Welcome to the Fourth Quarter 2022 Arista Networks Financial Results Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded and will be available for replay from the Investor Relations section at the Arista website following this call. Ms. Liz Stine, Arista’s Director of Investor Relations, you may begin..

Thank you, operator. Good afternoon, everyone and thank you for joining us. With me on today’s call are Jayshree Ullal, Arista Networks’ President and Chief Executive Officer and Ita Brennan, Arista’s Chief Financial Officer.

This afternoon, Arista Networks issued a press release announcing the results for its fiscal fourth quarter ending December 31, 2022. If you would like a copy of this release, you can access it online at our website.

During the course of this conference call, Arista Networks management will make forward-looking statements, including those relating to our financial outlook for the first quarter of the 2023 fiscal year, longer term financial outlook for 2023 and beyond, our total addressable market and strategy for addressing these market opportunities, supply chain constraints, component costs, manufacturing capacity, inventory purchases and inflationary pressures on our business, extended lead times, product innovation, and the benefits of acquisitions, which are subject to the risks and uncertainties that we discuss in detail in our documents filed with the SEC, specifically in our most recent Form 10-Q and Form 10-K and which could cause actual results to differ materially from those anticipated by these statements.

These forward-looking statements apply as of today and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. Also, please note that certain financial measures we use on this call are expressed on a non-GAAP basis and have been adjusted to exclude certain charges.

We have provided reconciliations of these non-GAAP financial measures to GAAP financial measures in our earnings press release. With that, I will turn the call over to Jayshree..

Thank you, Liz. And I am glad we avoided Valentine’s Day this time. Thank you, everyone, for joining us this afternoon on our fourth quarter 2022 earnings call. 2022 has certainly been a record year for Arista.

You might recall, in November 2021 Analyst Day, we had given you a guidance of 30% growth and instead have achieved well beyond that at 48% growth for the year, driving to an annual revenue of $4.38 billion with a non-GAAP earnings per share of $4.58, translating to an EPS growth of 58% for 2022. Indeed, a memorable year.

Let’s get back to some Q4 2022 specifics. We delivered $1.276 billion for the quarter, with a non-GAAP earnings per share of $1.41. Services and software support renewals contributed approximately 15.8% of the revenue. Our non-GAAP gross margin was 61%, influenced by our supply chain overhead and cloud titan concentration.

International contribution registered at 23.5% with the Americas at 76.5% in 2022. This was one of our strongest performing international quarters in recent history.

In terms of Q4 2022 verticals, cloud titans was our largest and first, followed by enterprise and then specialty cloud providers at third place, financials at fourth, and service providers at fifth place. In 2023, we will report the three segment sectors instead of the verticals.

Shifting to the segment sector revenue for 2022, cloud titans contributed significantly at approximately 46%, resulting in a triple-digit growth annually. Enterprise and financials together was strong at approximately 32%, while the providers were at approximately 22%.

Both Meta and Microsoft are now far greater than 10% customers at 25.5% and 16% contribution respectively. Clearly, we continue to enjoy a strong and strategic partnership with M&M. With that, I’d like to now invite Anshul Sadana, our Chief Operating Officer, to shed more light on our cloud titan performance..

Thank you, Jayshree. Our partnership with Microsoft and Meta grew even stronger last year. Both of these titans are in the midst of deploying our next-gen 100, 200 and 400-gig products at several [key tails] (ph) of their networks. The cloud is reshaping the Internet with their massive footprint, global backbone and edge partnerships.

We are proud to have our products designed into pretty much all of these use cases. In addition, our business with the other titans continued to grow as well. We had additional design wins in backbone WAN and edge [Indiscernible]. This past year, we ramped our 7800R3 series, high-density 400-gig, near-lossless spine.

We also introduced several new products based on Tomahawk 4 and our deep buffer Virtual Output Queue systems based on Jericho 2, 7280 and the 7800R3 modular systems.

While we will continue to add 100 and 400-gig products to our portfolio, we also launched our first 1 rack unit 25-terabit product, with 800-gig ports that can be broken out as 2 x 400-gig. These products have good use cases and high-speed applications, such as artificial intelligence.

EOS, our high-quality resilient network data lake-based operating system, has also matured and now supports cloud scale with multiple copies of the Internet routing table. We co-develop with our cloud customers who greatly appreciate Arista engineering expertise.

This past year, we furthered our partnership with Microsoft with SONiC support on many of our high-volume switches. Our work with them on automation and monitoring our skills is very well received for Azure and Bing deployments.

At Meta, we have our co-developed platforms, such as the Tomahawk 3 7368 and 7388, which helped them improve throughput and datacenter power efficiencies. [FPaaS](ph) and EOS are deployed with very high reliability in the cluster fabrics using these products.

Our deployments in their backbone and in generative AI and recommendation engines with the 7800 series are now smoothly deployed in production. We don’t control macro. We don’t control our customers’ CapEx plans. But when they do spend, we are there with them to make these next-generation cloud networks successful.

AI is a good example where we are continuing to grow into next-generation architectures with our cloud customers. The use cases we are involved in are generally core to their business and not an optional spend. Our cloud journey has come a long way over the last decade.

This is still a very exciting market segment given the pace of innovation and our partnerships here. Back to you, Jayshree..

one, our core cloud and data center products built upon a highly differentiated Arista EOS stack that is successfully deployed across 10, 25, 100, 200 and 400-gig speeds. This drove approximately 68% of our revenue with strong cloud and enterprise spending cycles.

We believe that we will continue to gain market share in the high-performance switching and have already grown from the teens to the 20s. In the 100 and 400-gig category, we have now earned the number one position according to industry analysts. We have also doubled our 400-gig customers from 300 in 2021 to over 600 in 2022.

Our second market is network adjacencies comprised of routing, replacing routers and our cognitive campus. We doubled our campus orders to exceed $400 million in 2022, but we did fall short of our revenue due to extreme supply chain shortages. We maintain our campus momentum and are aiming for $750 million in revenue by 2025.

Our investments in cognitive campus spines, clients, wired and wireless have generated significant customer interest and demand based on CloudVision and CloudVision CUE. Considering this is only our third full year of shipping versus incumbents who have been in the market for 15 to 30 years, we are very proud of our execution.

Our vision for a cognitive campus with network-as-a-service and edge-as-a-service based on NetDL is resonating extremely well and being embraced by our campus customers.

We have also successfully deployed in many routing edge and peering use-cases, such as securing data in transit with TunnelSec encryption, precision and performance for mobile networks, cloud exchanges and metro Ethernet.

Enterprise customers can now deploy EOS with a single EVPN protocol, whether it’s for data center, data center interconnect or WAN, delivering multiple profiles. Just in 2022 alone, we introduced 6 EOS software releases, 600 new features across 50 new platforms. Stay tuned for more in 2023 as we will be introducing new WAN transit functionality.

The campus and routing adjacencies together contribute approximately 14% of revenue. Our third category is network software and services based on subscription models, such as Arista A-Care, CloudVision, DMF Observability, advanced NDR with Ava sensors for securities.

Arista’s subscription-based network services and software contributed approximately 18% of our total product line. We are proud to note that CloudVision exceeded 2,000 cumulative customers, up from 1,500 the prior year and is really a compelling data-driven platform delivering network agility, continuous integration and operational excellence.

Arista’s non-cloud wins continue as well. While Arista’s 2022 headline has been the massive contribution from our cloud customers, we are pleased with the momentum of our enterprise and provider customers as well. Arista continues to diversify its business globally with multiple use cases.

Helping our prospects and customers realize these operational benefits with modern software and automation has been a recurring theme. And so let me shed a few examples that we have earned a seat at the table at. Our first example highlights the universal cloud network wins in the travel industry.

Like many conversations, the customers’ initial ask was to gain more visibility into their infrastructure. We presented our DMF, DANZ Monitoring Fabric solution, but it quickly transitioned to a general data center for all of Arista’s platform offerings. The customer chose our Layer 3 leaf/spine EVPN design for their critical VDI environment.

The customer also leveraged CloudVision for their Day 0, Day 1, Day 2 operations using our chassis spine, R3 leaf and out-of-band management to reduce their operational risk. Our second win highlights the financial customers’ choice to proceed with Arista’s best-in-class cognitive campus with wired and wireless solutions.

As with every campus opportunity, it was competitive. CloudVision once again was a key differentiator for us as we quickly became their trusted adviser. Our virtual training environment, such as Arista’s Cloud Test, gave architects the relevant hands-on experience.

Our low [CVE] (ph) count and commitment to single EOS with high-quality was unmatched by our peers. Arista continues to make inroads on regional Tier 2 and Tier 3 service providers. Regional service providers are in the middle of expanding and looking for reliable compressed routing footprint.

This third win highlights the evolution of our EOS driving stack, where customers are now deploying EVPN services on top of their MPLS segment core network.

Arista’s high-density, 100-gig MPLS routing, together with long-range optics and a fully automated deployment using CloudVision and zero-touch provisioning, delivered that cloud-like operating model. Our next win is an international one in the education sector for high-performance computing.

HPC demands low latency, deep buffers and real-time visibility. This customer chose Arista for providing a highly elastic, VXLAN-based leaf/spine pod with best-in-class performance.

Consistent technology between our spine and edge leaf, anchored by our flagship 7800 chassis and combined with CloudVision-based real-time telemetry, compliance and automation, really created a lasting impression.

Our final win for this quarter’s announcements is an exciting international one in the government sector, where Arista’s 400-gig Ethernet was selected instead of InfiniBand for big data Hadoop cluster deployments. In this case, the customer chose us for 100, 400-gig solution with built-in encryption capabilities.

The customer saw clear differentiation in our automated operations, hitless upgrade and full real-time telemetry, ensuring comprehensive visibility of workloads in the fabric. As we enter 2023, Arista is well positioned as a game changer in data-driven client to cloud networking.

A key part of this transformation is to make our cloud first principle and bring that to every aspect of the data network. Software functions such as routing for WAN, Zero Trust security and observability are moving into the Arista U.S. stack.

We are building upon our cloud network heritage to bring proactive platforms, predictive operations and a complete prescriptive experience, unifying datasets from multiple sources. Our NetDL architecture and AVA, or autonomous virtual assist, using AI and ML and natural language processing techniques is a very compelling combination.

Together, this architecture can gather, store and process multiple modalities of network data. And this way, network operators can reconcile all their different silos. 2023 is the start of Arista’s 2.0 journey. Arista 2.0 is our migration from best-of-breed products to best-of-breed platforms as we address our expanded TAM of $50 billion ahead.

We are uniquely qualified to bring modern software principles to build that world-class data center and data-driven networking. It is based on that foundational focus on quality, availability, AI-driven deployments with top notch support.

And as we undertake this 2.0 journey, we are excited to work with a collaborative ecosystem of our partners, and customers worldwide to realize this vision. In summary, I am so proud of our team’s execution across multiple dimensions despite one of the worst supply chain backdrops ever witnessed.

A special thank you to our customers for their patience and support to us last year and to all the Aristans for their hard work and Herculean efforts. Our tireless mission taught us valuable lessons and we expect to emerge stronger.

We reiterate our 25% annual growth outlook that we mentioned at the November 2022 Analyst Day as we now aim for $5.47 billion in 2023 in terms of revenue. Now I will turn it over to Ita for financial specifics..

revenues of approximately $1.275 billion to $1.325 billion, gross margin of approximately 60%, operating margin at approximately 40%. Our effective tax rate is expected to be 21.5% with diluted shares on a post-split basis of approximately 316 million shares. I will now turn the call back to Liz.

Liz?.

Thank you, Ita. We will now move to the Q&A portion of the Arista earnings call. [Operator Instructions] Thank you for your understanding. Operator, take it away..

[Operator Instructions] Your first question comes from the line of Jason Ader with William Blair. Please go ahead. Your line is open..

Yes. Thank you. Good afternoon, everyone. I just wanted to ask, I guess, Ita, for you on the order trend. We all know that the revenue is incredibly strong right now because of all the lead time supply chain issues, but maybe some visibility on how orders are trending versus revenue..

Yes. Jason, as you know, we don’t really talk about orders and backlog. I think we did talk about kind of healthy demand across the various pieces of the business. And obviously, we’re reaffirming the guidance for 2023. So there is good support for that. Jayshree, I don’t know if you want to add anything to that..

No. I think you said it well. Order trends in 2022 were good. We will wait watch and see if the macro has broader effects in ‘23, but our guide and our tone effects that we are pretty positive at the moment..

So no impact from macro of significance thus far on orders?.

When we have something to state, we will, Jason. So far, we don’t..

Okay. Fair enough. Thank you..

Thanks, Jason..

Thank you..

Your next question comes from the line of Amit Daryanani with Evercore. Please go ahead. Your line is open..

Thanks for taking my question and congrats on the quarter. I guess, when I think about this 25% growth in calendar ‘23, how do you think it stacks up across the three verticals for you folks? That would be really helpful in the sense kind of where do you see the strongest versus weaker growth.

And then on the cloud titan side, as you think about growth in ‘23 and maybe even beyond, do you think that’s really a function of what their CapEx plans look like on the networking side or do you think there is a bigger narrative around the share gain potential against white box solutions, especially as workloads get more complicated, that could help you as well? Thank you..

Okay. Well, I’ll take the first one, and I’m sure Anshul will have a few words on the second. How does this break down? If you look at 20 – let me go back to 2021. We had a very nice even split, and cloud titan was actually kind of on the low side. It was 30%, if I remember right, 30-30-40.

And if you look at 2022, which we are called titan was outsized when the 30 went to 46. If I had to guess, I would say we’d be between those two numbers. I still think we will have a very healthy cloud titan mix.

But enterprise momentum continues to be strong, and you’ll see a contribution from that as well as the Tier 2 specialty cloud providers and service providers as well. So I think it will – my guess is it will look somewhere between ‘21 and ‘22 in terms of split. We will see as the quarter progressed.

In terms of the CapEx and the impact of that cloud titans, look, we don’t exactly and equivalently track to CapEx, but eventually, CapEx is an indicator of future – of our future cloud titan progress. I don’t believe at this point that our progress is coming from white box or specific things like commodity, things like that.

It’s really coming from, as Anshul pointed out, a very strategic seat at the table on new use cases like AI workloads, which has a multiplicative factor on our bandwidth. So I believe we will have a real seat at the table, especially with Microsoft and Meta. And we will continue to see what the use cases are that we can imagine beyond ‘23.

But we’ve been working on this for 10 years, and I think it will continue to be strong..

Thank you, Amit. We can take our next question, operator..

Our next question comes from the line of Paul Silverstein with Cowen. Please go ahead. Your line is open..

Thanks. I hope you’ll into a clarification.

I just want to make sure you said Microsoft was 16 and Meta was 25 or do I have that backwards?.

Yes. 25.5 on Meta and Microsoft, 16..

Okay.

Now for the question, what portion of your cloud titan revenue in general and how much of growth in Microsoft and Meta was – if you know it, what’s your sense for how much of that was AI-driven? Any visibility as to the growth in AI and its impact on demand for your switches and various use cases over the course of the next few years with your cloud titan customers in general, including Microsoft and Meta?.

Yes. We see AI as a very, very important use case and workload for all our cloud titan customers. Clearly, it’s in the first innings. We’re just beginning. So very much like cloud networking 10 years ago, we see AI as an additional use case. It is a very, very small portion of our use cases so far. So a lot of upside ahead..

Is it possible to quantify, Jayshree?.

Too early to quantify. It’s not material..

Okay, I appreciate..

Thanks, Paul..

Your next question will come from the line of Aaron Rakers with Wells Fargo. Please go ahead. Your line is open..

Yes. Thanks for taking the question and congrats on the quarter as well. I guess, maybe this is for Anshul, building on the last two questions. Is that – as you look at kind of adding up the Meta and Microsoft contribution and you compare that to 46% total cloud titans, your other cloud titan contribution is still pretty small.

So Anshul, when you’re engaging with other cloud opportunities, maybe you can unpack that a little bit.

What’s opening up the opportunities for you? Is it AI or is it something else that you’re starting to see? And how do we start to think about that as an incremental growth driver?.

Sure, Aaron. First of all, we are proud of our achievement for the first two M and M with the contributions there. On the other titans, we have been engaged fairly well with them. That business is also growing, but it pales in comparison to Microsoft and Meta, but it is not insignificant compared to other opportunities in the market.

And we continue to chase those. Those partnerships are very, very strong as well. At some point in the future, if the opportunities materialize, any of these customers decide to go big in the market and buy switches from the industry like us, I think we will perform very well. We start to wait out and get to that opportunity. It’s not clear at say yet.

Whether it’s happening in a year or 2 or 3, I don’t know. When it happens, we will be there. And we will do well in where we are today with them, which is essentially routing use cases or DCI use cases or WAN or edge. And we touched on this topic before, too. But if there were shift buying more from the outside, I think we will be ready..

Good. Thank you..

Your next question comes from the line of Jim Suva with Citigroup. Please go ahead. Your line is open..

Thank you. Jayshree, and Ita and everyone, congratulations on great results. My question is, I think it was Ita made the comment of expect a deceleration in revenues as we progress throughout the same – throughout the year just to get to the 25% revenue growth.

I want to make sure I heard that right because that would then also mean that even with very, very difficult year-over-year comps for revenues, you wouldn’t expect them to go negative at all. And I guess when we look at that deceleration, it kind of seems like a steep decline to get to an average of 25%.

So can you help me with my math there or the missing pieces? Or is it some conservatism? Or I’m just kind of wondering, but it definitely doesn’t seem like negative growth is in the works..

Yes. No. No, we didn’t talk about negative growth. If you look at the trend last year, you’ll see it really accelerated post Q1, right? So that’s why you’re seeing a much stronger growth rate year-over-year with our Q1 guide, then you will move through the year.

So I think after Q1, it’s better to start to look at it as a quarter-by-quarter – on a quarter-by-quarter basis and kind of earlier revenues quarter-by-quarter. There is certainly no kind of negative growth in that. I think you’ll get a better answer if you kind of just grow kind of quarter-over-quarter from there on out.

Q1 was a much lower revenue number last year back on the trend..

Great. Thank you for the details and congratulations and Happy Valentine’s to all of you..

Thank you, Jim..

Thank you, Jim.

This is all about comps, isn’t it?.

Your next question comes from the line of Samik Chatterjee with JPMorgan. Please go ahead. Your line is open..

Hi, thanks for taking my question. Congrats on the results as well. I guess, I had a quick one, which is....

Can you speak louder?.

Yes. Hopefully, you can hear me now.

Is this better? Can you hear me now?.

Yes, much better. Thank you..

Yes.

So I was just going to ask you on your large cloud customer, Meta, and their recent announcement around architecture changes related to data centers and trying to run AI workloads and non-AI workloads together on the same data centers and some of those related announcements if you’ve been able to dissect that and sort of have any thoughts about how that might impact their spending in relation to switching and routing equipment, particularly as it relates to your portfolio.

Thank you..

Yes. So Samik, I’ll say some few words and obviously, Anshul can get into detail. We don’t foresee any major architectural changes in the build-out of the AI clusters. Clearly, we continue to work with them on the front end of the network.

And on the back end, these have been based on the flagship 7800 spine, the AI spine, where you can have a distributed AID for it can be going straight into the spine. And when you have the hundreds and thousands of GPUs, you need a lossless fabric that has all of the congestion control and bandwidth management required.

So in the short-term, no major change in architecture. In the long-term, as these customers look for efficiency, we look for these AI fabrics to get larger or more distributed, but there will naturally be an evolution as the market grows, but no dramatic shift or change, just more of the same.

Anshul, say few words?.

Samik, just keep in mind, Meta slowed down spending a few years ago, right? So there is some catching up to do to sort of the spend that got missed out. So you have to sort of go back what’s an average it out to understand the trend. And second, Jayshree mentioned from what we know so far, we don’t believe there is any change in the networking spend.

The CapEx optimization they are discussing are either tied to how the buildings are built, facilities or letting go of nights to our projects..

Got it. Thank you. Thanks for taking my question..

Thank you, Samik..

Your next question comes from the line of Tal Liani with Bank of America. Please go ahead. Your line is open..

Hi. I want to ask about the other part that no one is asking about, the non-cloud titans. So if I back out cloud titans, non-cloud grew 14.6%.

And the question is, first of all, on last year, did you allocate components to cloud titans? And was this area more pressured than cloud titans when it comes to allocation? So if that’s the case – or what is the answer about what happens this year, this coming year or this year on the non-cloud titan portion? What drives it to accelerate from the 14.5% growth of last year? Thanks..

Got it, Tal. So first of all, absolutely not. We don’t do any allocation. It’s very much a first in, first out algorithm. And many of the cloud titans clearly were the first in, so therefore, they are the first out.

Our enterprise customers and the momentum as the demand is very high, and we fully expect that they will get their turn in this year, in 2023. But given how constrained we were in supply, this is the way it worked out in terms of revenue..

Is there – what are the underlying driver for growth acceleration, the driver – outside of components, better component supplies, what are the underlying growth drivers for 2023 versus 2022?.

I think they are very similar. You heard me talk about some of the enterprise momentum. Our customers are really looking for consolidation of their data centers in terms of a better automation, better telemetry, better consolidation of their operational advantages in the data center. Campus is a huge use case.

Routing and bringing all of the routing features that we’ve been working on for over 5 years to bear has been a third one. Observability and securities, another use case, our telemetry with CloudVision. So very similar themes to 2022 that we’re seeing in ‘23..

Great. Thank you..

Thanks, Tal..

Your next question comes from the line of Fahad Najam with Loop Capital. Please go ahead. Your line is open..

Thank you for taking my question. I had a couple of clarifications.

The cognitive adjacencies that were, I think, 14% of revenue is it fair to assume it’s fairly split evenly between CapEx switching and routing?.

Sorry, Fahad, can you repeat the question? I couldn’t hear..

The cognitive adjacency to revenue that you gave, I think it was 14% of revenue, if I’m not mistaken.

And I’m just wondering, is the split even between campus and routing?.

Approximately, both of them were large contributors. So I don’t have the exact percentages, but we think campus over time will become larger. But at the moment, I would say it’s 6 or 1.5 a dozen of the other..

Got it.

For my question, how should we be thinking about – with AI and machine learning becoming more pervasive and cloud titan architectures and this prospective displacement of InfiniBand with Ethernet, how should we be thinking about the TAM opportunity? Because how big does this InfiniBand replacement opportunity, so to speak?.

Yes. No, I think the InfiniBand TAM today has a very – we use HPC $1 billion to $1.5 billion TAM. And it didn’t address AI workloads. I think the advent of this new application is going to open up the whole AI, networking and fabric TAM to much greater than InfiniBand.

So not only do we have an opportunity to replace InfiniBand, but we have a greenfield opportunity for new AI fabrics and clusters. So it’s both, not just a legacy InfiniBand opportunity..

So roughly how big do you think the opportunity is?.

I don’t think – there have been some market studies on this. Some people say $2 billion a year, some people say $4 billion, some say it’s going to $8 billion. So I think it’s still too early to call. It depends on how quickly the adoption of AI fabric happens in all of our large customers..

Thank you. Appreciate the answers..

Thanks, Fahad..

Our next question comes from the line of Pierre Ferragu with New Street Research. Please go ahead. Your line is open..

Thank you. Good evening. I wanted to catch up on what you said, Jayshree, about like routing – edge routing and gearing. And this opportunity still comes back as an interesting and intriguing area.

And so my question would be, anything you can give us in terms of sizing, how significant it is today? And then beyond that, could you give us a sense of how you understand like the long-term market dynamics in there? So it’s a market where all the legacy routing players are very strong, have like a very strong existing ecosystem.

And I’m still not exactly clear on what market dynamics create the opportunity for Arista and how we should think about it in the long run.

Like is there an opportunity to replace incumbents in peering – in large peering markets? And if that’s the case, how does that work? Is that like operators buying from you? Is that coming from other types of clients, like cloud players? So what – how does the opportunity shape up over time?.

Yes. Anshul, I’d love your perspective on it. Let me kick it off. We think the router market is much bigger than the routing market. The router market is the more legacy market that’s being served by a number of traditional industry experts for 20 years and mostly servicing the service provider market.

And that’s a very traditional market that Arista has been participating some in, but we don’t expect to be a major player in traditional service providers. However, we’ve added so much routing features. Routing is now part of our switching system. It’s sometimes hard to separate it. It’s the same hardware, different software.

If you just look at the last year, we’ve added Ethernet OEM capability, VPLS, timing with SyncE, EVPN, MPLS gateway and multi-cap VPN, edge services, routing scale, you heard Anshul talk about, that can go over 4 million routes.

So our portfolio is really transitioning to supporting 400-gig deployments, and routing in the cloud scale is something we are very successful in. So on one hand, we’re not super successful in the traditional service providers. On the other hand, we are hugely successful in the cloud.

And then in between, we are finding ourselves moderately successful in a lot of the enterprise and specialty cloud providers.

Anshul, you want to add a few words?.

Sure. Period of another angle here, if you look at how we started to enter this market through some of the CDN companies like Netflix and Spotify, these companies have an SDN approach to edge. It’s a scale-out architecture.

You can take a simple router from Arista and scale it out, and the automation and the SDK we provide allows our customers to do that, which is why we do very well in these use cases versus the legacy full-feature traditional router. And our cloud customers, the titan, the Tier 2 cloud, the providers, all like these architectures..

Great. Thanks for your answers..

Thanks, Pierre..

Your next question comes from the line of Michael Genovese with Rosenblatt Securities. Please go ahead. Your line is open..

Great. Thanks so much.

I guess just sort of theoretically in an AI data center, I mean let’s just – current way of doing chat versus an AI chat, can you give us some sense of the switching intensity increase in the new use case with AI? Is there a multiplier to put on the switching or the networking to think about the higher amount of content and spend for AIs?.

Sure. Michael, I’ll take this one. It’s way hard to generalize. It must have a single number, but AI equals so much more. But I’ll give you an example of something that Andy talked about at the last Analyst Day.

And if you look at the recent pattern, which Meta published some papers about some of the time, the GPOs were sitting idle because they were waiting for the [indiscernible] to come back. So networking becomes the bottleneck and [indiscernible] you can add more bandwidth then you essentially become non-blocking.

You can do your job can run faster and you can use your GPUs in a much more efficient manner. So a rough order of magnitude with GPU clusters need about 3x more bandwidth than a traditional compute network today. But again, that’s a generalization, doesn’t apply to every use case. But if you need a single number, that’s the one I would use..

Thank you..

Your next question comes from the line of Meta Marshall with Morgan Stanley. Please go ahead. Your line is open..

Great. Thanks.

I just wanted to get a sense of – on supply chain, what you’re seeing there in terms of did it loosen faster than you were expecting in Q4 and that was part of the upside or just how you’re looking at conditions kind of improving throughout the year and maybe just that kind of release to gross margins as we think about throughout the year and kind of the overhead of the inventory currently.

Thanks..

Thanks, Meta. I’ll comment on it and Anshul, you don’t have a few words, too. Look, supply chain hasn’t eased up enough for us. Maybe we have more demand than others, and that’s why we’re feeling it more. But having said that, our Q4 numbers would have been even better if supply chain had eased.

And our Q1 gross margin is a reflection that supply chain is still an overhead on our cost, right? We expect Q1 to be the absolute worst. We’re going to improve thereafter every other quarter. So supply chain is going to be using in the back half of ‘23.

And as you know, at the Analyst Day, we gave a guide of – Ita, we said 61 to 63 for the year?.

Yes..

So we fully intend to improve our gross margins every quarter thereafter after, potentially hitting a low in Q1, which is an indication of supply chain improving. But at the same time, remember, another huge factor in our contribution to gross margins is the healthy cloud titan mix.

We’d like to keep it healthy and ease supply chain, and that will give us some improvements..

Great. Thanks..

Thanks, Meta..

Your next question comes from the line of Alex Henderson with Needham. Please go ahead. Your line is open..

Great. Thanks. And congrats on super quarter. I wanted to push a little bit more on the supply chain issue that – just talking about.

I get the point that the gross margins are the worst in the first quarter, but when do you think the balance between availability and your backlog starts to come into balance so that you can actually ship what orders come in and the duration on your backlog, which I know you don’t talk about, but conceptually starts to come in line so that we’re back to a fairly normal book and ship environment?.

Paul, I’ll let Ita answer this, but I wouldn’t call our current environment approaching normality for some time. So we hope it will be second half that the supply and the demand catch up. But I hope it catches up because we improve our supply, not that demand goes down. So we wanted to also improve for the right reasons..

Alex, I think the goal, obviously, is to improve – have supply improve and then improve manufacturing and improve efficiencies, and we will be working on that as we go through the year. I don’t know what the final normal will be. We will have to see.

I think just given everything that we’ve been through from a supply chain perspective, it’s probably – maybe there is a little bit more lead time visibility that will end up in the system at the end, but we will have to see..

I think what we can safely say is we are getting comfortable that lead times will improve throughout the year. Will we get to normal lead times? I think that will still take time because we’ve got to work through our demand..

If I could just one clarification. Did you say you had a decommitted in the fourth quarter? I thought I heard that in the presentation. Thanks..

No. Decommits on the supply side. I mean. We’ve had a some thousand starts on the supply side, for sure, if that’s the question..

[indiscernible]. It had to do with our supply constraints. Component vendors are constantly decommitting..

Okay, thank you..

Thank you, Alex..

Your next question comes from the line of Matt Niknam with Deutsche Bank. Please go ahead. Your line is open..

Hey, thanks for taking the question. I just want to follow-up on the question on macro that was asked earlier. Are there any regions, verticals where you’ve seen any maybe greater-than-usual slowness in ordering because of macro? And then maybe if I can sneak one in for Ita.

On the free cash flow trajectory, broadly speaking, just curious if there is any broad color you can provide around working capital and primarily asking around inventory and whether that’s still a drag or whether you expect to maybe convert some more of that to cash this year? Thanks..

Yes. I mean, I’ll take the cash piece of it first. Yes, I’m not sure that we start to see it kind of come down just yet. I think probably, at least for the first half, we will probably still be building inventory. I mean we do have some kind of key components that are still long lead time. And we wanted to build buffers, so we will continue to do that.

And then hopefully, in the second half, it’s probably at least kind of flattened out. But again, we will update that as we go quarter-by-quarter. But I think there is definitely a piece that’s still going to be a long lead time that will kind of hold inventory a little bit higher than what we might like for the time being..

And your question on macro, like I said before, we will call it when we see it. We are not seeing anything major and significant yet. And customers are watching, we are watching, and no major trend I can point to..

That’s great. Thank you..

Thanks Matt..

Your next question comes from the line of Tim Long with Barclays. Please go ahead. Your line is open..

Thank you. Just kind of a two-parter on the campus business. First, I think you guys have talked about doing a little bit better in the wireless LAN area. So, curious if you think that having a better wired and wireless portfolio kind of accelerates the share gains potential in that area.

So, was that something that was maybe holding back some wins that could help in the future? And then secondly, I think at the Analyst Day, you talked a little bit about SD-WAN. I am just curious if you can give us an update on when you might start to see another leg to the campus strategy in what’s a pretty high-growth vertical. Thank you..

Sure. So, Tim, on the wired and wireless, we are obviously much stronger on wired because there is a very natural affinity to the Arista EOS stack. So – and we also have a full portfolio, 1RU, 2RU, all the way to a chassis with built-in encryption. No other company, maybe except one, has that. So, we are very competitive there.

On the wireless, we are sort of the new kid on the block. And we have – as I said, if you just look at our campus entry, we are a new kid on the block. This is our third year. So, we I think are going from being a toddler to an adult now here very soon.

So, we believe we have a strong portfolio also differentiated by CloudVision, both wired/wireless coming into the same spine architecture that we articulate and designed for the data center. So, we feel very, very good about our portfolio being strong.

I think more of our efforts will go into go-to-market and reaching these customers because much of what we have done to-date is, if you will, low-hanging fruit with our familiar customers and our existing base..

Okay..

Your next question comes from the line of Ittai Kidron with Oppenheimer. Please go ahead. Your line is open..

Thanks and nice finish for the year, ladies. A couple of questions for me. First of all, for you, Ita, on the cash. I just want to piggyback on some of the previous question on the account receivables. Clearly, they have ballooned here on the year.

Are the cash payment terms of the cloud guys any different than a normal enterprise? And what percent of this account receivables do you think you can recoup in the year? And then for you, Jayshree, on campus, clearly, supply chain is a little bit of a hurdle there.

Cisco has taken action to redesign some of its solutions to products and components that are much more readily available.

Is that not a path for you? And if it is, what can you do on that front to alleviate the supply and more easily address demand?.

Maybe I will take the cash one first. I mean a lot of the DSO growth is really around those service renewals that we saw at the back end of the quarter. If you think about those and how they flowed, they generate almost no revenue. But obviously, they are in AR, they are multiyear, so it causes the AR to spike.

We will collect kind of a lot of that in Q1. Good healthy am or balance target, I think into Q1. There is no change in aging or anything else. It’s really just the timing of those service renewals and the fact that they end up in AR at the end of the quarter..

Yes. So, Ittai, thank you for wishes, by the way and Happy Valentine’s Day. We listened to you and made sure the earnings call was not on Valentine’s Day. To your question, absolutely, we have our choice of vendors and redesigns. Redesigns take time, and qualifying them with our customer takes even longer.

So, we have chosen multi-pronged approach, where we do have redesigns that we can invoke, but we are also improving our relationship and partnership with our supply chain vendors. Anshul, your team has been working on that. I think your vendor this has gone from tens to hundreds, if I remember right..

That’s right, Jayshree. This is the first time we are close to almost 100 suppliers where we talk to them directly. Even if we don’t buy the components from them, we control the relationship and the technology and the roadmap....

So, to answer your question in the campus specifically, both with redesigns and with our supplier partnerships, we fully expect to come back and not fall short of our numbers in ‘23..

Very good. Thanks. Good luck..

Thank you..

Our next question will come from the line of Ben Bollin with Cleveland Research. Please go ahead. Your line is open..

Thanks for taking the question. Good afternoon everyone. I also wanted to piggyback a little bit on campus. Jayshree, could you talk a little bit about how customers are responding as they are facing the increase in lead times or the lead times overall? It’s better market share opportunity.

Any risk that that shares perishable? Do they choose to opt to renew with who they have? And then you talked a little bit about go-to-market on campus. What are you doing differently, or what are your thoughts on where that goes from here? Thank you..

Yes. No, those are very good questions, Ben. I would say, currently, we are gaining share because others are messing up. Whether it’s changing to a software model or not able to supply, Arista has been the benefactor of that. They are still small numbers, obviously.

But it’s difficult to imagine that we are at risk of losing share when we have such small share. Our goal is to grow share at the moment.

What was your second question or the second part of that question?.

Go-to-market strategy..

Oh, what is the go-to-market. Well, in the near-term, our go-to-market has very much been to target our 9,000 cumulative customers. But we are building a mid-market strategy. We are going to work closely with channel partners. Those things take time. So, I would say our initial go-to-market is our enterprise customers.

And over time, we will have a more mid-market strategy..

Thanks..

Your next question comes from the line of James Fish with Piper Sandler. Please go ahead. Your line is open..

Happy Valentine’s Day, ladies. Great quarter. Just going back to your commentary on cloud titans being kind of between ‘21 and ‘22 levels just given the overall growth, it does suggest a bit of an acceleration for everybody else.

I guess what’s driving that confidence? Is it just mainly what’s in backlog? Is it additional hyperscaler wins, including with AI or enterprise share gains or something else? And then, Ita, just for you as a follow-up on the cash flow, is there a way to think about kind of a normalized cash flow level or where you expect inventory turns to get to by the end of the year? Thanks..

Yes. Maybe I will take that one first, Jim. I am not quite ready yet to call kind of a turns number for the end of the year. I think inventory dollars probably grow, certainly through the first half. And then hopefully, we can flatten out from there.

We will look for optimization, but there is still a fair amount of kind of long lead time items that we need to kind of carry and buffer. So, I will come back to you as we kind of go through the year. But I think certainly for the first half, you should be looking for inventory to probably continue to grow on an absolute dollar basis..

Thank you, James, for the wishes. I think in one word, I would say momentum. Our enterprise customers are really looking for an alternative to what they have got. There is a lot of fatigue in the system.

And what’s driving my optimism, where there is backlog from prior demand or present demand, is they are really hungry, and Arista presents that alternative..

Next question operator..

Your next question comes from the line of David Vogt with UBS. Please go ahead. Your line is open..

Great. Thank you everyone for taking my call. I just want to pivot back to Meta for a second.

And so in addition to the new architecture that they have been talking about, and I think Anshul addressed it, the company also talked about potentially using more colocation and maybe other public company assets to kind of meet its capital intensity needs going forward.

Would just love to kind of get your thoughts on how that impacts your spending going – they are spending on Arista gear going forward.

And then just going also on the titan mix percentages, if the rest of the business is growing at the rates that we think it’s going to grow in 2023 to end up somewhere between the ‘21 and ‘22 level, does that suggest that the titans business in total grows kind of in the low teens in ‘23 off of triple-digit growth in ‘22? Thanks..

And just to answer that one, it is definitely not going to be triple-digit in ‘23. We can say that with certainty. That was a beautiful year and one for the history books.

Anshul, you want to take the rest?.

Sure. On the Meta question, David, experience and so on, I think the high-level message to us is similar around their business efficiently as efficiently as possible and optimize. So, projects are nice to have. Obviously, those are getting cut back.

And as you mentioned, in base like colos and so on, you don’t need a very large architecture to start with. If you only have a 3-megawatt side, as an example, you have a smaller cluster size. But our products already fit very well in all of these use cases.

So, we don’t believe there is any significant impact to networking from what we can tell today in the near-term, right. We don’t have visibility, that’s many, many years out today. But the message we have been given is basically no big impact to networking as far as we are concerned..

Great..

Thank you, David..

Our next question comes from the line of Tom Blakey with KeyBanc Capital Markets. Please go ahead. Your line is open..

Yes. Thanks for squeezing me in here. I have a question back on the F&E line financials in enterprise, the drivers that I think maybe Ita was getting at many questions ago.

But I was wondering how much like rip-and-replace type of wins are kind of like starting to rear into here this – implied, in my mind anyway, an acceleration in the growth in the F&E line.

And specifically, the new cloud test product that you launched at the end of last year, if that’s kind of more of a 2023 driver and again, that kind of rip-and-replace type of wins, which is a large opportunity. And enterprise is more of a ‘23 driver or if it’s more ‘24. And then maybe just quick for Ita.

As enterprise mix is up, just remind us what the gross margin and operating margin impact should be for mixing more towards enterprise, that would be helpful. Thank you..

Maybe I will take that one quickly first. I mean I think the gross margins, we have kind of talked about it, improving as we go through the year, and kind of the mix is obviously part of that. Operating margin is pretty neutral actually between cloud versus the rest. So, I don’t know that there is any big driver there..

No. We have much lower sales and marketing on the cloud, more technically driven. So, it’s not the same. Going back to your rip-and-replace for financial, I think it F&E means financials and enterprise, just to clarify..

Yes. Exactly. I am just talking specifically about the new cloud test product where you can emulate an existing network and then just kind of plug and play the Arista product over an existing install..

Okay. So, one of the common spreads we are seeing in enterprise and financials is that they want – that nobody is getting more staff to do their job. So, they want more tools to automate and bring their SecOps, DevOps, NetOps, all of their operations together.

And this is where the Arista introduction of our continuous integration, continuous design and continuous test has really been strategic because not only do you have to give them a tool for automation, but you also have to work with them and train and teach them how to deploy it.

So, these end up not necessarily being rip-and-replace, but sort of a gradual evolution where they will identify the first use case of first data center that they will do this on, and then it will expand – land and expand to more use cases.

So, most enterprises are not a rip-and-replace, but it’s a use case that we begin with and then gradually evolve to go into a rip-and-replace as their depreciation gets completed on the existing legacy year. So, it’s a multiyear type of deployment, and it usually begins with a couple of use cases..

Thank you, Jayshree..

Thanks Tom..

Thank you, Tom..

Your next question comes from the line of Erik Suppiger with JMP Securities. Please go ahead. Your line is open..

Yes. Thanks for fitting me in and Happy Valentine’s. On the Meta front, I am just curious, they have talked about adopting more of a modular kind of scalable architecture. I am wondering if that changes any of the buying behavior on the purchasing patterns.

Does that smooth out some of the purchasing from the likes of a Meta? And then secondly, Ita, on the balance sheet with your purchase commitments, do you have control over how much inventory you take on, or as the inventory becomes available, do you get – do you take it in, in which case might we see your inventory balloon if more of the inventory becomes available?.

Yes. No, I think – I don’t like balloon as a word. I mean there are certain suppliers where lead times are [Technical Difficulty] inventory. So, we will continue to do that. I think on the purchase commitment, we talked about this a little bit at the Analyst Day as well.

I mean as lead times start to move around, obviously, we will work with the contract on those take [ph]. That’s why, I mean over time, that number should come down as aging to lead time with the contract manufacturers..

Okay. And on the Meta question, the Meta architecture already is quite modular with – you talked about design for development than terabits 7388. It can go up to 256 ECMP, 256 net bus. The cluster sizes are smaller, they don’t need 256. Maybe they can start with 16 or 32. So, we already built into the model today. I don’t believe it has any impact on us.

Same thing on the 7800 AI spine, they can add a number of line cards based on the number of GPUs or RACs that they are connected to. So, we are very, very efficient already and this fits very well in their model..

Thank you Erik. Operator, next question..

Your next question comes from the line of Sami Badri with Credit Suisse. Please go ahead. Your line is open..

Great. Just for me in two quick ones. First one is for Ita. Can we just talk about the benefits of pricing from some of the price increases that you guys have put through to the portfolio and the effect it had on gross margins? And then the second question is for Jayshree.

Jayshree, you have given us kind of a ballpark visibility, I guess some kind of quantification in the number of months that you see visibility with some of your biggest customers.

Could you give us an update on that same type of visibility?.

Yes. I think on the pricing piece of it, I mean for sure, we are getting some benefit from the pricing. But as time goes on, it starts in – the dynamic environment, it starts to be harder to track that kind of when it gets lost in the overall growth in the business. But we did check, and there is definitely some uptick for pricing there.

It’s just not something that we are kind of tracking on an ongoing basis..

And in terms of visibility, Sami, in the past, we have seen as much as a year’s visibility. If I were to guess, I think as the lead times improve, that visibility will reduce. Maybe it’s down to three quarters now. And the visibility was very much tied to planning cycles.

And when the planning cycles were longer than a year because our lead times were longer than a year, then that – then we got greater visibility..

Thank you, Sami..

Your next question comes from the line of George Notter with Jefferies. Please go ahead. Your line is open..

Hi there. I am curious about why you guys think you should take share from InfiniBand going forward in AI and HPC environments. I am just curious about what the logic is there. Thanks..

Yes. There is two big reasons. I think in the past, Ethernet was always striking in terms of performance and bandwidth to InfiniBand. Today, as we start talking about 400, 800, 1.2 terabits, the options on Ethernet are much greater and very cost-effective than anybody is there.

The other is, I think historically, InfiniBand has been more for high-performance compute use cases. We are very bullish on the AI workloads and its impact on Ethernet, where we don’t believe InfiniBand has any particular advantage and, in fact Ethernet does..

Thank you, George. We have time for one last question..

Your final question comes from the line of Simon Leopold with Raymond James. Please go ahead. Your line is open..

Thanks for taking. I wanted to maybe dig a little bit into the campus business, particularly whether or not that unit has been more constrained, and therefore, recovery bounces back. And ultimately, wondering if really an increase in campus in the mix, I know you gave us a $750 million target by ‘25 million.

Wondering if that’s considered a headwind to gross margin or whether it’s more about the market verticals that affects your margins? Thank you..

Yes. No headwind to gross margin. Our campus business has good gross margins. I just – as we said, on the product side, I feel very good that the campus can execute. On the go-to-market side, we have more work. So, I am giving our self some optionality that if we do the work really well, we could exceed the $750 million.

And if we can’t, then that will be the more likely number..

Great. Thanks Simon. This concludes the Arista Networks’ fourth quarter 2022 earnings call. We have posted a presentation which provides additional information on our results, which you can access on the Investors section of our website. Thank you for joining us today, and thank you for your interest in Arista..

Thank you for joining, ladies and gentlemen. This concludes today’s call. You may now disconnect..