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

Hi, good morning. I'm Samik Chatterjee, the networking and hardware analyst at JPMorgan. For the next session in our coverage universe, we have the pleasure of hosting Arista Networks. As most of you can see on your screen, we have Anshul Sadana, who's the COO of the company. Anshul, thanks for taking the time to participate in the conference. It's a pleasure to host you.

I wanted to just kick it off here, and I know the audience probably will have questions as we go along, but just getting to the biggest change that we've noted on the last earnings call in relation to your commentary, you sounded a lot more upbeat about cloud provider spending. That's driven a lot of questions, obviously. Just to then kind of dig into that a bit, how should we think about that improved outlook in terms of visibility into some of the upgrades, 200-gig, 400-gig that you talked about related to some of the more standard capacity expansion, more data centers coming online, which is more of the usual kind of CapEx trajectory. So if you can help us break that down. And again, thanks for taking the time to do this.

Sure. Thanks, Samik, for having me back here. Look, our cloud customers have always been great development partners with Arista. We've done a lot of work with them in the last 10 years, and we are now onto our next-gen products with them. And it's an important cycle for our customers and for us because this allows them to get to next-gen technology, not just 400-gig in terms of speeds, but it has 100, 200, 400, has routing features in there, it has other savings for them in terms of power efficiency that we're able to hit as well, and -- which is why this will be good for the customers and for us. You are correct that compared to the past, we are, I would say, more satisfied with the cloud customers and the recovery on that side at this time. That's simply just execution, right? We needed to make progress on the execution and get to this stage where you feel confident that, yes, the next-gen is here, the customer is trying to deploy and move ahead. We've completed lots of successful POCs, or proof of concepts, with these customers. Many of them are deploying our products in their pilot sites. That's an important step before they go live or go to high-volume production. And there are other dependencies that are out of our control, like including availability of ZR optics for data center interconnect, and that's coming around really well as well, and which is why I think lines up very well for second half sort of start of this next-gen.

Before I move further, and Anshul, thanks for that, that sets the backdrop quite well, investors do -- investors who logged in or dialed in do have the question, if you think of asking a question, [Operator Instructions], and I can ask it on your behalf as they come in. Anshul, just moving to the other topic that has been, and I know you've answered probably numerous questions on this, is market share through the upgrade cycle with cloud titan customers. That's been kind of the key discussion point here for the last 6 months or 12 months almost. Now you talked about the execution. It's giving you higher visibility. So what are you seeing in terms of visibility related to your market share with the cloud titan customers, as you call them, I think?

Sure. Well, there are 2 aspects here, but let me first clarify the first one. So a lot of investors are looking at these reports and looking at, hey, can I look at 400-gig share, can I look at 200-gig share? What should I be looking at to see how Arista's execution is? But for us, many of these next-gen architectures actually include 100-gig facing downwards towards the server and 200-gig or 400-gig facing up towards the rest of the network. So which is why it's important to look at the combination of all of these high speeds together, 100-gig, 200-gig, 400-gig. And I think we believe we'll do well in that group for this next-gen cycle. With these cloud titan customers and overall in the market, we believe we'll maintain a healthy share. So we feel our position is strong, and we expect healthy results as well in this space. There's been a lot of noise, I would say, from competition. But the reality is, this market has always been multivendor. And if there is some displacement of share between 1 or 2 of those vendors and one new one is now making a lot of noise, it's not as big a deal for us because we believe we'll just maintain the share we have overall with these customers. And the rest is all about our execution and the customers' willingness to spend.

Again, as I move forward, if you have any questions, anyone in the audience, please send them in. Anshul, so you mentioned new competitions always being there. Some are making more noise around this cycle. I think, largely, when we talk to investors, there is a broad expectation that Cisco will gain some share in the upcoming 200-gig, 400-gig cycle because of offering silicon diversity to cloud customers. So how important of a driver do you think that is? I know Broadcom is largely aligned with offering -- Arista is largely aligned with offering Broadcom chips. So do you need to have a similar strategy of where you offer silicon diversity as well, if this is going to be a material driver for Cisco to gain share?

Well, as you know, Samik, Cisco has been building ASICs for a very, very long time. So this is simply the next-gen ASIC for them. There is more marketing attached to it, but it's not a new thing for Cisco or the industry. We've done well with our merchant silicon partners, including Broadcom, and we continue to offer what we believe are the best solutions to our customers. And our customers heavily influenced our choices as well. So this is not simply us building a product and then finding out who buys it. The cycle starts much prior earlier than that, and our customer feedback has been they like the products we're building, they like chips we get from Broadcom. And as long as we continue to outperform our competition, I don't think we have anything to worry in this space. As I mentioned in the first comment as well, it's all about execution, less about the marketing that's going on around these stories or these solutions, and we feel good about our position right now.

Let me follow-up on one of the comments you made recently about one particular hyperscaler moving back from white box to branded equipment in -- potentially in the future. I know you've kind of, in terms of timing, sized that up as something that will take multiple years to come through if that does come through. But if you can share your insights on what are some of the challenges that these hyperscalers are still facing with using white box on a consistent basis? And what could be the drivers that essentially get them to evaluate branded equipment, again related to white box.

Absolutely. When companies that have a lot of resources and a lot of money that they can invest in these areas, they make a decision based on certain factors. So build versus buy is generally about costs, about flexibility, about customization. And then at least one scenario, what we are finding is, it's not just the actual cost, but it's also the opportunity cost of these engineers. And these companies have larger initiatives to take on. And if they can invest these resources in other areas like compute or storage or encryption or databases and so on, the cloud is a fairly complex group of technologies that they need to offer to the enterprise, they can actually go ahead and invest these resources in other places and have a better ROI or better outcome for their business versus just building a product that they otherwise could have bought from the industry. Now in some cases, customers have not been very objective about it because they [ just have religion ] that I have to build it myself because I know it will be better and they stop being objective. In one of these engagements, we are finding the customer to be very objective about this. They're very open to the analysis and the evaluations, but we just have to go through that journey with them. If you've been building on your own internally for a lot of years, a long time, then it's not easy to quickly transition because your internal stack, your operations team, your tooling is dependent on your internally built products. And if that changes dramatically, you have to retool the way you provision, the way you deploy, the way you make changes to the network, the way you look at security patches, things like that, all of that has to change. And the customer, in this particular case, is evaluating through all of that as well, what changes do they need on their side to make this happen? So I think, in summary, they are -- this customer is actually quite open to a build versus buy analysis, not for every layer of their network, but at least for certain layers, where there is enough complexity to justify buying from the outside. Proof of concepts will happen. There'll be a lot of technology debates on other changes they may need to the product to fit into that space, and then they still have to go through that decision process on their own. But what makes me feel good is that they're actually much more engaged than ever so in the past and are willing to have a true objective comparison, and then they'll make a fair decision as well. Now because they've been in this cycle for a long time, it may take them a couple of years to actually go through this journey to try and transition, make the decision, do pilot sites, do production, go high-volume and which is why in setting all of your expectations, I've been saying this in the past as well that these things take a while, but the fact that this is happening is a healthy change for Arista at least.

Got it. No, that's helpful. Just following up on that. You mentioned the complexity of the layer really driving where -- which layers they decide to kind of move back from white box to branded. So obviously, the implication of that is, you're thinking of like the DCI and the spine being the more immediate kind of opportunities or the bigger target opportunities relative to top of rack or leaf. I just want to confirm that when you start that -- see that change happening, it will be from the upward layers down more and move down into the data center further.

We haven't mentioned exact -- the layer of the network yet because the customer hasn't given us the permission to. But you are correct, these are higher layers of the network that have more complexity in them and merit solutions from the outside because the cost of building internally is actually quite high as well. And in addition, you have to keep in mind, these are not point decisions, you have to make the decision for the next decade because it's the decision for the product cycle and then the next 3 generations after that. And which is why these are taken very, very seriously, both by them and by us, but I actually feel good about our position. As I've mentioned before in previous discussions, it's not a done deal. It may not happen, but 5 years ago, we couldn't even have this discussion. Today, we're having very healthy discussions with this customer. We now just have to go through that journey and execute well.

In terms of just finish off the conversation here related to kind of cloud hypers -- cloud companies, not just the titans, but including the smaller cloud providers here just outside of the upgrades to 200-gig, 400-gig, how much of a runway does 100-gig have? Who are the primary customers that you can still aim to gain share with when it comes to 100 gig, particularly if you look at the cloud vertical in itself.

So let me get this right. When our results have some doubts, there are lots of questions about the cloud. And then when the results are good, there are lots of questions about the cloud. Just kidding. Let me talk about 100-gig. So 100-gig is actually still growing in 2021. So 100-gig is not done with its growth. There are projections from several market analysts on how by 2022 or 2023, 100-gig starts to taper off and other technologies like 400-gig start to take off. But the reality is, the larger customers, Tier 1 cloud, Tier 2 cloud, some of the larger service providers, some of the largest financials are still deploying 100-gig in high volume. And in fact, the connectivity to the server is largely 100-gig, whether it's 4x25 gig service based or 2x50, but it's largely 100-gig, and that will continue to grow for some time. The upper layers of the network, where you are combining data from a lot of racks, lots of servers, lots of story spots is where you need faster speeds. And that's where 200 or 400-gig, and in the future, even 800-gig will make a lot of sense, and that's really what the architectures show. So the way we think of our next-gen products, when we look at the 7800R3 series, when we look at some of our 7060X4 Series products and so on, all of these products are using the next-gen chipset, but they have ports that are 100-gig, 200-gig and 400 gig. So going forward, we'll actually just continue to analyze this combined in that group of high-speed together rather than any discrete ports. It's almost the same thing. It's really your choice as a customer to decide whether you want to use a 400-gig port as 1x400 or 2x200 or 4x100.

Before I move to the enterprise topic, let me again remind investors that if you want to send in a question, just use the Ask A Question button there. Anshul, actually, before I move to enterprise as a topic, let me ask you 2 more near-term questions, which I know have been top of mind here. Particularly, the first one being supply chain constraints, and obviously, you did elaborate on it on the earnings call as well. But since then, I think it's an increase in terms of the conversation just given Cisco reported more recently and talked about some of the challenges as well as some of the higher pricing that they're seeing from their own suppliers. So just want to get a quick check there as have things kind of worsened from the time we spoke on the earnings call? Have things gotten better? What are you really seeing the constraints on in terms of your product portfolio?

Sure. Well, the supply chain is highly constrained, no doubt about that. We've been talking about this now for at least 2 or 3 quarters, and there's been a ripple effect for lots of reasons, right? All of us can read the headline news in Wall Street Journal every day. Almost every day, there is a new article on how either a car company or a refrigerator company or a dishwasher company no longer has chips to make their products. And this is certainly having its ripple effect through our supply chain as well. We did increase our commitments and our buffer for components several quarters ago, and I've been doing that consistently to keep up with these lead times, but lead times still keep going up. The longest lead time for some of our components, the chips especially, is running close to 52 weeks, which is a year. Now the good news is that our customers also understand this and are willing to plan with us and partner with us and sort of plan for this kind of a supply chain in the near term. But things are still unpredictable. If you look at the second wave of COVID or the third wave going on through many of the countries in Asia, there are more lockdowns. And it has ripple down effects, and you'll see some again in 3 to 4 weeks, where new suppliers come back and say, sorry, I couldn't meet up to my production commitments because my factory was shut down or slowed down. And we should expect that ripple effect for the near term. Our expectation is that shortages will last through this year. They could easily last longer as well. It's going to be hard to -- for us to predict when we recover from this cycle, right? The demand supply mismatches have a way of overcorrecting on both sides, but we just have to now go through that cycle and make it to the other end.

I know you talked about -- on the earnings call, you didn't give an updated revenue guide just on account of the supply chain constraints limiting some of the visibility there. But is it primarily impact to revenue and visibility in terms of meeting the demand that you're seeing? Or is there a component of this, which is on the margin as well or headwinds to margins as well, where suppliers are raising prices on you, and you'll have to find a way to either pass it through or that impacts your margins?

It's still a bit too early for us to make a concrete comment on margin impact. If prices go up and stay there for a sustained period, then that certainly has an impact to margins and prices as well, but we're not yet there. We have to go through that cycle with our suppliers and see whether these are short-term impacts or long-term impacts and have the same discussion with our customers as well. But what's going to happen, Samik, is that a lot of times when you do your planning, you plan for, okay, this is my customer demand. This is the supply of products I have. This is what you commit to the customer. These are your delivery dates or your ship dates and you come up with a revenue plan. But in the current scenario, because of the impact of shortages in the semiconductor space as well as commodity items, there are suppliers who often come back and change their commit dates. They'll come back and say, sorry, I can't give you your commodity items, including PCBs and power supplies and fans and so on, it can take longer. So you have to assume there is some risk in these deliveries, and which is why I -- it's hard to come up with a very concrete number and you just have to go through that cycle and execute well. And you can have better visibility near term, it gets harder as you push this out, which is why -- I think, for 2 reasons. This is one constraint due to which we didn't give guidance for the full year. And in general, we don't want to do that every quarter, right? We guided last quarter for the full year, and then we'll just execute through the year and come back and do it one quarter at a time.

No, no, that's helpful. Anshul, I do want to ask you one more on kind of separate outside of the revenue drivers here, particularly given the amount of kind of pushback or concern we see on this front from investors, which is around the operating leverage. What we've seen is you've grown revenue roughly 20% year-on-year in 1Q and 2Q, the operating margin although has been very similar to last year. And I realize you're holding to your target model of 37% operating margin. But given that we've not seen much margin expansion, despite the strong revenue uptake, it obviously has raised some questions from investors about the operating leverage. So just wanted to see how you're thinking about operating leverage in the model as well as like when you think longer term, are operating expenses really going to grow in line with revenue because that would be very different from how generally companies operate.

Yes. I'm actually surprised with that question because some investors ask us the other question, which is why you're not investing more to grow your top line. Look, it depends on which markets we're playing in, and what -- sort of what market share to expect and what investments are needed to get to that result. And some of the new areas of investment for us incurring more enterprise growth, more campus growth and so on will continue to require more investments as well. So I would say, it's too early to try and push the operating margin higher with increased revenue. If we can invest more in R&D or go-to-market, then we will so that it can result -- see results in higher revenue, which obviously has better net margin and EPS for the company as well. But I wouldn't get attached to it that much on a quarterly basis because the R&D cycles come in cycles for us as well. When you look at 400-gig, right, there is a lot of heavy uplift required to get to that next-gen product. At any given time, we're building 48 to 50 new hardware products. We have over 1,000 software engineers in the company, building new features for our customers. And we have to continue doing that, right? The game here is innovation and building the best product for your customer, not just trying to focus on net operating margin. At the same time, we are not a company that likes to waste money. We're fairly frugal. We are fairly efficient. And if we see the opportunity, we'll invest. And if we don't, we're not going to waste our money. We'll save the money for a future opportunity, which is what you see us doing as well. But if anyone has more concerns on this and want us to increase our operating leverage and operating margin, my only comment would be, be patient because that cycle will come. As we grow our revenue, you reach a point where the leverage increases as well.

No, that's fair. So let's talk a bit about the enterprise opportunity here. And again, before I do, investors, if you have any questions, please send them through. So Anshul, just like as you expanded the enterprise portfolio, obviously, you started with the data center, now you have a campus portfolio. When we think longer term, it does raise the question, do you really aspire to be like a full-fledged solution provider like Cisco, for example, who you already compete with in a lot of fronts? Or is the thinking here that you pick and choose which parts of the network do you want to be in, where you want to be best-of-breed and largely remain kind of a solution provider with only certain pieces of the network that you can provide solutions to, but don't become a full-fledged solution provider. So I just want to get your thoughts around the longer-term kind of how you're thinking about positioning with enterprise customers?

Absolutely. When you look at enterprise networking and enterprise solutions we provide, we have to stay best-of-breed. This legacy approach of trying to provide the entire IT solution from one-stop shop no longer succeeds. Customers are changing their operating patterns. They're changing their architectures. Their needs are changing. If we're trying to go after the entire installed base the customer has, in the next-gen when they do the refresh, some of these technologies are moving to the cloud. So connecting to the cloud becomes far more important than trying to provide a drop-in replacement for the 20-year-old solution the customer has. In addition, there are certain technologies where there are far better alternatives already available out there. A customer that is using a legacy teleconferencing solution is moving to Zoom. It would not be wise I would say, in my opinion, for us to try and build something of our own today and try and compete in that space. But what we are doing is partnering with companies like Zoom, partnering with other companies that provide sort of access control information to these customers, providing integration with the ServiceNows and Splunks of the world so that the customers can visualize and analyze the data that they otherwise get in discrete parts of the network, in compute storage, firewalls and so on. Partnering with firewall companies to get them better security rather than providing just a one-stop shop for perimeter security. We've developed a zero trust framework for our enterprise customers that includes Arista solutions like Awake, but also allows the customers to integrate with other partners, whether it's CrowdStrike or a Palo Alto or a Fortinet or a Check Point and so on. So you have to evolve your ecosystem with times. And the customer base today is telling us that they actually do want to separate out the stack and get to best-of-breed. But their challenge is, they cannot integrate everything on their own. So we have to provide that integration or the API hooks and more automation to them to do that and take them through this next-gen architecture and the next-gen journey, which we call the Cognitive Campus. So in our way, this is coming along well, but it is an iterative cycle. You can't come back -- you can't think of bidding things that the customer or from our standpoint are going to fit 10 years later. You have to sort of, okay, next 3 years, what do we need? And the next 3 years, what do we need? And do it incrementally to increase your footprint as well as the value to the customer.

I still don't see any questions, and I'm surprised nobody wants to ask any more on the cloud market share. So I guess there is not as much of a debate anymore, which -- compared to a few months ago. But let's keep one more on the enterprise side, Anshul. I think you've expanded -- you rolled out the Campus portfolio there. You acquired Mojo. You acquired Awake. But in terms of the sales motion and where you're seeing really customers, like what gets you in with a new customer? Is that still largely the data center capabilities, and then it pulls through some of the rest of the portfolio? Or are you seeing new customers come in just on account of some of the Campus solutions and not necessarily just because you already have a position with them on the data center.

That's a good question. We actually see a very healthy mix of existing Arista customers who like Arista products and EOS and CloudVision in the data center, and they're now leveraging that for the enterprise Campus side as well. But we also have new customers. We have customers that have warehouses. We have customers that have health care institutes, and we have universities in addition to large enterprises deploying Arista Campus solutions. And they're quite happy with the execution. The products are rock solid. They work really well. They provide more visibility than any of the incumbent solutions they've had, and they're easier to operate. For the enterprise, and this is -- I don't have the reference right now, but a stat I've often heard is that for every dollar an enterprise customer spends on CapEx, they will spend $3 in OpEx through the life of that gear, right? So in the end, product decision is only 25% of the total spend. You have to go help them in operationalizing this technology as well, and that's where enterprises are very attractive to Arista. So the acquisitions are certainly working out in rounding out and making the solution broader for these customers. That makes us more strategic for them. And then they look at us as the best networking out there for their needs. And the enterprise, going through this pandemic and coming back, is changing as well. When you look at how people are thinking of employees coming back to the office and reshaping workplaces and workspaces, I think our architecture plays a significant role because it's a lot more scalable, a lot more changeable than the legacy stack. So as a result, if you need a high-density area where you have hot desks or people are coming in for a collaboration area, you have slightly different networking needs there versus the rigid cubical and office structure we used to have, and our cognitive solutions absolutely enable that. So we're seeing some of that benefit as well. Customers like that forward thinking and allowing them to change versus forcing them to buy legacy.

All right. So Anshul, I did get questions over e-mail to ask you. So I guess, somebody is saying the Ask A Question button might not be easily visible. So let me take -- there are like 5 questions in here, but let me take a couple of them because you've addressed some of them already. So one of the questions is, as 400-gig takes place, is white box gaining share in the 100-gig at the same time?

The answer is just no. I don't see any new decisions happening for 100-gig in the white box space at this time, especially not with the large cloud companies. I don't know if there is any small companies out there making those decisions, but in the larger cloud titans and so on, what you'll see is, there are -- as almost everyone knows, there are 2 large companies in the United States that build their own white boxes today, especially for top of rack type of use cases, and they have been transitioning to 400-gig as the -- at the switch. Now the switch might be used to connect into, as I mentioned, 2x200 or 4x100 downstream to the server, but from a market share standpoint, you'll see this as a 400-gig switch. And those are getting deployed. So you'll see that transition happening, but that's largely a transition for 2 companies that already built their own switches, going on to their next internally developed switch. So there is no real change in market share or market momentum as such. For the rest of the space, I actually don't see any big change decision going on, on our white box for 100-gig. I think 100-gig remains stable as where it is. And in fact, I think Arista remains in a very healthy position in the 100. And going forward, 200 and 400 as well.

Got it. Got it. The next question in that e-mail. So what are the unit economics of 400-gig switches versus 100-gig? Are 400-gig switches at a higher price? That's the first part. And I'll ask you the second part of that, which is more on margin, but let's go through the first part.

Sure. When you look at the transition that happened from 40-gig to 100-gig, a lot of people try to equate that transition and are anticipating the same thing will happen now with 400-gig. But there is one key reason why that is unlikely. And that is the answer to the question this particular individual is asking, which is 100-gig, when we went from 40-gig to 100, it costs -- 100-gig was the same cost as 40-gig, same power as 40-gig and fully backwards compatible to 40-gig. So there is no reason to buy 40-gig at that time. But 400-gig related to 100-gig costs around 2.5x more and is almost double the power. And you have to deploy different types of optics to be backwards compatible. So as a result of that, it's not as straightforward. So customers who want to go to 400-gig are largely the ones who absolutely need the 400-gig speeds and the technology underneath today. And if they don't, they will either fan-out in 100-gig mode or deploy discrete 100-gig switches. So a 2, 2.5x cost delta is significant. And as a result of that, I don't think the market will simply go there just yet.

Given the price economics that you outlined, I think the second part of the question is largely around that, that as you sell more 400-gig, is it accretive to gross margin?

So the market is very, very competitive. And as a result of that, I don't think there is any room to increase margins in such a competitive market. I think market margins largely remain stable. Some -- I will answer the other part that some people have talked about, which is, if the 400-gig switches cost more and the ASPs are higher, doesn't that mean the market is expanding quickly or the revenue is growing quickly, which you have to note that you can also deploy a 400-gig switch as 4x100, and then the cost of 100-gig goes down. So as a result of that, you have to balance the 2 out. And when you balance the 2 out, I don't think the market dynamics are that different than what many of the market analysts have been projecting, which is somewhere in the 9% to 12% CAGR range for the high-speed data center ethernet switching market.

Okay. Let me -- I think we have like a minute here. So let me finish up with a couple of competitive questions here, competitive landscape questions here. One, I know you've talked about the enterprise market. So going back to the enterprise market here, you talked about it, but I mean essentially the way the landscape looks like today is everyone is trying to invest in the enterprise landscape and take share from Cisco, which is the incumbent. And there, the narrative definitely has picked up with like Juniper becoming aggressive, HP becoming impressive. And so any thoughts around that? It does seem like a bit more -- a bit tougher and maybe it requires more investment. The second part of that question, how does CloudVision now compare to some solutions like Apstra, which were acquired by Juniper because there seems to be much more of a multi-vendor alignment in terms of the solution that they are trying to provide to enterprise customers.

Sure. The way to look at the overall competitive space is, customers actually have been anxious for other networking companies to be an alternate supplier to them, but they needed a better solution, not just an alternate solution because the alternate solutions provided in the past were largely cheaper, but were more fragile. And no one really wanted to get fired for making a decision if they had an outage, but the Arista offering is rock solid. And when you have support from the largest cloud customers, the largest banks in the world, the largest service providers, the largest content providers, and they're all coming back and saying, these are really good products, in fact, they're better than what they've used from the incumbents in the past, you're more likely to succeed. So I think building good technology is still extremely important. Some people assume that the enterprise customers are not so smart, that's just not true. The enterprise customers are extremely smart, but they may not have as many resources on the engineering side. So that's coming -- comes back to helping them operationally and helping them succeed. And I'll answer the CloudVision part, but this is also where we actually spend more time with our enterprise customers, helping them making successful with the deployments, with the integration, and they're extremely thankful for that. So the solutions are very sticky with them as well. Now a lot of ways we do this, you have to worry about provisioning and change control some of the other things I mentioned, you need a turnkey solution for the enterprise. They're not going to go into 20 different tools and 1,000 APIs to do this integration on their own. That's where CloudVision comes in. CloudVision integrates with every Arista device automatically. It just comes out of the box, it just works. And tasks that customers typically do with 20, 30, 50 engineers can be done by one individual, so a significant improvement for the enterprise as well. The alternates out there, you mentioned one, but every company is trying to build their own solution, just haven't been that good because you need a rock-solid foundation to collect network-wide data, keep it in a database like structure, process it without any sort of synchronization errors, without any bugs and being able to sort of quickly roll back or make these changes and integrate with the ecosystem. We haven't seen others do this. They claim they do this, but when our customers have tested their solutions, they barely work. They're very, very fragile. They work with one version of operating system, but you try to upgrade and the entire solution falls out. That's really not a way to build a rock solid foundation and automation infrastructure. The other comment I would make is, many of these alternate solutions are essentially trying to evolve from the legacy network management, which used to use SNMP to trying to be the next-gen network management, but that's not the same as cloud-native automation. But we've learned quite a bit from our cloud customers and applied that into CloudVision, how do you solve these problems in a very systematic manner so that you don't run into these issues again. As an example, let's say you run -- you get a security advisory from your favorite vendor, saying the following advisory was released last night. Your network is now vulnerable to hackers. Here is the security patch, please go ahead and deploy it. What do enterprises do? They open up a ticket, schedule tasks, they have to test the patch, and then they have to deploy in production and so on. And the average cap for many enterprises runs in months. We're talking 3 to 6 months by the time they can actually patch all of their assets. So now for 6 months, they actually don't even know what's exposed, what's not because you're not talking about 1 advisory, you're talking about 10, 20, 50, 100 that they have to track. In CloudVision, that's fully automated. Every night, CloudVision downloads the latest information from arista.com and tells the customer, here is the latest, these are all your assets that are still exposed, and by the way, click here and we will do everything for you automatically without causing an outage. That saves a ton to their security teams and their compliance teams, and these solutions are just not available from the alternatives. So don't get fooled by marketing, customers do test these products, and they're not comparable.

Anshul, that's all we have time for, but thanks for this time. This was very helpful, very insightful. So thank you, again, and thanks for participating in the conference.

Thank you, Samik.

Thank you.