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

Great. Thank you, everyone, for joining, and good afternoon and evening, and thanks for sticking around for the UBS Global TMT Conference. I know it's late in the day. My name is David Vogt. I'm the U.S. enterprise hardware and networking analyst here at UBS. And we're pleased to have with us Ita Brennan, Arista's Chief Financial Officer; and Liz Stine from Arista's Investor Relations department. Just before we get started, I'll make a brief comment around the disclosures. So during this call -- I'm going to do it from memory. During this conference call, Arista's management may make forward-looking statements related to the financial outlook of the company and the potential opportunities and COVID-related issues that the company is facing. These forward-looking statements apply as of today and should not be relied on as representing views in the future. So with that out of the way, why don't we get started? [ Ita is reading the disclosure in person in conference call ].

You did a very good job.

Great. All right. So maybe let's -- obviously, there's been a tremendous amount of interest in how strong your business has been operating over the last couple of quarters. You just hosted an analyst event fairly recently. But I think a lot of investors that I've talked to sort of haven't really triangulated on how the business is actually operating by segment. So maybe we could just start with your recent results and your sort of 2022 framework and help investors understand what are sort of the driving forces of your sort of 30% revenue growth outlook to start for next year?

Yes. Look, I think we'd like to have multiple different scenarios of how we can get to kind of these outlooks that we talk about. So this one is no different. I can kind of walk you through the building blocks and maybe which ones we think are more likely, right? But we certainly don't know exactly how it's going to play out. We've had come off of kind of a period where the enterprise business, which is now equal in size to the cloud business, that business has been growing very steadily. We've talked about the campus piece that's included in that and that we're going to double that again in 2022 from 200 to 400. We've also seen the data center piece of that enterprise business grow consistently. So that's kind of one of the building blocks that we believe based on kind of the pipeline that we have and what we can see with customers and with opportunities in that business, that business can continue to grow and contribute kind of healthily to the growth rate. I think the providers business, which we don't talk about a whole lot, but that business has also returned to growth kind of second half of last year and continues to kind of indicate that it will continue to contribute to growth. And I think probably the biggest difference is that you're now seeing strength of cloud and cloud demand as well, right? So we had come through this period kind of Q4 '19, first half of through kind of 2020, where you saw that spend be somewhat constrained. Their spend cycles for the first time at a market level were kind of flat to down in some quarters. You're now seeing that kind of return, and we've started to see that demand return, and we can kind of see that with the visibility we have into next year. So I think it's a combination of those 3 factors and exactly how relatively they'll play out, I think maybe cloud grows a little bit faster. I think that seems to be kind of where we are now as we look forward, but some combination of those get us to that kind of 30% outlook that we provided.

And then -- and we get asked this question all the time, and I'm sure you must have gotten it during your meetings today. So when you think about the enterprise growth that you saw over the last couple of years and the strength that you're going to see in 2022 and hopefully beyond, what has been the sort of the competitive response by maybe some of your peers to your strength in that particular market, where it's been sort of the bread and butter of maybe some of your larger competitors in the past?

Yes. I think what's been interesting is we've been growing into that kind of that position with enterprise. We've always had the theory that the cloud principles, the cloud networking principles that were so important to the cloud customers, automation, quality, openness of the software stack, et cetera, would eventually be important to the larger enterprise customers and even beyond that at some point, right? I think what we've seen over the last couple of years is that as all of these companies become more -- have a more of an online presence, have more -- they're more dependent on IT, that's exactly what's happening, right? So I think what's enabling us to kind of see this growth is really the rightness of the Arista solution, EOS, end-to-end CloudVision as kind of the management tool single pain end-to-end, but simplicity is really resonating with enterprise customers. We've expanded the sales team as well in that time period. And we now have kind of the end-to-end product offering, WiFi campus all the way through to routing. All of those things, I think, are causing us to have that better performance in the enterprise. They're somewhat hard to respond to in the sense that the product offerings that the competitors have are kind of -- the offerings that they have, and they're not architected for that same kind of single operating system end-to-end in the same way as Arista is, right? I mean, obviously, there's a large incumbent that has a very strong position. They have their strengths in the market with go-to-market channels, et cetera. That's something that we will have to focus on as we go forward. But I think architecturally, it's somewhat difficult to respond to the Arista solution. Liz, I don't know if you want to add anything to that or not?

Yes. I mean I think you think about why we're winning with EOS and with CloudVision, I mean that's -- it's an operating system that we spent back 2004 -- since 2004, creating and building. And it's really the quality that operating system offers customers. And quality is hard to replicate and having disjoint product sets with all different operating systems and management platforms isn't going to be the answer to something that's EOS on every box with one CloudVision management and orchestration and automation tool. So I think that the enterprise is really seeing -- they've been ready for this for a while from a quality perspective, and they're really seeing the value of it. Whether they started in the data center and expanding to the campus or they're having a new campus opportunity now and are operating as a seat at the table.

Got it. I want to come back to the campus in a second. But maybe just -- I don't think you've given specific numbers, but you just mentioned that cloud is relatively the same size as enterprise going forward. And cloud has been sort of an underperformer as of late, as you referenced. So when we think about -- I think you said this at the Analyst Day as well, cloud, we're seeing an incredibly strong cycle in 2022. So cloud is going to be, I guess, pulling more than their fair share of the weight in 2022. And you have relatively good visibility in 2022. But in a normal environment, I know this is not a normal environment, what kind of cyclicality do you expect in that particular cloud business, right? So you -- in the past, there's been a year or 2 or 18 months of incredibly strong demand and growth, offset by maybe some digestion in the period shortly thereafter. How have the cycles changed? And how are you guys thinking about it longer term, given that you felt confident enough to give sort of a longer-term forecast at your Investor Day?

Yes. I mean, look, I think pre kind of end of 2019, you really hadn't seen any cyclicality in cloud, right? And it's just been kind of this something to the right consistent kind of growth as they invested and built out geographically, built out different use cases, different capabilities, long-haul metro, different scenarios. And really that kind of end of '19, first half of '20 is the first time that you saw kind of a slowing in their spend, right? And we had to navigate through that. Obviously, now we're at the beginning of this new product cycle, they've come out of that period, probably more constrained from a capacity perspective. They have and they're ready to make significant investments, they saw strong demand in their businesses in the face of that. So we're kind of -- there's a lot of drivers now for why we're seeing this uptick with them. And I think that will -- that has a long road ahead of it in some ways, we're in the early innings of that, right? But we do think that eventually, over time, there's a little bit more cyclicality to that business now than where it had been, say, pre-2019. And we have to accept that there will be periods of time where they will spend more aggressively, like in product cycles, in technology cycles in periods where maybe their growth is less, right? And so we incorporated that into our kind of outlook when we talked about kind of the mid-teens CAGR to 2025, that assumes some kind of cyclicality in cloud, where we'll have some periods like now, where you're seeing good demand from them and other periods, maybe we're more reliant on the enterprise and other pieces of the business to drive the growth. I think over that period of time, it's still growing -- we've got a reasonable CAGR. And obviously, you're still driving a very good kind of operating margin, cash-generative business, but it may not all -- that piece of the business may not always be up into the right. It may cycle even though over time, it's growing when they're investing.

And so maybe just to help clarify. So when I think about public cloud spend, I know directionally, CapEx is not necessarily a good correlative indicator for you, but it's directionally relevant. And when you think about the next cycle, so we're entering potentially a relatively strong 400-gig cycle, maybe not as strong as 100 gig has been, but you're going to have a relatively strong 400-gig cycle. How would you define sort of the duration of the cycle versus what we've seen with the 100 gig versus potentially 800 gig down the road? So are they becoming more compressed right, where companies, hyperscalers and enterprises spend significantly for many, many years on 100 gig and then maybe the 400-gig cycle is a little bit more compressed than maybe prior cycles and then you're getting into that sort of cyclicality again from the 800-gig perspective?

Yes. I think there's a couple of different drivers at play here. Obviously, supply is going to be part of this, right? Because it's going to elongate the cycle effectively, just because you can't get to the deployments that you want all stacked on top of each other, right? So the supply is going to play a role in elongating the cycle. I think the -- we've talked about 400 gig kind of being deployed initially into DCI and some of these high capacity, high-speed use cases first and then extending into the data center, when you have the next rev of silicon and the next rev of products. That's probably going to elongate the cycle a little bit versus 100 gig, where 100 gig was kind of deployed broadly early, the economics were such that you saw that kind of happen earlier in the cycle. So it feels like -- and again, we'll see how it plays out, but it feels like this 400-gig cycle is maybe not quite as steep and dramatic as 100 gig was, but we'll be -- maybe a little bit shallower and will elongate longer, right? Now again, we'll see. Obviously, nobody has perfect visibility, but I think that's how it's setting up from a both supply and technology perspective.

Got it. And then since you mentioned supply, when you think about your near-term supply constraints, obviously, things move around a lot what. We were joking earlier in pre-call about where the headwinds might be and the bottlenecks might be. Anything on the margin has gotten better, worse? Obviously, it takes a complete solution to ship a million different components, to ship a particular product. So it could be something different today versus something different last week. But just maybe can you help us understand kind of where we are if anything really changed. I think publicly, you said, what, second half of 2022, maybe early 2023 where things get a little bit better?

Yes. I mean I think we've talked about kind of back end of 2022, early 2023. Again, there isn't perfect visibility to that, but we can certainly see that in between now and then, we don't see big drivers for things to change, right? Our strategy, we started out initially taking a hard look at key components and being aggressive around supply and those key components probably early on in this cycle, which has helped, and we continue to true up on that. Last quarter, we broadened that out to directly deal with other providers, other suppliers that normally would flow through the contract manufacturing supply chain, et cetera, and deal with those directly. And we're now managing kind of some supply commitments around those suppliers as well. We may take control of some of that inventory through the contract manufacturers here over time because it's having a commit is not necessarily as good as having the actual inventory. So in certain components, we may do that just to make sure you can complete the BoM. So yes, there's a lot of moving pieces I think the customers are being collaborative in terms of helping with supply and helping with prioritization, helping with accepting some of the price increases that we've had to make. And then the rest of it is really blocking and tackling on the supply side. So I think it's obviously an area of huge focus both for customers, suppliers and ourselves, and we'll just have to keep -- I think there will be surprises. We need to be able to identify those. They already respond to them, be flexible around that. I think that's kind of the game, and we'll probably be playing in that environment for a while.

Can I go back to your comment about purchase inventory and purchase commitments versus inventory right? So the model has been sort of rely on your contract manufacturers to take in the inventory at your sort of command or discussion, but those purchase orders are cancelable, right? Or not cancelable from your point of view. What I meant is there's not an agreed upon price effectively, right? So they can adjust price. Is that how the agreements are struck where like a year from now, if you have a commitment, one of your CMs can say this particular component is up 10%, 15%, 20% in price, and that's going to flow through back to you? Is that how the deals are typically structured?

I mean, if you think in a normal time, obviously, those lead times are much shorter. And you place orders for -- everything was happening within a quarter, and there wasn't a lot of price movement that happens in those time frames. I mean now we're in a world where you're putting out orders and getting commitments on lead time and in theory on price, right, for that longer period of time for good or for bad, right? Now the reality is we've seen some repricing of backlog orders by some suppliers. We've seen -- and I think, ultimately, when supply starts to get better, whenever that happens, you'll see some repricing of maybe backlog on orders, but we're certainly not anywhere near there yet, right? So in this world, you get your commitments, you try to get your commitment on price. We have seen some repricing of that, right, which you have to respond to, to get the supply. But I think continuing to execute those long repurchase commitments is the best way to manage this and get in the queue and be consistently in the queue over some extended period of time.

And then when you think about your -- follow-up on the inventory component side, a lot of companies have been sort of slowed by the ability to get just-in-time inventory. It's been a great supply chain for a decade. And this little bit of a wobble has caused, I think, a lot of consternation, a lot of companies that we talk to. So to your point about taking maybe some more in inventory, I know it's early days and this is an untested sort of strategy. But how would you feel -- how much would you feel comfortable having on your balance sheet? Is there turns metric that you might think to target? Is there an absolute level of cash flow that you need on your balance sheet that's committed to working capital? Is it -- just help us understand what this could look like over the longer term.

So I mean our strategy on the kind of what we call key components, right? So the higher-value components has been that we buffered those even in the past, right? So we would always -- you've seen in our financials, you'll see our working capital bucket. And in there were stuff like the ICs and other things that we would buffer because their value is so significant. The lead times were long. And if you didn't do that, then the contract manufacturers don't have the wherewithal to do that, right, to buffer those, right? So we -- we've always kind of acknowledged that and had that kind of strategy for those high-dollar items. Now obviously, those lead times are way extended, right? And so the purchase commitments associated with those have gone up, right? Hopefully, we don't build that inventory per se because we should be shipping it as we get it. If we can get all the other components to line up as well, right? So I think what we're engaged in now is looking at some of those other components and saying, okay, do we need to help the contract manufacturers, hold some inventory and do some more strategic kind of planning around some of those components so that you have a complete BoM when you have those key components. I think that's kind of the -- I would think over time as things get better, and they will get better. I mean, they will, over time, at some point, we'll start to see this loosen up, then lead times will come in on some of those commodity components, and that probably goes back to normal, and we'll continue kind of our buffering strategy for the key components.

The key components, got it. And then maybe just, obviously, on silicon. Obviously, you have a great relationship with a key provider of silicon. Where do you think they stand in terms of hitting their commitments for you in 2022, given the guidance that you laid out there? And is there some opportunity for some incremental capacity that could lead you to be able to ship more than you might originally have thought when you gave your initial guidance for 2022? Like is there a little bit of a buffer there or you're shipping to what you think you're going to get, and that's kind of where we are for 2022 at this point?

Yes. I mean, obviously, we tried to take a risk-adjusted view of that outlook, right, and think about not only those key components because ironically, because we were early and we were more strategic around how we dealt with those key components, that's maybe been a little bit easier, and we've been more surprised by some of the other stuff that's shown up, right? But we did try to take a risk-adjusted view of that 30% outlook. So I think there's room for maybe some upside or there's room also for some surprises, right? It's not that it's all going to go in one direction. That's what we're trying to balance, right? So I think we feel good about that kind of outlook. And obviously, we're striving to improve it. But again, there are surprises that keeps showing up too, right? I think -- yes, I think it's balanced. We try to make it be balanced. And then if we see improvements, obviously, we'll continue to talk about that. But it's not just a one-way street.

Got it. And then you mentioned sort of long lead times. I just want to maybe go back to some of the conversations that you have with your hyperscaler customers or the cloud titan customers. Obviously, they're all spending relatively aggressively in the near term and into next year. Has the tenor or the duration of the conversations change with them over sort of COVID and where we are today from a supply chain perspective? And you've talked a little bit about this, I think, in the past, but do you have a little bit more visibility into what their sort of -- their needs are to get the sense that, that's sort of a potentially permanent change in how the communication or the dialogue would exist between the 2 of you or you and your customers? Or it's more of a temporary sort of phenomenon that probably reverts back to the normal way of doing business once we get past the supply chain initiatives?

Yes. I mean I think, look, right now, supply is top of mind for everybody all the way through the stack, right? It's the thing that's occupying everybody's minds across all levels of the organizations across the entire chain, right? And that's making it front and center and a lot of energy is going into trying to figure out the best view and the longest view that you can have around kind of demand and supply. So I think that's super helpful right now. It's necessary. It's -- I think it heightens people's understanding of the complexity of the supply chain and all the various pieces that go into that. I'd love to think that it stays this way from a demand visibility, when lead times come back in. I don't think that's likely. I mean I hope we don't go back to that very short-term view that we had previously, maybe that understanding of how it all works. Now it kind of makes that a little bit longer. But I don't think -- and whatever the right number of years' time forward is that we'll be sitting here with like 60 -- 50, 60 weeks of visibility because once lead times come in, I think that's just automatically going to contract the business.

Right. Well, along those lines, too, from a pricing perspective, so you're talking about taking pricing up, you're not going to get the benefit, as I understand it, until probably the second half of next year as it flows through your model. Customers are very sophisticated, and we've talked about this. How -- what's the kind of customer reaction to pricing? Obviously, I would imagine they're being amenable today to your point about supply chain. But how do you think that plays out as we go through 2022? And hopefully, there's some alleviation of some of the supply chain headwinds and that maybe there's some further discussion about maybe walking back those prices potentially as the supply chain gets better? I mean is that a reasonable outcome? Or you think prices are a little bit more sticky than maybe we might think?

Yes. I think we've kind of implemented the price increases as a very kind of collaborative transparent view of what's driving the cost increases. And that's given us, I think, some pretty good adoption of those price increases from customers. But by the same token, I think once -- and this will take time, right, because you have all these purchase commitments and backlogs with suppliers, et cetera. But I think over time, once those prices, those costs do start to come back in line, there'll be an expectation that some of those incremental price increases would also kind of reduce in line with that. And I think that's probably reasonable from our perspective. But again, it's going to take time because it's the same way as it's taking us time to get to the inventory that has those price increases. It will take time to get through kind of all of the inventory that will be in the channel or in the pipe, if you like, when things start to improve as well.

No, that's helpful. And then maybe just touching on margins along those lines. So I know it's been incredibly topical from an investor perspective and not just yourself, but a lot of other companies are experiencing sort of headwinds that are driven by the supply chain. I think you have a unique circumstance where your gross margins are going to be sort of affected by mix. You've talked about it. Can you kind of walk us through within sort of the construct of that 63% to 65% gross margin range that you talked about pretty regularly? How you see 2022 playing out, given the risk -- given the mix as well as sort of supply chain headwinds that we're facing today?

Yes. I mean I think we've talked about kind of the progression, if you like, on the cost stuff. We'll have -- we'll probably be second half of the year before we start to get the benefits of some of the price increases and stuff that we've done because we didn't reprice our backlog. We will see the impact of -- if we do mix towards cloud, like we talked about, that will have a negative drag on gross margin. I think over time, that 63% to 65% is still a good way to think about the business model. But in a quarter where you stack all those things on top of each other, if you happen to have a quarter that's particularly heavy cloud mix plus the cost south and you haven't started to get some relief on the pricing, then you will test the bottom end of that range. I think we've been clear about that. And -- but again, over time, I think those 63% to 65% is still a decent model.

Right. And so I mean, is it a reasonable assumption to make -- obviously, in a given year, you're going to have different growth rates between cloud and enterprise and some of the other customers. And so the primary variability in gross margin over the long term is solely going to be mix for the most part, right? If we're in a normal stable environment. So in 1 year, if cloud is significantly larger part of the mix, gross margins maybe below that midpoint of the range, below 64%. And if enterprise has been through the roof, 30%, 40%, 50% growth, you're going to be at the top end of the range. Is that the best way for people to think about between now and, let's say, 2025 just in a given year? Without giving specific guidance...

If you could kind of remove all the noise from all the other things, which is tough to do right now, but I think if you look maybe backwards in time, that's certainly been the biggest driver is that when we've seen pressure on gross margin, it's been periods where cloud has just been a bigger part of the business. And then when we've seen ourselves at the 65% [ loss ] because the enterprise has been a bigger part of the business.

And is there any -- from a software mix perspective, how does that factor in over the next 3 to 5 years, right? From a software perspective, most companies are moving in that direction to add more recurring component to their product offering. I mean is that a material driver in any way, shape or form to mix, to gross margin mix going forward? Or is it still just completely the right way to think about it, embedded in all of the products that you sell, and so it's already sort of captured in that sort of 65% to 63% range that we're talking about?

Yes. I think if you step back from the accounting and just think about the product offerings and the products that are being sold, the stand-alone kind of software offerings and the incremental kind of software offerings tend to be directed towards enterprise, the enterprise part of the business. So that's part of kind of that gross margin being healthier, right? Whether it's ratable or upfront, I mean that's just going to depend on what that offering is and where the accounting comes out, right? But the value of the software, whether it's some of the visibility tools that we have, whether it's CloudVision, et cetera, that tends to play out more in that enterprise book and it's kind of one of the reasons why that gross margin is higher.

Got it. And then maybe -- I don't think we've ever discussed sort of the implications of margin from campus. I mean, obviously, it's early days, $200 million revenue, growing to $400 million of revenue. Are campus margins comparable to enterprise margins where the -- I think the revenue is reflected in the enterprise segment. Are they sort of core enterprise margins right now? Or is it subscale, and we still have an opportunity to take some margin up there as well?

Yes. The margins on certain -- the direct sale kind of campus sales have tended to have margin that's similar to kind of the data center sales in that part of the business, right? Now obviously, we'll sell campus products to other parts of the business as well over time. But in the enterprise, they tended to have similar margins. And over time, we'll start to ramp some channel, more channel activity around that the campus products. And there, you'll probably see some offset in gross margin line maybe and then the sales and marketing efforts will be kind of in the channel to that gross margin line. But today, it's very similar because it's mostly a direct sale.

Got it. And then maybe just along those lines on a campus perspective, obviously, I would imagine selling campus to existing data center customers has been sort of an easier sell. Have you disclosed or have you sort of given investors a sense for kind of the mix between campus-led sales versus traditional data center-led sales, where it's an existing customer versus a new logo or a new relationship?

Yes. I think we haven't done that in some time, right? But there has been, I think it's, for us, a surprising number of kind of campus-led sales kind of early on in this kind of evolution. But the bigger part is still selling campus to the data center, to the existing customers, right? So as we've got lots to do to penetrate the rest of that. And then as we're adding new data center customers, now the conversation is just broader and includes campus in kind of those initial discussions, right? So that's kind of another driver. And then ultimately, it's looking at kind of how you drive the channel to lead some of the sales efforts around campus, particularly in that kind of medium-sized customer space and maybe less other than the larger enterprise space.

So that's actually a good point. I get a lot of questions on -- I think you laid out at the Investor Day, campus is going to double this year. And then by 2025, we're going to get to about $750 million of campus revenue, which sort of implicitly is tremendously strong growth, but it does suggest a little bit of a deceleration in '23 and beyond. Is the growth in enterprise -- campus, excuse me, should it reflect sort of the broader growth of Arista's revenue growth profile as we get into the outer years? Meaning a lot of -- as the base gets larger, it's going to be more difficult to grow and so it just sort of reflect the overall customer relationship, spending trends of your existing relationships. Obviously, you're going to add new logos but it's not going to be materially -- material outperformer, right? So if I think about 2022, you're adding $200 million of incremental revenue based on your guidance, which is a pretty big step function. Obviously, that's going to become less going forward. So it's kind of -- my guess, my question is, is it all going to kind of get lost within the broader enterprise, and it will just sort of reflect the growth rate of the core business going forward in the outer year?

Yes. I think over time, certainly, that's what happens, right? I think the -- like a $400 million business is -- I mean that's a pretty big business if it was a stand-alone business. So I mean the growth will slow as that becomes bigger for sure. And I think we'll set new kind of milestones as we go here. We won't lose sight of kind of what's happening in that business. But I think the growth will slow from -- you're not going to grow 100% forever, right? I think that's kind of what the target is saying, it's going to -- so do we know exactly what the slope of that line is going to look like, we'll have to see. But again, it's indicating that we do believe, it's not going to grow at 100% forever.

Yes. But I mean 100% -- yes, I wouldn't expect you to grow to 100% back to back, but if I think just from $400 million to, let's say, $750 million, that's sort of a mid-20s sort of compounded annual growth rate over '23, '24, '25, when it sounds like you're just getting traction, right? You're just starting to invest in the channel. And it seems like a little bit of a deceleration, probably sooner than I would have thought over the next, call it, 2 to 3 years. I mean, is it just -- you just trying to be conservative in terms of penetration rates and where we are in the sort of supply chain constrained world? Like there's just not enough people, you just can't go out there and build the channel as effectively or you want to be a little bit slower and more prudent in terms of investing in this business right now to keep margins a little bit more robust than they might be if you had full speed ahead to drive that channel expansion.

No. I think, look, we want to drive this as fast as we think it can be driven, right, and be effective in terms of what we're doing. I think the exact slope of that curve is still TBD, but we do believe that there will be -- over time, there will be some deceleration. The channel and building the channel for this is going to take some time. You don't necessarily need that for the $400 million or even for maybe the next milestone. But over time, you will need this, right? And so we are kind of engaged in doing that now and making sure that, that's in place and can contribute when we need it to. But that's still work in progress, right? It has to be targeted. It's -- I think we are making some progress, but it's early, right? So that's definitely an area of focus for us is, in fact, how do you do that so that you can address some other opportunities while you're still kind of benefiting from that larger enterprise space, which is more obvious.

Got it. I've got some questions from the audience, but I've got a couple more that I want to get through first. Your balance sheet is in great shape. Cash flow is strong, '22 will be another strong year. You've talked about buying back stock to offset dilution. Given the supply chain issues, maybe this is more of a theoretical philosophical question, but when you think about how much cash you need to keep on your balance sheet to run this business, given what we've just all lived through, do you have a general sense for how you're thinking about what kind of excess cash that you need to keep on the balance sheet versus cash used for M&A and buybacks and other sort of initiatives over the intermediate term?

Yes. I mean I think the way we think about it is what do you need for the business, what do you need to be competitive, if you like, to some of the larger players and be in a position to respond to things that come along. I think the supply -- the whole supply situation is just a good example of that, right, where we have the means, we want to have the means to be able to kind of have an equal position and be aggressive in that market. And certainly, it helps to have kind of the type of balance sheet that we have. I mean, obviously, over time, we have started to return cash. We'll continue to return cash. On the M&A front, we are -- we look at M&A as a lever. It's kind of netted out to some of the kind of, I would say, smaller technology adjacent-type acquisitions that have kind of been extensions of that enterprise push. We'll continue to look at M&A as an option. The bar is high, just given kind of the technology and culture fit that you're trying to achieve. But we'll -- but it is a lever and we'll continue to look at it over time. So I think big competitors want to be competitive and have the means to be competitive. We will offset dilution on the buyback and M&A is kind of -- it's driven really by the product, the technology of the business more so than it is the financials, but we want to have the means to be able to do something there. They feel strongly about it.

Got it. And just to be clear, obviously, you have a high-class problem, your stocks are up a lot. It's harder to use your cash flow to offset dilution simply because your stock is higher. But I mean, -- do you think there's an opportunity to not just offset dilution, but just to modestly reduce sort of your diluted share count, using what, $3.5 billion of cash on the balance sheet. 2022 is going to be a really strong year. I'm just trying to get a sense for if there's another lever that you could pull to drive earnings growth a little bit faster in 2022 than people might be thinking.

I mean we have been somewhat opportunistic from time to time on top of that kind of dilution offset, right? So I think we'll see kind of how things play out here. But we have been -- we're not opposed to that when it makes sense.

Got it. So I want to take this question from the audience. It's a couple of questions. I'll just pull them together as one. And the question is around your cloud titan contracts and your customer relationships. So when you enter a given year, like 2022, obviously, we're in a completely different environment. We have good visibility. If we take -- the question, I guess, is around what I asked earlier. If you take out the near-term challenges from supply chain, typically how long before COVID did customers give you an indication for what their orders -- what their plans would look like? Was it 90 days with sort of a typical sort of relationship discussion where like they gave you this is what we're going to do over the next 3 to 6 months? Or is it varied by customer? Because people are trying to understand what it could look like as we get back to more normalized operating environment in '23 and '24.

Look, I think in the pre-COVID, pre all of the stuff environment, we had -- like on an order book basis, it was very short, right? It was kind of -- you were lucky to have like a 90-day window, right? It now seems ludicrous than seem long, right? Then we would buffer the components, the key components for those products. And then with other lead times, you could turn upside in -- within a quarter, certainly much more than a quarter, right? So that's how that business was a high turns business. The long lead times were really those key components. And when you buffer those, you could be very responsive. So when you're responsive, you gave them a lot more flexibility, right? And that was the model. They would -- we would work with them on projects and understand the project visibility beyond that maybe for another quarter, maybe 2 quarters. But really in terms of book of orders, it was shorter. They tried to do forecasting, but again, it just wasn't as critical, right, because you had this responsibility to respond to upside, right? Fast forward to today, and it's a complete opposite of that, right? It's very difficult to respond to upside. You need a solid forecast for 50-plus weeks, right? So the question is, does it go back to exactly that? Or does it go back to something that's maybe a little bit -- with a little bit more visibility than that. I mean I hope that we keep some of that, and that maybe there's a little bit more collaboration around the forecast and stuff now that there's more understanding of how complex the supply chain is. But I don't believe that it stays at 50 weeks once lead times come in, and it's just not going to be the focus.

Great. And then we get this question a lot and someone just e-mailed me, when you look at, let's say, third-party data that talks about the growth in the public cloud, I know you're not off public cloud, but just using that as a proxy. Do you -- does the management team look at sort of that forecast, whether it's a Gartner, an IDC or Dell'Oro, whatever the case may be in terms of what the spending environment could look like over the next 3 to 4 years in terms of what the hyperscalers should be spending money on, given the growth in the public cloud market? I mean does that get factored into your planning and your forecasting for the next couple of years? Or is it more of a bottoms-up analysis customer by customer by customer? And I guess is there any sort of top-down overview on top of it as well?

Yes. I mean, for sure, we work with those guys. We look at their analysis. We try to understand their analysis, right? The challenge is predicting kind of they have the same problem that we have when you try to predict out 2, 3 years, it's difficult, right? So yes, we absolutely look at that data and learn a lot from their perspectives, but it's not that they know certainty what's going to happen out there in that -- and those time frames either, right? So it's definitely part of the mix, right? You're always looking at it to say, okay, does it make sense? It's what we're coming up from a bottoms-up perspective makes sense. But it's very difficult even for them to be predictive of what's going to happen 2, 3, 4 years out.

No, that's helpful. And then maybe just one final question from the audience that was sent to me. So when you look at cloud versus specialty versus enterprise, are there any holes in your product portfolio in any sort of maybe the end customer market verticals that going back maybe to M&A, it's more of an M&A question, I guess, that jumps off the page where you're looking at it going, if only we had this, we could be more competitive in this particular vertical? And where are you in terms of your overall portfolio from a philosophical perspectively? We're just going to operate the business for the next 3 to 5 years with some small bolt-ons, and we feel like we're competitively well positioned, given all the work that we've done over the last 3 to 5 years.

Yes. No. I mean there's a constant, I would say, road map -- R&D road map of stuff in front of us, right? Whether it's just the mainstream kind of -- I think the routing is a good example, maybe just because it's got a lot of pieces to it, but like there's been a complete rebuild of routing protocols that's happened with the cloud customers over the last period of time, right? So there's a lot of work that's gone into doing that. There's some service provider specific routing features that are more kind of to address their specific needs that are different to that, that we've been working on. I think the visibility is a good example. The Big Switch acquisition where we were handing off kind of we had a certain amount of visibility in the Arista platform, but we're handing off kind of that more -- that higher level of visibility capabilities to competitors, acquiring Big Switch recovery we could have done that ourselves, but we knew the team, we knew the technology there. We thought that was a good -- that made a lot of sense to kind of acquire that and added to the platform. So I think that will continue, right? There's going to be -- I think about low latency with Metamako, there's been -- I think you'll continue to see cases where the teams will interact with other technology teams that they respect and admire and believe that they can solve something that we don't need to solve for ourselves, right? And so there's an openness to that, for sure. But again, it tends to be something that has to be adjacent to be technology aligned. That to date, at least has rendered it to be relatively small, right?

Great. No, that's helpful. I think maybe we'll just leave it there, but I'd like to give Liz just 30 seconds. Is there anything that we didn't cover that you guys want to impart your wisdom on the investment community that maybe we didn't touch on? Or do you think that maybe people haven't been focused on with regards to your sort of strategy and your outlook going forward?

Yes. No. I mean I think we've touched on the key pieces of. But I think there's a better appreciation now of kind of the role enterprise in that piece of the business is playing. So I think that's good because it's important, right? Everybody is obviously fascinated with cloud, and we get that. And of course, we wake up every day now and think about supply. So we're not surprised that kind of everybody else is doing the same thing. I think that's going to be with us for a while. We'll be happy for the day when there's no supply questions, but I think it's a ways away. Liz, I don't know if you want to add to that or not.

No. I think we covered all the main topics supply, cloud, enterprise. I think it is important that to kind of stress that the one differentiator people miss is just the impact of the architecture difference of our operating system and EOS and I think that, that gets so glossed over. Sometimes, if they don't understand, it would take a complete architectural shift for somebody to kind of copy or emulate what we've done with the OS, and it's really sticky in the enterprise. And what it's allowed us to do for those enterprise customers and the visibility and the automation that's given their miss is why we're seeing success there.

Great. No. And with that, I think maybe we'll end it there, Ita. Thank you, Liz. Thank you for your time. You've been incredibly generous. We appreciate it. And thank you everyone that joined, and we'll catch up soon. If anyone has any questions, please feel free to reach out, and have a great evening.

Okay. Thank you very much, David. Thank you, guys.

Thanks, guys.