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

Good afternoon, everyone, and thank you for joining us this afternoon at the UBS Global TMT Conference. My name is David Vogt, the enterprise hardware and networking analyst at UBS. And we are pleased to have with us today, Jayshree Ullal, President and CEO of Arista. Jayshree, thank you for joining us today and hosting a keynote address at our conference. But before we get started, I wanted to turn it over to Charles Yager from Investor Relations to read the disclosure statements.

Yes. I just want to remind everybody that this webcast is covered by our safe harbor forward-looking statements. You can find them in our Investor Relations deck or in our quarterly SEC filing of the 10-Q statement. So please refer to those, and please continue. Thank you.

Great. Thanks, Charles. And thank you again, Jayshree. And just one housekeeping issue. If anyone on this line has any questions, please feel to e-mail them directly to [email protected], and we'll try to work them in towards the end of the call. And so again, welcome, Jayshree. Thanks for joining us.

Thank you, David. It's a pleasure to be here. I'm sorry we couldn't do a video. But hopefully, we'll make it up with good audio.

Yes, perfect. So without further ado, let's just jump right in here. I know it's late in the day for some people. On your last earnings call, in lieu of an Analyst Day, you highlighted 3 major contributors that you give -- that give you confidence in sort of a multiyear growth cycle. Maybe we could start with your core cloud and data center business and the 400-gig opportunity at a high level to get everybody sort of up to speed, that might not have been on the last call. So the Ethernet switching market is undergoing a transition, whether it's cloud service providers, telco service providers or large enterprises, investments are being made to upgrade infrastructure. Given your relative position and market strength, can you kind of touch on how you see the market evolving over the next couple of years as we move from 100-gig to 400-gig and beyond?

Absolutely, David. As I said in the last earnings call, I think Arista is really gaining strength from 3 major markets that we're now going to be playing in: The core data center, which is our largest market; our adjacent markets, which is routing and campus; and an emerging high-growth sector with network software and services, which is our recurring revenue with A-Care, CloudVision and CloudEOS. The recent acquisitions this year of Big Switch networks and Awake Security. But coming back to your question, the core is obviously very core to our business. It's very central. And I believe we are winning not just in 100 gig, but also in 400 gig. And I'm confident that Arista is going to emerge an early leader in 400-gig networking, just as we did in 100-gig, given the commonality of that database. We believe the 100-gig market is at least 10x larger than the 400-gig in 2021. But often, they're really in tandem and combination, and they're not separate decisions. So the TAM altogether can be significant at, I don't know, depending on market studies, in the range of $4 billion to $5 billion. You may also know that Arista has been an early product leader with over 20 products in our portfolio and approximately over 50 customers growing daily, weekly, monthly. So to answer your question more exclusively, the gating factors for the 400-gig delay of at least a year has been really due to 2 factors: the high exorbitant cost of 400-gig optics, which I think was both expensive and very limited availability, and both of those are changing; and obviously, the delays in qualification due to this 100-year pandemic we're all in with COVID-19. So as I've often said, I think the second half of 2021 can be a turning point for 400 gig, and we're going to win in high performance, whether it's 100 or 400.

Great. And then maybe just on the qualification comment. Obviously, a lot has transpired over the last 6, 7, 8 months. Maybe you can kind of give us some thoughts or some insight on what's changed recently to give you greater confidence in terms of the trajectory that you laid out today and on your last earnings call. Is it easier to get individuals to take the time to actually perform the required work to qualify products to get people in the door effectively? Or is it just the expectation that once we get sort of a better working environment with a vaccine potentially next year, those gating factors become a little bit easier to overcome?

Well, I think it's a little bit of all of the above, Dave. When you look at the shock we were all in, in our systems, when the door shut in March, I don't think any of us thought here in December, we'd be still dealing with this pandemic with almost everything shut down. So I think the qualification cycle has been limited to essential workers that can actually physically go to labs and test, much of what was not happening in the first half of this year. But life must resume and things must go on. So somewhere in the fall time frame when our confidence got stronger, some amount of limited testing has been happening. And of course, we have another shutdown now due to outbreaks for COVID. But I think we will continue to see slower testing, but we will continue to see some proof-of-concept testing in the first half of 2021 as well. And life must live on with the pandemic, which is what gives us confidence and improved 400-gig upgrades in the second half of the year.

Right. I mean, can you share any color or commentary if there's any difference between some of the major vertical markets that you compete in, whether it's the Cloud Titans market or enterprise? Or is it pretty much uniform in terms of the challenges and the opportunities that you face from this pandemic at this point?

Yes. I think broadening your question more to not just 400 gig, but cloud networking and what has Arista experienced in cloud networking. Again, I think much of the industry and customer base, including ourselves, was processing and learning what was going on in the first half. But -- and we were also having a lot of supply chain shortages and long lead times. So we were just in survival mode, to put it bluntly, right? As people started planning and realizing that this pandemic was going to be with them for a long time, not a short time, people had to move on and make decisions. So as you saw with our Q3 results, we experienced an overall strength in diversity across our high-tech enterprise, health care, M&E, Tier 2, Tier 3 service providers, Tier 2 cloud providers; and of course, the volatility of our Cloud Titans don't go unnoticed, but they're a very strong market for us and also came back strongly. So all of this would suggest that every vertical was affected by the pandemic, but every vertical is returning back strongly with a huge strength in diversification, both across our verticals and across our product line as well.

Great. And then -- and you mentioned sort of the constraints that you operated under over the last couple of months. And I think you touched on it on the last call, you saw improvements in the supply chain. But I think you did mention that you still were somewhat constrained, I think, if those are the words that you used, with some lead times entering into Q4. Any kind of commentary or thoughts on that today?

Yes. Nothing different than what I said in Q3. Our lead times are improving. It's a lot better now in November when I had the earnings call, than it was in May through August. But as you know, we've had a second rev of pandemic here, and there's been quite a few odd breaks in locations where we have contract manufacturing dependencies. So things can change in a dime. So while our lead times have improved, we're in no way out of the woods yet.

Right. Fair point. Good point. And then when you think about proof-of-concept and testing, obviously, your enterprise business has been exceptionally strong recently. And we were just talking offline about customer wins and some opportunities that you have within the large enterprise, not just in the 100-gig market, but in 400 gig. Has the COVID pandemic changed the behavior of the enterprise market to the point where the market opportunity or the addressable TAM is larger or more accessible today than it might have been 6, 7 months ago? Is there something unique about the digitization efforts for large enterprises that have been sort of pulled forward that are an opportunity that you can take advantage of?

I think that's a very thoughtful question. What the pandemic has done in an odd sort of way, it's given all our enterprise customers a chance to think, a chance to be more thoughtful and plan what their network looks like. Because otherwise, they're usually firefighting. So we've seen some major Tier 2 clouds just think about what they want to do and have a chance to build better networks, and build that leaf-spine architecture or extend the universal spine with additional monitoring capabilities and observability. Or a media and entertainment that relies so much on video broadcast now realizes that this is a way of life. So we won a large enterprise there on precision time protocol scale, EOS stability. There's also a lot of fatigue in our installed base of customers who pretty much only had one choice for the last multiple decades. Another example is many of our customers are looking at how successful they've been with us in the data center. And now, they're looking to move some of their workloads to their cloud and yet enhance their network monitoring capabilities. And through a combination of cloud EOS with AWS, and CloudVision, they're making the right sponsorships and decisions on what workloads go there. So -- and then even second-tier insurances, financials, banks, education, public sector, these are all examples where we're able to block and tackle from the traditional legacy infrastructure and really show that they have -- that you can see their age, you can see that quality, you can see the problems, and they want to move from the proprietary-ness of that to much more of a highly automated, agile, analytics-driven open network architecture with best-of-breed. So all of this suggests there's a number of examples in Q3 alone, that the COVID has given them time to be more thoughtful, plan better. And while we wish they were in their buildings more and testing more, that planning, I believe, we will see the fruits of that labor this year and next year.

Great. And then maybe just sticking to sort of the core cloud and data center market. I know we get a lot of questions from investors in terms of the pace and the timing. And obviously, there's a degree of uncertainty given the economic climate and the pandemic backdrop that we're all living through still to this day. But when you think about -- to your earlier comment about the second half of 2021, can you maybe expand upon that for a second? What gives you confidence, whether -- is it capacity constraints that you're seeing from some of your customers? Is it just demand trends that are just data-driven that require sort of the investment that you're articulating maybe if you can expand on that a little bit. So we have a better sense for what are some of the underlying drivers that you're seeing.

I think that's a really good point. That's a really good point. I think today, what we're seeing, especially in the data center, and obviously, to a different degree, on the campus, is that the network is a mission-critical bandwidth point. It used to be that when you did occasional applications, you didn't really consume the network. But the enterprises are driving video. They're driving collaboration. They're driving cloud-native applications. These's are not things that are separate anymore from mainstream enterprises. So while mainstream enterprises may have their proprietary applications in mainstream, they have to deal with a cloud networking strategy for cloud-native apps, where some of their workloads may move to the cloud, but many of them have to have cloud principles on their enterprise. And as you know, we went to some of the leading-edge customers and financials, and as well as the hyperscalers and Cloud Titans and deployed that. But it's not just for Cloud Titans anymore. Enterprises have their smaller version of Cloud Titan expectations, where their staff is not getting larger, but the pressure to build a mission-critical network is getting greater with the really chokepoint and predictable throughput and performance requirements of compute, storage as well as the pressure of applications. So this has been around for a while, but I think the culmination of everything coming together has made it even more important to build a sound network. And the network is often not the biggest point of cost, but it can be the biggest point of choke and is the biggest point of availability requirements. So putting that right foundation is key for them to build that cloud-first strategy.

Right. And when you think about some of your larger Cloud Titan customers, obviously, there's an expectation that there's going to be a significant investment spend in 2021. And I think you said it repeatedly in the past, and you just sort of intimated it just now that the network spend is not obviously the lion's share of the spend. But against that backdrop, is that the right way to think about it sort of directionally that investment, whether it's from a large Cloud Titan or second-tier cloud provider, bodes well directionally for your demand. Maybe it's lagged by a quarter or 2, but that gives you some confidence in the second half of next year.

I think so. I think we've had some lumpiness starting in Q4 2019. And as you know, between pandemic and some of our volatility of the Cloud Titan spend as well as some of our deferred revenue comps, it's been a tough 4 quarters. However, I think any time there's toughness, the tough gets stronger, and I think that will lead to a better 2021.

Great. And then another question that we get pretty often from investors is, and I'm sure you get it, fairly consistently is the generic threat from white box. Obviously, they've taken share in the switching market, in large part, by players like Google and Amazon. But obviously, maybe just give us a sense from your perspective, where we are today, what do they bring to the table today that you can't bring to the table or maybe a competitor can't bring to the table? And how would you sort of characterize your competitive advantage going forward when you think about the transition and sort of the next evolution of data center spend. Is there something that is digitally native about -- I'm sorry, go ahead.

Yes, go ahead.

No, I was going to say, obviously, you guys were the leaders from a digital-native perspective.

Yes. Of course, I get asked this question many, many times. And I think white box will always be a natural tandem to Arista boxes, right? There really are 3 choices for our customers, cloud customers, enterprises, many of them use Arista EOS because it is the best CloudEOS in the market, bar none, for turnkey enterprises because nobody can build their own, right? Now there are other times we can use an Arista SONiC or Switch abstraction infrastructure interface with Arista Hardware, and we can enable things like cloud typing operating systems, whether it's Facebook FBOSS or Microsoft or the industry SONiC. So what -- we're unique in that, we don't look at that as a threat. And then the third choice is Arista customers can use containerized EOS options along with other cloud operators now. So in no scenario, are we ever saying that the white box is nonexistent or Arista's nonexistent. Very often, it's a combination of the 2 rather than an either/or. You have to remember, going back to history, that the Cloud Titan simply did not have choices in 2005, when most of the white box projects were started like Google or Amazon. There was no non-blocking, high-scale, highly programmable, high-reliable switches. It was all vendor-specific closed proprietary platforms. And in fact, Arista was inspired by this gap. And we've had a decade of commitment to open networking since our very inception, whether you look at open Linux or EOS SDK, eAPIs, containerized EOS, SONiC support FBOSS support, O&L support, and even our most recent announcement of the Arista SAI endorsed by Microsoft. So developing a programmable software on open platforms, whether it's on the data plane management plane or consult plane, very much defines Arista. And whether that happens on an Arista blue box or white box, this is our culture. So in my view, the marketing rhetoric on white box is far greater than the competitive reality, but it's important to know that when white box happens, Arista embraces it, and we coexist with it, and we co-develop with it. And I think this often gets less understood, if you will.

Right. So maybe let me ask it a little bit different way, and maybe we can go from there, too. So how much value do you think -- or how much value can we ascribe to the actual hardware piece of the business going forward. And what I mean by that is, obviously, if Facebook wants to use their own operating system, Microsoft, the same or the industry, how does that change your dynamic going forward from a business model perspective? Obviously, a ton of inputs, a ton of variables. But just maybe help us think from a top 50,000-foot level, how you think about it going into a discussion potentially for an expansion of a relationship with a Cloud Titan player?

Right. I think it's important to understand software has to run on something. So there'll always be hardware, right? So I think the customers will choose Arista for its reliable software, whether -- in some cases, the software will run on a standard merchant silicon white box platform; and other times, they will not have the ability to do all that porting and would prefer that it'd be Arista hardware and Arista software. So our business model will always have a combination of software and hardware. And the proportion of hardware will greatly depend on the customers' ability to build that integration and invest in the engineering to do it themselves. So even the best of the Cloud Titans, can they do it? Yes, of course, they can. But will they do it? Well, they'll do some, and they won't do others. They will rely on Arista to do both the hardware and software. So I'm very proud of Arista's SDN and HCN philosophy, if you will, where we build both programmable hardware and the best hardware and software. And we need that combination to build the world's largest hyperscale, high-performance terabit switching platforms. And as for the mix, I think it will continue to be [ 6 to 1 ] and half a dozen of the other. I don't see a shift in that. On the enterprise side, I see it almost always be not a do-it-yourself, but give me your turnkey solutions. So the hardware and the software will go together. So while 80%, maybe even 90% of our investment is in software, I would say the hardware comes along with it because you have to run it on something. It doesn't go away.

Right. No, that's a good point.

A very critical part of the consumption model that gets lost, when you talk about SaaS and SASE and all that good stuff, it's great, you need that. But it still needs to run on hardware.

Point taken. And then I don't know how much you can kind of comment on this. We get this question a lot, and I'm sure you and your management team got this today at our conference. Any thoughts on how sort of Facebook and/or Microsoft play out over the next couple of years? Obviously, is it simply they're deploying effectively kind of a slightly different variety each of 400 gig. So in the case of Facebook, it's multiple channels of 200 gig and Microsoft probably, what, 400 gig straight out, out of the gate in 2021. Is that how you're thinking about '21 into '22 at this point?

Yes. Stepping back a little, both Microsoft and Facebook will continue to be very important, relevant customers to us. Microsoft has always been a top 10% customer, and Facebook, suddenly our top 5%, sometimes, even top 10%, right? And our cycles to networking are very much tied to Facebook's server cycles and compute cycles. And as you know, we did very well when they were deploying a lot of that, and then they decided to skip a service cycle. So you're absolutely right. You can only skip a service cycle so many times before you start needing the compute. And we believe that 2021, especially in the second half, will favor Arista networking in combination with compute and server cycle and CPU cycle decisions Facebook will make. Typically, like you said, lagging a quarter or 2 behind, which is why they make them in the first half. We could get some linkage in the second. Microsoft is a different set of use cases. We are -- we have a very tight partnership, executive to engineer. We are really working as an extension of Microsoft. We're co-developing with both Facebook and Microsoft with their engineers. But we are in many, many use cases. So we will work with them on pure zero SONiC use cases. We will work with them on spine and surplus spine, 400-gig use cases. And don't forget, in many of these cases, the switching and routing is coming together to form a universal spine. So the WAN routing, which is something that Arista has not historically been in, is also a very unique opportunity for us for that Cloud Titan and an important area for 100- and 400-gig use cases. So we see many roles, many networking use cases that can really bring to bear Arista's rich software capabilities and features, both in switching and routing.

Great. So that's incredibly helpful. So just in the interest of time, maybe we can touch on sort of these technology adjacencies that you laid out last quarter. Obviously, campus being front and center. Obviously a much smaller component or a much smaller piece of your overall business. Maybe just to give investors -- to level set investors, where are you coming from? Why did you decide to enter this market given that this is an incredibly challenging market, historically speaking, given the market dominance of the incumbent? And when you look at your product set and your competitive offering, maybe just kind of outline how you think you're going to be successful beyond sort of this initial $100 million to $200 million of revenue that you're sort of on the trajectory to generate?

Yes. No, good question, Dave. Everybody starts at the bottom rung of the ladder. So this is our first $100 million, and we look forward to $200 million and many more thereafter. But I think it's important to understand that Arista is not taking an old school cable plant approach to campus. It's really about building, a cognitive smart building or cognitive homes, cognitive offices, cognitive workspaces, responding to our customer drive for quality that we brought into the data center that they also want to see in the campus, and also responding to the explosion of devices and IoT, but 1 user no more equals 1 device. All of us are running around with at least 5 devices, whether it's your laptop, you have iPhone, your watch, your jewelry, your badge readers, whatever they may be. So we don't believe the future of campus needs to have separate architectures for wired and wireless controllers. And -- where you're sneaking in and out of different methodologies. Our Universal Spine that we had launched in 2019 for VLAN, VXLAN, Layer 3, connecting to a unified cognitive edge, whether it's unified for wired or Wi-Fi, across all clients, is really the reason that Arista is entering the campus, because we are bringing a software-driven automation through the campus like we did to our data center. It's the right strategy for the next decade, and it's clearly playing to our strength for programmability where we bring that heritage of high availability, zero touch automation, segmentation; with the acquisition of Awake, threat detection and malware detection; with the acquisition of Big Switch, more and more contextual cognitive observability to the campus. So we're winning many customers in the enterprise, and we're especially proud of the [ SASE ] and the [ back to spine ] launch at our November call with 5x the performance of our competition. 3RU, better footprint, leading integrated embedded security, the most improved sale over time with hit list communication and deep visibility and telemetry. So our approach to this cognitive concurrent, always on for our clients so that you can operate client to cloud, is very different than certainly the way I did campus 20 years ago, which was much more of a wiring plan decision tied to boring clauses. Today, it's much more [ data ] driven.

So then along those lines, when you think about sort of the other technological sort of adjacencies that you're talking about, obviously, you did the Awake acquisition, you just mentioned security. Maybe kind of talk through how this all kind of fits together in terms of your broader strategy. Obviously, acquisitions have been a relatively smaller part of the strategy going forward. What are you -- looking out over the next couple of years, what do you see as sort of an incremental set of technological skill sets and capabilities that you need to really drive this campus business forward? Is it more M&A, or...

Yes. I mean, I think it's never easy to do acquisitions, and Arista has adopted a string of pearls methodology to our M&A where our core DNA, whether it's for campus or data center or routing, is really steeped in our engineering, our innovation, our quality. This is some of the best and brightest engineers and leaders I've ever worked with in my life, in my 3 or 4 decades. But we will augment that with specialized skills to enter new markets, right? So if you look at every one of our acquisitions, it wasn't let's go do something we're already doing better; it was let's augment the principles we stand for, which is common culture, common team, outstanding technology, and bring adjacent value to our customers. So I believe we've had a high rate of acquisition because we reject 90% of them. Because when we double-click and invest in an entrepreneurial team, I'm personally involved with it for the first couple of years. And I think it's very, very important to build both an organic and an inorganic R&D engine. It's the lifeline of our company, and culture is a key part of making it successful. So that's kind of my philosophy. Now specific to security, you may have heard me say this many, many times: but for us, it's as important to partner with our security vendors, whether it's Palo Alto Networks or Zscaler or Forescout or many others, as it is to build a secure network. And we bought Awake because we wanted to turn security from a noun to an adjective. 50% of the devices out there, you can get them malware. And 3 years ago, Gartner stated that prevention is futile, unless it's tied to detection and response capability. You can throw the game of guacamole where you throw firewalls at it, you throw cybersecurity at it. And so I guess you could say Arista was awakened by Awake, where threat actors continue to evolve their tradecraft, but have we, as an engineering team, evolved our ability for threat hunting, malware detection, distinguishing good intent from bad intent, and normalizing activities across devices, IoT and even OT proliferation. And this is really where we were really drawn to the Awake team, because it's not just security, it's the combination of network and security, where they can distinguish between a good intent and a malicious intent. And the situation only gets worse as you get proliferation of IoT and OT. So most people throw the word AI, but what we saw in Awake was they really built a foundation that is AI-based, network processing-based and focused on the sophisticated threat hunter to build that autonomous AI-driven human platform. But at the same time, they made a combination of AI and how human expertise work together with their special query language. So I have -- I'm expecting big things from Awake. Rahul, here's a shout out to you. And I think they'll be a critical part of our Zero Trust networking initiatives in the future.

Got it. So it sounds like, obviously, Tier 1's going to be a critical component of the business going forward, but you still work with your existing partners and other components, other security components, whether it's Palo Alto. And so it will be sort of an integrated solution on a go-forward basis. Because I guess, if you talk to the -- if you talk to the security vendors, obviously, they feel like they're moving deeper and deeper into the network effectively, I guess, maybe not to put words in their mouth, but obviously, just I'd love to get your thoughts on. Would love to get your thoughts on sort of their sort of encroachment on the network side of the business and how you think that plays out.

Yes. Yes. I would say good luck for Arista to encroach into security or for security vendors to encroach into networking. They're very different skill sets. However, I think it's important to think of security in sort of 3 angles. One is how do you embed that into the network? And this may be things like encryption support or wireless intrusion protection or segmentation, which is a very natural form of security for Arista because we've been segmenting networks all the time about what groups talk to who and what and why. It's all about ACLs and VLANs and VXLANs in many ways. So doing macro segmentation, working with the Palo Altos to enforce that on the firewall, or with VMware to enforce it on the host, this is our heritage. So we'll continue to build embedded security, if you will. That's network-based securities. Or building just cognitive secure networks. The second is, let me just discuss, working with the best-of-breed security players. We can't be everything to everyone. We will work with customer strategy. Many of them have a strategy to go with Palo Alto firewalls. We will absolutely work with that. Or Check Point or Fortinet or Zscaler cloud services. Our aim isn't to be one of those vendors. Our aim is to take our state-driven architecture and then connect it to what they're doing. And then the third apex of the triangle, if you will, is the autonomous threat detection, where the network and the security don't just come back -- come together in an embedded fashion, but you really need that AI and human expertise, [indiscernible] nucleus and that trade craft of bringing the best-of-breed network detection and response for malware as well as the network to work together. And it can also work very well together with our deep DANZ monitoring and Big Switch capabilities. So I see embedded security, and I see network detection and visibility as really Arista a whole lot. And we'll continue to work with best-of-breed vendors. So I think we're being more pragmatic and sensible about where we are strong and where our partners are strong, and we'll bring the best of both to our customers.

Great. Great. And obviously, just to be sensitive to your time and the audience's time, maybe just a quick pivot to another end market that might be right for innovation, next-generation routing. Maybe you can kind of give us a sense for where you are after several years of investment and how you're thinking about that market as we go into 2021?

No, I think that's a good one. And I want to echo your several years of investment. This one didn't happen quickly. Our service provider strategy has been shaping over the last 3 to 5 years. And unlike the cloud strategy that happened faster, service providers move these same cloud principles just like Cloud Titans do. But it takes longer because they can't start with a clean slate and they have the incumbency. There's also the other problem that the traditional definition of a switch and a router in the past were very different products. Due to the innovations in merchant silicon, we can bring the best of fitting and the best of routing, whether it's Layer 2 protocols or Layer 3 VPN protocols or peering or traffic engineering, the combination of merchant silicon and EOS really allows us to develop a telco-based cloud for routing and switching with density, scale, power, footprint, economics. And this is not only true of classical service providers but also of CDN. A good rule of thumb over there is the streaming can be 4K flows. We're all living in this world today. And 4K flows can be 10 gigabits. You have 1 million to 10 million users, and you can end up using 10 terabits of capacity. So for all these reasons, routing, itself, is going through a transformation. And the demand for content, the demand for more POPs, higher bandwidth types, last mile access, high-bandwidth encryption, VPN services Internet growth have been exploding. So Arista's investment in simplified routing, and I use the word routing as opposed to router, to streamline agile services on such platform through options to different use cases have really been our strategy. And now we're starting to see the fruit of that success. We see at least 4 use cases. We see telco cloud and data center interconnect would be an actual extension of where we already are. We see the new web and edge, and this could be with Cloud Titan; it could also be with other providers. Peering edges of data center to colo or colo to cloud. And this is a very natural place. And finally, we see the core backlog upgrading to spine. So we're trying to see a mix of customer wins here with DC, DCI peering with both Tier 1 customers as well as some very interesting Tier 2, Tier 3 and of course, Cloud Titans as well. And so we are encouraged by the beginnings of this. And it's taken us a few years, but I think Phase 2 will be much faster as these are now getting qualified, and we start to see more production deployments of them.

Great. And then just in the interest of time. I actually got this question sent to me by a couple of people. As it pertains to campus, I think it definitely makes sense to just maybe address it. So -- and just turning back to campus for a second, when you're thinking about when you're discussing with customers sort of the technology that you bring to the table, are customers -- let me rephrase this correctly, are customers looking for programmability? Or are they looking for sort of the traditional Cisco solution where, I guess, Cisco is basically holding their hand effectively during the entire process from RFP to actually deployment? And how do you guys think about that?

That's a very insightful question. I think when you look at Arista's 6,500 customers or more, we certainly hold their hand, but they're -- generally, the leading edge, early thinkers, technology-savvy. And while they like to us to hold their hands, for sure, what they want most from us is quality and agility and automation and analytics capabilities and a very, very best-of-breed product that works. So the Arista class of customers that's already deployed us in the data center has qualified EOS, understands the power of CloudVision and a network time machine that can do change control and all of these capabilities, compliance, et cetera. So they are naturally drawn to us. I'll give you a little story on this one. Two years ago, I asked our customers, our top, let's say, 50 to 100 customers, "Would you like us to enter the campus?" And they all said, "No, Jayshree. Don't do that. We've got a product there. We've got multiple products. We got Cisco. We got HP. We really love you in the data center. We wouldn't want you to be distracted." And so in the second -- the next time I asked them, which was last year in 2019, asked the same question. And this time, 50% of them, maybe even 70% raised their hand and said, we'd like a product. And I said, "What changed?" They go, "Well, we're not getting the software quality and the consistency." And frankly, they were -- there was a lot of fatigue on existing incumbency and products, many of which were forced to go into new products and new subscription models that they didn't like. And this year, just before COVID, we had a similar customer meet and they said, "Where's your product? You're late. We want them now." So our customer base, the Arista customer base, is sophisticated, savvy and ready-to-use Arista product, right? Now I think there's a second half to your question, which is, but not everything is there is the customers. There is the whole white space we have of non-Arista customers, where we will have to work with our channel partners. We will have to work on building and bringing to them a much more plug-and-play CloudVision with the right approach, right dashboard, right UI, right compliance and right designs, and can hold them. So obviously, Arista is investing significantly in customer engineering, both pre and post-sales and professional services. And while our partner program is in its early stages, it will be an important piece to augment this, along with our technology partners.

Great. That's incredibly helpful. I think we are about out of time. So I just -- I think we should probably end it there. Jayshree, thank you again for your time. We understand that you're very busy, given all that's going on from a COVID perspective and a market perspective. So again, thank you for your time, and thank you for everyone joining. And if anyone needs to reach out, you can reach out to me e-mail or call me directly, and I think we'll just end it there. And everyone, have a great evening.

Thank you. Everybody, be safe. Thank you for this.

Have a good evening.