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

Hello. It's my privilege to introduce the Vice President and General Manager of Cloud Networking Software for Arista, Doug Gourlay. Doug will give us a quick background of his responsibilities, and we'll jump right into some questions about Arista's software strategy.

Thanks, John. So basically, I look after building a software line of business at Arista. Product lines like our CloudEOS offering, what we're doing in monitoring, management, observability, looked after the Big Switch acquisition we executed over the last year and so on, areas like that. And I'm happy to chat with you about where we're going in software and how that really defines who we are as a business and really defines a lot of our success criteria in the market.

[Operator Instructions] So I guess the first question I have for Doug is I worked on the IPO for Arista and have covered it for many years before that. And clearly, the EOS, Extensible Operating System, was always the crown jewel on the differentiation, the competitive differentiator for the business. Maybe if you could speak to since the IPO in 2014, how has that evolved? How has the competitive advantage of Arista evolved? And is that still the key to your success?

Sure, John. Comically, I worked on the IPO of Arista too, just on the other side of the table. So I joined Arista in 2009, I took a small break, and I came back at the end of 2018, early 2019. And from the beginning, I mean, really going back to 2004 when Ken Duda was founding the company, the entire thesis was our industry could do better, right? I mean like we just felt that the networking industry has let down the networking operators, the consumers, the engineers, the people who use and depend on that network reliability. And it was just this foundational belief that our industry needed to do better. And that started with a better operating system. And if you look at the type of software models that Google's developed, using a large mono repo or there's tremendous code reuse within that environment, yet, it all composites into a very consistent end-user experience, we took principles like that. We took principles from how incredibly large-scale software development organizations have operated and brought that into the networking context. And some of them, it just sounds boring, but it's age-old principles like in enterprise resource planning systems and financial management systems, we use a database, right? We use a database to handle payroll, to handle supply chain management, to handle moving money, and that's because they work and they work incredibly well. So we applied really well-known principles. We applied the latest in cutting-edge distributed development technologies and said, "Let's build a different architecture for a network operating system, one that allows us to achieve incredibly high levels of reliability, but also incredibly high levels of code reuse, not wasting people's time, not boring our engineers to death, so that people like enjoy what they're doing in their careers and their jobs. Let's automate testing because nobody likes punching the button 100,000 times and see if it breaks the 100,001st. Better left to a robot, better left to an AI or automated system to do that." We built that infrastructure. And yes, you're right, that was the core principle of our announcement in our IPO in 2014. Now what's changed since then? We've added CloudVision as a network management platform that works very hand in hand with our core EOS product, but developed to the same standard. We've taken that CloudVision product line and expanded it to being delivered as a service. Because the ongoing day-to-day operational management of those types of resource computational storage, analytic intensive systems is complicated. And we can make that easier and more consumable by a larger number of customers in a SaaS delivery vehicle. And then as you've seen, really in the last 2 years, we've started expanding the aperture of capabilities we're delivering through strategic mergers and acquisitions, expanding us into the wireless markets, recently, the observability fabrics with the Big Switch acquisition and enhancing our portfolio in the low-latency trading world which has been a core market for Arista for over a decade now. But yes, what's changed? Some things. But so many things have stayed the same. And sticking to that strategy is really core to our success executing on that strategy for the long term.

Well, one of the things that's changed significantly on the customer side, obviously, is cloud computing, whether it's public or private cloud computing and how that's changed in enterprise IT usage. How has that presented opportunities or challenges for you and your competitors as, obviously, there's a one-way path to significant utilization of cloud as an infrastructure IT?

I think as anybody probably on this call knows, the cloud market as a customer base has been an exceptional set of customers for Arista, and we've done a lot to build and empower the architectures deployed in some of the major clouds. That being said, it has created, spot on, a fundamental shift in the consumption model of IT resources, and -- from the enterprise perspective. And we look at that enterprise customer and we said, "Wait, we need to take these architectures that have worked for the cloud providers because we're seeing cloud providers that are building networks, larger than the largest carriers in the world, about 1,000x more efficient from the number of operators necessary to manage them." And what is it that worked there? Well, there's consistent design patterns, consistent architectures that enables automation. How do we take that and bring that to the enterprise? So we started doing that with our universal cloud network architectures. We then realized that it's not just the physical how we interconnect the boxes. It's not just what routing protocol I use, which matters, matters greatly. But it's also how do I operate it? How do we give features and capabilities to that enterprise-class operator that lets them run their network like those large cloud operators do? And then fundamentally, how do we change and embrace the unit of consumption of the network in the cloud, which is a virtual private -- I buy a VPC in Amazon. I buy a VPC at Azure. I buy a -- or VNet in Azure. I buy a VPC in Google. How do I do that in my enterprise network? How do we enable consistent units of consumption so that the enterprise networking teams, buyers, right, the app team, the platform team, the systems teams, they can consume that network in a common model? One of the key capabilities we've brought out within the last year or 2, it's what we call CloudEOS. So we took that same core network operating system and enabled it to be purchased through the public cloud providers in a pay-as-you-go on-demand model to normalize that networking connectivity abstraction to get consistent cloud architecture, consistent cloud operating model and then enabling a consistent unit of consumption, a consistent cloud experience for connecting up these virtual private clouds, they being large multi-cloud networks, multi-region networks with a consistent operating model, so the customer can keep that same benefit of same EOS software, same troubleshooting model, same operating model, across their entire distributed environment, whether in the public cloud or on premises.

Well, that might have a lot to do with the next question, which is the obvious question. Why are you winning today...

Is that a softball tee up? I mean is that what we're doing here?

Without mentioning your largest competitor and all of their multitude of products, why are you guys winning today? And how has that momentum changed, positive or negative?

Sure, John. And I'll tell you that when I'm talking to customers about new product offers and areas, should we go in, and I'll give a classic example. 3 years, 4 years ago, Jayshree was asking our customers, should we go into the campus? And they're like, "No, just nail the data center, get the data center nailed, and we'll be great." 2 years later, they're looking at or going, "You're late. We need you in the campus now." And she's like, "You told me no." And what we've realized is the software quality, the quality of experience we've delivered for our customers has given us a charter to support the enterprise network, right? And it's -- our customers are asking us, "Please support as many locations in our network as possible." And when I go back to them and we have that conversation now and I ask them why -- and I'm the guy who likes to invent a new feature, and that's been the hallmark of my career for 20 years. I got an idea. I go, "What do you want us to do?" I got these ideas, and they look at me and go -- and I go, "Why would you buy this from us?" They go, "Well, because you're Arista. If we buy it from you, we know it's going to work." I'm like, "So it's not the cool idea I had?" No, it's the quality. And there's -- I think that's created this charter to expand the footprint because it makes it easier for our customers to operate. And probably the simplest example I can give you, John, every time there's a vulnerability in our software or our competitor's software, somebody files what's called a CVE, a common vulnerability or exposure with the U.S. federal government, the National Vulnerability Database maintained by NIST in the Department of Homeland Security. Last year, we had 2. Now it's 2 times we released a software defect that caused our customers to have to do a software upgrade. We're really sorry about that. Next year, we're going to do some more. It's going to happen every year. It happens to everybody. Our job is to own those, support them, give our customers choices on patch upgrades and make it easy and streamlined. By contrast, just in one operating system, some of our competitors had 78 last year. So we forced upgrades 2 times because of software quality defects that we take to heart and we own and we wish we'd never do again, and they did it 35 times more. And that meant more outages, more upgrades, more software qualification, more downtime and incurred more operating cost. I think when customers see that and monetize it, and they see not only the combination of risk and true in enterprise risk but also that this risk lets some affect critical infrastructure that literally every single security control in their enterprises built on the integrity of that network, and if that integrity is compromised, their entire enterprise risk management is compromised. So when our customers look at that, yes, I think there's really significant quantifiable reasons when it comes to managing enterprise risk that causes customers to explore multi-vendor options as a necessary financial control, as a necessary risk mitigation technique. And then when they do a fair valuation across the vendor landscape, we end up looking really, really good as a safe choice, and frankly, the best alternative to the incumbents they probably have.

Yes, yes.

Well, we're getting actually a couple of questions interestingly that are the same of some of the questions I have, so I'm going to enter first some questions.

I mean you're actually on the right track with the softball questions. Okay.

Yes. One obvious question that I was asking but an audience member beat me to it was, how is work-from-home and COVID impacting you guys? Good, bad, ugly? What's been the impact the last few months?