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

We good? Great. Folks, thanks for joining us. My name is Simon Leopold. I'm Raymond James data infrastructure analyst here in person live at our conference in New York. It's nice to see people face-to-face again, it's been a while. We've got with us this morning, John McCool, who is with Arista and is the Chief Platform Officer. John? Thanks. Good to see you. I think we saw each other not that long ago at Arista's recent Analyst Meeting.

Yes, it's good to get the face-to-face segment.

It feels a bit odd, but it's nice. It's nice. So I've got an outline, we'll do sort of a fireside chat, but if anybody has questions from the audience try to get our attention, we'll repeat questions for the webcast. I guess I want to sort of start out with some of the high-level stuff and drill down. But Arista provided, I think, a surprisingly good forecast for 2023 for 25% growth. Honestly, I went to the meeting with kind of a mid-teens number thinking, okay, we'll probably go up to 20% so 25% was a positive surprise. Maybe help unpack what sort of led into that? What are the contributors? What are the key drivers? Do you feel that's conservative? What's influencing essentially to build up to that model?

Right. So I think just starting with 2022, I think we've seen a good rollout of our 100, 200, 400 gig solutions in the cloud segment. And there's clearly a cycle there that we're participating and benefiting from that growth. We talked about at our Analyst Day, not just the traditional network in the Cloud, but also new use cases, specifically around AI clusters. I think this is interesting. It's been a long time since new things got connected to the network, and I had to think back to maybe IP telephony, where you had noncompute devices connected. And now we're seeing GPUs that I think people would have thought about a closed system, maybe around InfiniBand being connected to Ethernet. And these are large clusters that have to have a close affinity with each other to do their thing so that's exciting. And then I think the we continue to drive the enterprise business with the campus offering as a new use case in the enterprise network, and that's building momentum and gaining new customers.

And so historically, Arista's sort of segmented the verticals. Maybe talk a little bit about what you're expecting trend-wise from each and how big they are. So the cloud titans have been the biggest, you've mentioned enterprise, but you've got some others. How do we sort of what's the relative contribution?

Yes, I think about the cloud segment that the Cloud Titans being the largest, and these can be 100x the number of servers then the next portion down. We think about the service provider and web specialty tier, which is also growing. I think we've seen a lot of the smaller providers of the Titans building out the same type of infrastructure moving to 400 gig and building out their networks and then the enterprise segment, which we've built that up to the substantial size.

And when we think about the expectations for '23 and the Cloud Titan trend, how much of this is really working down the backlog that you've established during '22 in part because of supply chain constraints and how much of this would you characterize as sort of incremental business? What's sort of the split?

Yes. I think we've had good visibility with the increased lead time, which kind of, in some ways, distorts a backlog metric because you have this variable time pace that makes that less clear. So these are projects that we've been involved in with the 400-gig cycle. So we have good visibility into those projects and the rollout in 2023.

So I actually received an investor question as I was walking in that fits right here. It is, is it conceivable or possible that some of those cloud customers have been building up buffer stock or building up inventory, sort of how smooth is the outlook?

I would generalize the answer. I think in this environment, it's been so supply constrained that people are really jumping at the bit for deployments in what we can ship and that continues not just around the cloud segment, but the enterprise as well.

And then one of the other sort of related debates is this question of quality of backlog. And so it's the idea that, yes, you have the backlog, but couldn't it go away? How do you go about assessing that?

I think, again, focusing more on rollouts and deployment as well as the projects we're participating in and getting focused on that and when do actually people need the product and when can we build it and when do we get supply?

And I wanted to maybe unpack a little bit around sort of use cases and you mentioned one, the AI machine learning. Is that exclusive to the Cloud Titans or is that a trend that you're seeing in your large enterprise customers and maybe what are some of the other use cases beyond that?

Yes. I think we've looked at the Cloud Titans in our business as really leading the definition of new architectures. And over time, we've seen them apply outside of the Cloud Titans to the specialty cloud providers as well as enterprise over time. So if we look back, we entered routing probably 5 years ago, it was really the cloud titans that wanted to interconnect data centers but operate them as regional data centers as one logical network and drove us into the routing segment. We took the routing technology and brought it into the enterprise, so we could do data center to data center and then ultimately into service providers. This trend, I think, is early in the cloud providers, but over time, I think we would see similar kind of architecture as well as the enterprise. So not there yet, very relevant in the cloud titans, but I think it's an interesting use case to apply across a number of verticals.

And I think if we look back, enterprises are substantial part of your business, remind me, percent, 35%, 40% of revenue.

Roughly roughly next year we can talk about that.

Not the first specific quarter. But and given that as kind of a reference point, and I'm having this discussion with every fireside chat, aren't we worried about a recession, aren't we worried about slowing from enterprises and candidly, the feedback I've gotten has been quite varied. So I'd love to sort of get your take on the recession fear and what it might mean for Arista?

Sure. I mean, I think we're clearly watching for that everybody's talking about it. I think in the context of the enterprise with respect to where Arista is, in networking market share is pretty distorted. There's been an incumbent that has an extremely large market share in the segments that we focus on, which is Fortune 2000 and up. And we come into the business really looking at a differentiated offering focused on reducing operational complexity and expense. So we have opportunities for share gain in that environment, I think somewhat independent of the overall market given where we are, especially in campus where we've been in the business for 3, 4 years, and they're still ramping.

And I want to get into maybe something that's a little bit more closer to home in terms of your day-to-day job is talking about a product cycle area. So we often get this question about the 400-gig cycle. So classically, it's where are we in that cycle? How long does it last? How is it different? And so why don't we start there and then we'll kind of unpack it a little bit?

Sure. I think 400 gig a day, really, the market is still pretty small, but dominated by cloud, cloud titan deployments. There's definitely enterprise and some media entertainment applications. But compared to the size of the cloud titans is still pretty small. We think a lot about the cycles. We've been in the networking side of the cloud since our inception and there's only been one period of maybe a pause or a hold and it's tough to say whether this is cyclical or and that was I think sort of people think about the end of the 100 gig, it was pretty clear 400-gig technology was coming. And then COVID hit at the same time so we had those 2 events. So it's difficult to find a pattern in that data, but I think we're still pretty early on, on this. And we talk about this cycle as 100, 200, 400 gig because the same chipsets and technology can be used across all 3 of those port configurations. So we, in fact, have products with the same silicon architecture that has 4x the 100-gig ports as the 400-gig ports and we have customers that are deploying in all 3 of those flavors this new technology.

Yes. I think that's an interesting point because that's one of the reasons I think the investment community is a little bit confused because you've been shipping 400-gig capable platforms for years.

But the 400 gig optics haven't really been readily available until maybe last year.

Yes. And we had a customer that did some pretty big deployments at 100 gig using that 400-gig cycle.

So some of this may be a little bit of a nomenclature challenge for the investment communities. There was a big break between 40 and 100 gig. So if you wanted to buy a 100 gig ready switch and deploy it at 40 gig, you waste at 60% of the bandwidth. So that really pushed the market to 100 gig to take full advantage of that engine, so to speak. Here you don't have that in elegance. You can use that same silicon for 4 100, you're not wasting any bandwidth and it's a smoother migration path with this technology.

Yes. No, I feel like that point hasn't kind of sunk in yet, so people struggle with where are we.

And share is tracked by port speed so that reinforces.

Exactly.

Perception.

So the other thing that I've observed in this industry over multiple product cycles is disruption invite share shifts. So when there's a new product cycle, somebody says, okay, we were doing 100 we're now doing 400. How are you seeing that change in that your networking competitors claim that 400 is sort of their insertion point and that they're catching up. What are you seeing in terms of the sort of shift in the competitive landscape as product cycles change?

Sure. I think that was a mantra a few years before a 100 gig existed. I think we feel very secure in our foothold in the early 400-gig market and leading that market as well as what we've done on 100 gig as a market leader. So I think we'll see what the 800-gig cycle brings, but I think we're feeling pretty good about where we are today.

And one of the things that I thought was an interesting topic at the analyst meeting was the transit WAN. So I think there's a lot of nuance to the different places people deploy networking gear. And so transit WAN is relatively new.

Correct.

So if let's start out by trying to explain it to a financial audience. How is this different from what you've done historically?

Let me put it maybe in the context of an enterprise, some generic enterprise somewhere in the planet. You have a core network that interconnects all your buildings facilities. You have a data center that connects the servers where you run your applications. You'd have some interconnect of that backbone network to your hybrid cloud. And then the last might be connection to some type of edge. If you were an insurance company might have a lot of branch offices. If you were a fast food restaurant, you might have a lot of other facilities that you're connected in. So think about an extension of your enterprise into these smaller branch offices. That network has had a lot of QT technology names over time. I think in the early days of networking, it was dominated by the concept of integrating services at the edge, specifically to run voice over IP. And then we went into a period where that T1 E1 links were the bottleneck and you had the era of WAN optimization so how can I optimize the links. And then as you started to get more redundant type of networks with LTE and IP backbones, cable modems, how can I optimize those links and get as maximum bandwidth and redundancy and you have the era of SD-WAN. So we have the building blocks to participate in that market. Some of those building blocks are access switches because you need them to connect when you're in the branch and then core routing technology to connect to the core. But what could be a differentiated solution and where could we have an impact. As you start to see hybrid cloud and the concept of transit through the cloud, that becomes an interesting assertion point for us. And we announced our intent to enter that market the key piece of that is it kind of completes the picture for us of a total enterprise networking solution from the routing core, the data center access to the servers, the campus and now connecting these edge locations.

And so it's not that this is a new category of technology it's a new use case for you. So who were the incumbents in that?

All the same incumbents we compete against today, the Cisco Junipers of the world as well as some specific SD-WAN kind of start-up smaller companies.

And you've talked about, I guess, the amount of TAM expansion. Do you have those numbers handy of how big that opportunity could be?

We talked about our overall TAM growing about $10 million in the next 2 years and another.

Billion.

$10 billion over the next 2 years. And another $10 billion after that with all of these elements altogether.

And the way I've thought about it, and I'm okay being wrong, so just say, hey, Simon, you didn't get that right. I look at this opportunity as really being focused on more of the point-to-point data center interconnect for hyperscale that this is not yet targeting sort of the typical telco metro routing applications where you're supporting consumer broadband and 5G and all that other kind of stuff. So is that in the road map? Do we think about you going from this sort of data center interconnect use case to more of a broad multiservice use case? Is that the intention?

This would be more a sell to the enterprise. So sell to the enterprise, VP of Networking that runs the entire architecture and solve the problem of how I connect my branches.

So I want to sort of pivot now to that enterprise campus part of that. So in '22, we expected revenues would double. I believe it was last quarter, you guys had to resin the forecast, blaming supply chain. You wouldn't have been the first, hopefully, you'll be the last, but I got that. What occurred really during the most recent quarter that led to the change in the forecast what kind of parts are you suffering shortages up?

The supply chain thing is more generic than just the campus piece, but given where we are in ramping that market, I think, had a bigger impact. If I look back over the last 3 years, we're out of the era of plant closures and mobility control orders that are keeping people away from work and really impacted by the semiconductor supply chain. I think around the large devices that are typically on very advanced process nodes, things have gotten more steady. I think supply is still tight, but more predictable but if you go to analog and power products that are used not just networking or even IT, they're used broadly in consumer applications, they're on 100-nanometer types of technologies and older. There's been a lack of investment in those kind of capacity across a broad set of suppliers, and they continue to be limited. And I think we foresee that happening into 2023.

What's your expectation for the duration of these supply chain constraints broadly because it sounds like most have said there's been some improvement, but we're not out of the woods. What's your prediction for when we're out of the woods.

All these suppliers are working on increasing supply and they all have plans for new fabs and increasing the size of wafers to get better utilization in existing fabs. And that's improving slowly. I think some of the demand reduction on the consumer side has actually been the primary reason that there's some benefit here. Just some context so I think when people think about improvement, they think we're getting back to normal. Normal days, worst-case lead time for semiconductors was roughly 24 weeks and today, things are ranging from 52 to 70 weeks. I think we have a long time before we see going back to that old normal. There might be some new normal and predictability would probably be the new normal as you might have extended lead times, but things show up when you expect them and there's no surprises.

But it does appear that in general, and I'm not now stepping back broadly, Arista, not camp that the company made a concerted effort to pay whatever brokerage fees you needed to pay to get products because you haven't missed on revenue your margins have come under pressure. So how much of that reflecting on it was intentional on your part to prioritize revenue over margin. Is that the way you think about it?

I think we think about it in terms of fulfillment of demand. So we saw demand. We made some purchase commitments that were significant to the supply chain in advance of this, which I think was helpful to get in line effectively. And then as you start to see missing shipments and you have the rest of the kit and you have capacity to build broker buys are a good way to augment that to fill in those gaps that's the way we think about it.

Okay. And I'll take you back to the campus topic that was just my own little deviation there. But when Arista first announced entry into campus, I definitely admittedly scratched my head and I get the logic of if you have 0 share, it's only one direction to go is up. And I get the idea that you've got existing customers who would be thrilled to buy other products from you because they like your data center products. So that, to me, was sort of the easy part that gets you established. What takes Arista to kind of the next level from a low single-digit market share player, you start facing tougher comparisons. So whether it's around feature differentiation or building channel, what's the strategy to take this up somewhere substantial?

Sure. No. I think maybe I'll step back for the investors, just to think about how we think about it. This enterprise network has been hypersegmented over time. As networking grew, there were a number of start-ups that focused on different technology areas. I'm focused on POE and access, I'm focused on WiFi. I'm focused on routing. And if you're a network operator, you're consuming all these technologies, you have a different operating system. Each of them might have at different times, their own security alerts. So now I have to upgrade and patch all of them separately. I have a different management network. I have to train my people across all of them. Arista's taking a different approach. We started with the data center. We added routing, but we did this in a manner with a consistent management stack and software stack. So one EOS release, that's our operating system, we'll run across all these various network pieces. So when we entered campus, I think people were looking for the knockout Wizbang feature some QLS magic or something, the magic was you could run it the exact same way as the data center, and you could upgrade it the same way, and you had a consistent management profile with an extremely strong high-quality operating system that's it. We're going to approach this win transit space with that same model. So the consistency of the environment is actually the big value added.

Now it's interesting because everybody targets the same one big competitor who sort of the argument is, well, the incumbent is complex, is hard to do business with. And the challenges I see it is the narrative from your closer competitors in market share, whether it's a Juniper or HP, Aruba, they seem to be telling a very similar story. So everybody can sort of keep nibbling away at Cisco forever and maybe that's a business strategy. I want to see how you think about competing against those smaller players that have a similar narrative, right, that it's simple, it's one screen, common management systems. How do you think about that?

Yes. I mean it is hard to compete against the narrative because it's very easy for me to say, I have better quality, right? So quality is an experience. Our sales teams look for an insertion point and what we've given them is more opportunities to insert. When we only had data center, it was difficult. If you missed that refresh window, you had to wait another 3, 4 years. Well, now you might have a campus opportunity or a transit WAN or a management network or some way to give the customer a way to experience the Arista effect, if you will. We found if we can insert which is probably the harder part, we're in a really good position for the next larger win. So you're absolutely right, competing with the narrative is difficult. You have to get a customer to have the experience and then we win.

And I tend to think of the Arista cultureas having a certain aspect of being frugal in terms of sort of you don't spend any money wildly. And often, you're underspending on OpEx and struggles hiring enough people and what I've sort of struggled with is when I look at companies that are selling into the enterprise and commercial market, they have to spend a lot of money on channel development, a lot of money on developing a sales force. And so far, you haven't sort of had to go down that path, but that's I see it as a risk. Internally, how do you talk about striking that balance of maintaining expense control, but being able to grow in those kind of markets and given your past history what it takes.

Yes. I mean we look in a new segment or let's say, we have a new use case to find that early success recipe with a handful of customers. And then we know we can scale it. And then we have the confidence to go spend accordingly and make the investments. If I see some of our sales teams and how they've expanded in the enterprise, they found some big accounts that they've worked on maybe in a region and then gotten a toehold there and then they bring in a partner, invest in the training of that partner specific to a region or an area. And then that partner is able to go find other accounts that are smaller and probably out of our scope of view, right? So that's been the effect. I think also, if I think about the conversation about competing against the narrative, you have to have somebody that experienced it. So having an anchor tenant, if you will, in the region, where there's some reputational success and now we've seen people leaving companies and moving in their careers to other companies and having the risk experience, it's getting easier to explain that because they can sell for us that it worked over here it will work over there. 5 years ago, we would go into an account like it's who's Arista, right? We have to go through explaining who we were, that's dissipating significantly.

Now do you see investment in the channel as an aspect of this strategy?

Yes, we're using the channel to fulfill and also to deploy. So yes, it's critically important, but we don't see ourselves as being kind of distribution-led finding partners and that model it's really direct driven in terms of the value and the partner coming and helping with deployment and bringing other technologies to pull it all together.

Now the other thing I want to ask about, I've got a question in the back. I'll repeat the question, but go ahead.

Yes. I don't mean to, John, drag you back into kind of the Mariola environment you're in, as you may guess from the question anything. Can you just tell us to bringing a lot of a question, what it is you guys have said about the macro you've obviously get your outlook for a lot of the sort of Tier 2 cloud people we have here. And while the medicine varies, many of them are talking about seeing some realize we send deal cycles and that's tended we're just being asked again for your view, inventory building. I think you addressed that just whatever the message is on the macro.

I'll paraphrase the question for the webcast. Basically…

[indiscernible]

Let's see if I do it shorter. Basically, what's the Arista messaging on how the macro is affecting the business broadly beyond my earlier question about potential for inventory builds, but how are you seeing the macro play out and change.

Yes. And let me connect that a little bit back to the distribution and resellers. We're very direct focused, and we're focusing on Fortune 2000. So we have, I think, good line of sight to actual deployments and schedules of deployments and especially around the situation where lead times are lonmg maybe even more visibility and some angst around those lead times. So that's helpful. Look, I mean, we're definitely watching like everyone else, the impact of macro. We're not very retailer consumer based in our deployments so that helps. We're not a mid-market company. So we might be shielded a little bit from even some of the visibility. It's not that it couldn't impact us if there was a deep and wide recession but as of now, we haven't been impacted.

So I want to make sure I ask you about software. And so I want to preface this with, please never call me a hardware analyst.

Okay.

Right. And because that's a bad word. And the funny thing is you're talking about the hardware guy so.

exactly.

Right. No, I get it. And so part of it is Arista's from its roots, you don't build your chips, your merchant strategy. So just to help folks out rough statistics, what portion of your engineering focuses on software versus hardware?

I don't have a Chris number, but it's probably 90% plus so it's all about the software. And from a quality perspective, my team make sure we don't mess that up.

So I knew the answer to that one. I just for the folks who sort of think that networking is boxes but you Arista has taken somewhat different tactic around sort of a software strategy than some of your peers, where they talk about recurring revenue and software licensing and you haven't gone down that path and maybe help us understand how you think about the software elements and monetizing the.

Sure. Yes. I mean we look at how customers want to consume and what makes sense to be in a subscription or ratable model and what needs to be consumed perpetual. It's very clear that the expectation of the market is when we sell a box that it comes with an operating system and that operating system has a perpetual license. We sell services that allow for upgrades and whatnot and service and support that's a component of that but that models widely accepted and I think that's what people want. Not really a model where they want to kind of look at over 3 years that I use a certain feature and then I upgrade my license for a new technology. We do sell like routing licenses for high-end routing features and applications and drive additional revenue on top of that. And that's well received. But we have WiFi licenses that are on a monthly basis because that runs in the cloud and it's connected with service. So where it makes sense, we're going to monetize where the consumption is appreciated by the customer, we'll do that. But if it's something around the box and what runs and what comes out of the box and we also want to simplify that licensing experience. So we've seen a lot of operational complexity of people not being able to run the networks because they don't know what features they use or what Box has what license. Easy to important it's easier to consume. Yes.

Great. So we've actually pretty much run out of time. So this was fun. It went by quickly. I'd like to close with sort of one sort of question that given the discussions you've had with investors recently, you have an opportunity to maybe try to expose something you think is either underappreciated or misunderstood that maybe we didn't get into, but sort of how would you like to sort of sum that up?

Yes. I think I'll just reinforce a point we talked about here on the enterprise side, where we talked about the different pieces and the WAN transit and the campus, I think it's important to think about how we position ourselves as an enterprise networking company and an alternate and really on a different competitive vector around operational ease of use and think about the enterprise business in that light.

Well, great. Well, folks, thanks for joining us. John, thanks for you time, thank you very much.

Pleasure.

This is Simon Leopold, wrapping up with Arista and John McCool. Thanks.

Thank you.