetera, where we did, because we were doing new paper. We had an opportunity to do new papers. So that number will kind of go to zero eventually, right? It's just as we continue to kind of process through those and some of the newer deals won't be on kind of contractual -- on a contractual basis. So that number should eventually kind of head towards zero.

Okay. That's helpful. 800 gig, and one of the things that we've been thinking about, and Broadcom, I think, made a comment last quarter that their Jericho3-AI and Tomahawk 5 silicon would ship over the next, I think, it was 6 months or something. Have we seen any shipments of those platforms at this point? Or is that like 800 gig, that 51.2 silicon, that's really a '24 dynamic?

Yes. We haven't announced any of those products yet, right? So -- but that's obviously underpinning a lot of the -- when we talk about new products in 2024 and beyond, it's a -- that's [indiscernible].

Okay. In the couple of minutes we've got left, I've got a couple of other quick questions. So services and support, 16.8% of revenue last quarter. I think you guys reported in the P&L like services revenue, 14.8%. The delta between that is software. I mean, we're starting to see -- you mentioned earlier, software -- is there a software monetization dynamic to maybe start to think about?

Yes. It's probably even a little bit more than that just because there is some software offerings that end up in the product on the P&L, right? Yes, so it's a little bit more than that spread. Yes, I think the software revenue kind of -- it gets dwarfed when the product grows the way it's been growing. But it's been keeping pace, right? So which is not nothing, given how fast we've been growing, right? So there is -- I mean there are this set of value-add software subscriptions like CloudVision, some of the visibility tools, some of the Awake -- ex old Awake capabilities, the new -- that have now been incorporated into CloudVision, et cetera, AVA. That all kind of gets sold as a software subscription license. And it's -- again, it gets dwarfed by the product sometimes, but it's a nice growing kind of software revenue stream, and it is contributing kind of to gross margin, et cetera, and it's helping kind of drive that enterprise campus piece of the business as well.

That's perfect. So the other quick questions in the few minutes I got left, so last quarter, there was no share repurchases, right? You guys generate a tremendous amount of cash. I think you had $4.5 billion of total cash. So -- should I read anything into that? How are you thinking about capital allocation? And ultimately, I'm also going to ask you about M&A. Like how are you thinking about the strategy for Arista on that front?

Yes. I mean we talked about this a little bit at the Analyst Day. I mean,we'll probably always carry more cash than maybe like a strict financial model would indicate just because, for things like the supply chain, right? It's important for us to be able to kind of step into something like the supply chain and carry as much weight as some of our much larger competitors, right? But that said, I mean, it is our intention to return cash, and we've talked about kind of offsetting dilution, returning 50% kind of our -- of the cash that we generate. That got a little -- with all of the inventory stuff, et cetera, we kind of backed off that a little bit. And then the other piece of this is just kind of how do you approach that? Are you -- because the stock is volatile, even -- mostly for reasons beyond our control, actually, like it just moves a lot. Do we try to optimize for that as well, right? And I think you saw us do that last year, where I think it was Q2, we were very aggressive and active on the buyback because there was an opportunity to do that. I think we'll continue to do that. So it is our intention to return cash, but we'll probably remain pretty opportunistic in terms of when we do that. And if there is volatility, we'll try to leverage the volatility.

And on the M&A front, is it -- I mean, I don't think about Arista as being kind of a platform acquirer. It's really got to fold under the umbrella of EOS, maybe CloudVision. It's more just kind of tuck-in strategy from an M&A perspective. Are you seeing any change in that thought process as the company has evolved?

I mean we look at stuff all the time and will continue to, and everybody is very connected into the ecosystem. But because of the platform, because of kind of the -- just the leverage that EOS brings and the opportunity that's there to kind of just organically kind of capitalize on that, it tends to drive us towards kind of technology, people, teams, smaller M&A, right? But again, we look at things all the time. I mean I think the team is very well connected into what's happening in the industry, and we do look at opportunities all the time. But that is kind of something that you have to solve for. You'd have to -- it raises the level of conviction you need to have, right, in terms of doing something bigger.

So one of the other questions on -- I feel like a lot of times, people think of Arista and they think about Microsoft, Meta, cloud, big type footprint deployments, but this enterprise momentum that you've seen is, in my opinion, underappreciated off that. One of the things last quarter, you grew EMEA revenue, I think it was 56%. You grew Asia Pac revenue 72%. Are you starting to hit a different stride in some of the international markets? Are you investing more there? Like is that starting to show up as a bigger growth driver for the company?

I mean, we are, for sure, investing internationally in sales headcount, et cetera. We are deploying internationally. I wouldn't overfocus on that quarter though, because some of that is still shipments, kind of supply chain-related, the patterns. I mean it's been kind of -- our international percentage has been kind of stuck at the 25%. But again, that's -- the overall is growing pretty aggressively, right? So those business in those regions is growing, too, right, and growing healthily. It just hasn't shifted the mix. But I think the quarter-over-quarter that we saw is probably more just kind of just shipment -- as much shipments as anything else, I think, in the short term.

So the final question I was just going to ask you is kind of just an open-ended question is as you talk with investors, you talk with myself and others, is there any attribute of the Arista story that you feel is not understood enough or maybe asked about enough? Because everybody is focused on AI. But is there, in that context, some things that you just think people don't pay enough attention to?

I mean I think the story is pretty well understood. I think the power of having kind of these 2 pieces of the business -- people focus on AI and on cloud, which is great. And it's a very important -- and they're very important customers to us, and we want to do as much as we possibly can with them. But then the offset of kind of having this enterprise, campus business, I mean, it is very -- over time, it can become a very healthy offset, those 2 pieces of business, be it at gross margin line, be it at kind of a growth offsetting some of the cyclicality. So I think that piece -- and we need to continue to execute on that, and I think we've been making progress. But if we can really do that right, I think those 2 pieces of the business kind of, as an offset to each other, really does help the model overall.

Perfect. I think we'll leave it at that. Ita, thank you so much for joining us.

Thank you. Thanks for having us.