Welcome to the First Quarter 2015 Arista Networks Financial Results Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded and will be available for replay from the Investor Relations section at the Arista website following this call. I will now turn the call over to Mr. Chuck Elliott, Director of Investor Relations. Sir, you may begin.

Thank you, operator. Good afternoon, everyone, and thank you for joining us. With me on today's call are Jayshree Ullal, Arista Networks' President and Chief Executive Officer; and Andy Bechtolsheim, Arista's Chairman and Interim Chief Financial Officer. This afternoon, Arista Networks issued a press release announcing the results for its first fiscal quarter ended March 31, 2015. If you would like a copy of the release, you can access it online at the company's website. During the course of this conference call, Arista Networks' management will make forward-looking statements, including those relating to our financial outlook for the second quarter of the 2015 fiscal year, industry innovation, our market opportunity and the impact of litigation, which are subject to the risks and uncertainties that we discuss in detail in our documents filed with the SEC, specifically in our most recent Form 10-Q and recent Form 10-K, and which could cause actual results to differ materially from those anticipated by these statements. These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. Also, please note that certain financial measures we use on this call are expressed on a non-GAAP basis and have been adjusted to exclude certain charges. We have provided a reconciliation -- reconciliations of these non-GAAP financial measures to GAAP financial measures in our earnings press release. With that, I will turn the call over to Jayshree.

Thank you, Chuck. Thank you, everyone, for joining us this afternoon for our Q1 2015 Earnings Call. I am pleased to report our fourth consecutive beat as a public company. Consistent with prior quarters, customer demand for our 7000 Series products drove our results that exceeded the consensus. From a geographic perspective, our customers in the Americas generated 71% of our sales. A significant and major contributor to the international increase came from the U.S. cloud customer expansion to both Europe and Asia. Without this, our organic international was 20% and the Americas was 80%. Revenue grew 52.8% year-over-year to a record $179 million. For the first time, service contributed more than 10% of the overall revenue as we are now breaking it into a separate category. We delivered non-GAAP gross margins of 66.1%, resulting in a doubling of our non-GAAP earnings per share to $0.50 since Q1 2014, growing profitably in our competitive and dynamic industry. We now have over 3,200 customers and continue our acquisition trend of 1 to 2 new customers per day. This quarter, our key highlights include, we announced our flagship Extensible OS as a disaggregated cloud-based-software-only purchasing model. EOS as a Subscription, called EaaS, offers customers balanced CapEx and OpEx flexibility. We expect the results of this model will be more measurable in the next 1 to 3 years. We formalized our converged infrastructure solution with Supermicro systems for high-performance 10- and 40-gigabit Ethernet computer networking. We forged technology alliances with Lawo at NAB 2015 for frame-accurate IP-based video routing and Infinera for metro-optical networking solutions. We believe that Arista is the first data center networking vendor to receive application marketplace certification from ServiceNow called Store. Customers can now have a more complete configuration archival, versioning and merging of DevOps and NetOps for cloud business platforms. In March of 2015, Crehan Research published their 2014 market share numbers for high-performance 10-, 40- and 100-gigabit Ethernet switching, minus blades. We believe this is a more holistic switching category than individual port speeds, which are less precise. Arista grew both in port share from 6.7% to 9.3% and in revenue share from 5.3% in 2013 to 7.7% in 2014. While quarterly trends may vary, we continue to expect that we will gain market share in 2014 -- 2015. Finally, for the first time, leading analyst, Gartner, recognized Arista as a leader in the 2015 Magic Quadrant for Data Center Networking. It was based on a number of factors, including our high growth, our technology solutions and our flexible software. Given that there's been no leader in the data center sector since its inception, we are particularly honored by this validation. Now I would like to turn it over to our Interim CFO, Andy Bechtolsheim, for Q1 2015 details. Andy?

Thanks, Jayshree. Before I discuss our financial results and guidance, I would like to note that except for revenue numbers that are GAAP, all financial numbers I will discuss are non-GAAP unless stated otherwise. A reconciliation of our selected GAAP to non-GAAP results is provided in our earnings release. Also we, starting this quarter, will break out service revenue, which has become more significant to us and will break this out separately along with the related costs. Let me first comment on the 8-K that we filed earlier today, disclosing that we're going to amend our 10-K for 2014 due to a classification error in our 2014 statement of cash flows. The error was the result of a duplicate reporting of excess tax benefits associated with our equity incentive plans, resulting in a $17.4 million understatement of cash from operating activities and a $17.4 million overstatement of cash from financing activities. There was no change to the ending cash balance and there was no impact to our net income or the balance sheet. The error occurred due to inadequately designed controls surrounding our statement of cash flows, which resulted in a material weakness in our internal controls over financial reporting. We plan to remediate this control deficiency in the coming quarter by including additional review procedures to ensure the accuracy of the calculations and related disclosures in our cash flow statement. Now let me turn to our Q1 results which reflect our continued market momentum as our customers embrace our innovative and best-of-breed software-defined cloud networking solutions. From a top line perspective, total revenue for Q1 was $179 million, an increase of 52.8% year-over-year and 3% sequentially. We grew profits faster than revenue with net income of $35.5 million, up 117% from a year ago, and fully diluted EPS of…

Thanks, Andy. And thanks for doing a fantastic job as our acting CFO this quarter. Andy and I are also pleased to report that we have completed our search for a Chief Financial Officer for Arista. Effective Monday, May 18, 2015, Eda Brennan [ph] will be Arista's new CFO. Eda [ph] brings over 2 decades of finance, executive and operating skills that align closely to Arista's culture, values and goals. We welcome her enthusiastically and wholeheartedly to our leadership team. So in conclusion, we are off to a good start this year. I am not only proud of our results, our customer traction and our growing market share, but very encouraged and excited by Arista's increased market opportunities ahead.

We will now begin the Q&A portion of the Arista earnings call. [Operator Instructions] Operator?

[Operator Instructions] Your first question comes from the line of Jess Lubert of Wells Fargo.

Let's see if I can squeeze 2 in here. But I had a question first on the outlook. You're guiding below seasonal trends during the June quarter, so I was curious if you pulled some business forward into Q1 that was maybe impacting Q2 or if there was some other reason you're expecting less seasonality in what is typically a stronger period, especially for some of your cloud Titans? And then the second question, I was hoping you could provide some more details regarding which products are being made available on a subscription pricing basis, how broad of a customer set is interested in this method of consuming your offerings and the impact that the availability of subscription pricing is likely to have on growth going forward.

All right, Jess. I'll take your first question. I think our guidance is cautiously optimistic and conservative. Not having a new CFO, we stayed with the current consensus estimates for the most part. The cloud Titans were a large contributor to Q1. Q1 is historically our weakest quarter seasonally. And then Q2, Q3 and Q4 are all strong quarters. So I wouldn't call our Q2 as the only strong quarter. Hopefully, all 3 will be. And so we feel comfortable with that guidance. I think we have an opportunity to meet it and maybe beat it.

And then on the subscription pricing?

So specifically, we introduced EaaS specifically on our high-volume platforms, the top-of-rack switches, the 7050 Series families, the main one, that comes in different form factors, 48-, 64-, 32-port, 40-gig, and the goal here really is to start offering our customers options. We've always said we're a software company, and we can now make software available on the Arista blue box, if you will, on a monthly subscription basis. Most people, most comfort buyers, like to see the 2 bundled, but I think, as I said, I think this will take time to absorb. But particularly with top-of-rack switches, it makes a lot of sense because the lifespan of those switches are 2 to 3 years rather than 5 to 7 years.

Can you maybe just talk about the customer that's showing interest in this? Is this mostly cloud customers? And what are you hearing from...

And you've now come to your third question, Jeff, so you will be off the charts now already. But no, typically, cloud customers tend to look at scale and volume. This would apply more to entry-level customers who are starting with small projects or smaller companies that don't have immediate large-cap expend and are looking for that CapEx, OpEx balance and flexibility in their budget.

Your next question comes from the line of James Faucette of Morgan Stanley.

I wanted to ask, I guess, a couple of related questions and hopefully we'll count those as one. First, we've started to hear about new products from competitors based on some of your suppliers, Merchant Silicon, anticipated for later this year. And I'm just wondering if you can talk about timing for that and your expectations for being able to maintain market advantage and I guess, time-to-market advantage. And then I guess kind of consistent with that, also wondering if and where we might be seeing an expansion of the use of Arista's switching products in applications that are not necessarily directly switching, but are more akin to routing, et cetera?

Okay. So as you know, our software has always been architected to support multiple software architectures. Today we support 3. And we are very nimble and very agile and usually being the first vendor to support new silicon architectures. So our leading vendors here had been Broadcom and Intel. We will be looking to add silicon diversity, both in new chips from these vendors, and we will be looking at other vendors as well. But we tend not to preannounce it. Usually, the silicon vendors will announce the silicon, and product from Arista can be anywhere from 9 to 12 months after that. Andy, do you want to add something to that?

Yes. I mean, we obviously don't preannounce future products, but I will say, we expect to be among the first who will ship new silicon once it's production-worthy and if it's -- once we get production parts.

And we're particularly excited about not just the 40- and 100-gig options, but also the 25 and 50 that's emerging later this year. What was the second question?

Sorry I was just asking about if you're seeing any cases where customers are using your products outside of traditional switching applications like, for example, in -- that you just typically assume routing.

Yes. So to us, switching includes routing. We have always been shipping options for both switching at Layer 2 and wire speed switching at Layer 3, a.k.a. routing. So when you look at our leaf and spine, now, products, many of them, a large level of them go out with routing features. And this has been true for us for the last 3 to 5 years. We will be looking to expand into routing adjacencies beyond that as well, so routing features vary depending on whether you're on the leaf or whether you're on the spine or whether you're on the core. Arista does not address the core routing market today. We largely interoperate with Juniper or Cisco or Alcatel in that case. But some of these core features, one could argue, are changing and particularly with the advent of large-scale ECMP and MLAG and VXLAN, where you can scale to 6 million nodes at L2 over L3, there's no reason a data center interconnect could not be a core routing or L2 over L3 fabric as well. So I think many routers, the noun, will turn into routing, the adjective, and Arista's already in the middle of that.

Your next question comes from the line of Brian Modoff of Deutsche Bank.

Can we just talk a little more around the new products? And we got Tomahawk coming later this year. How do you see that impacting your business? And can you give us your opinion to what you think about its clients? And then how do you see the 25- and 50-gig port offering? Do you see that as additive? And then finally, just kind of back on your guidance, you're saying then you feel like your guidance right now is really just the fact that you have a new CFO and you're not about to put aggressive numbers out there for the new CFO to deal with?

Yes. So I think just to reiterate, we feel good about our guidance, we want to be able to meet it, and there's always room for improvement. But I'd remind you that we're still growing at double digits on a fairly large base of revenue. So don't knock that guidance too much. But going to back to your question on new products, Tomahawk, yes, very much an important silicon for 25- and 50-gig. I think much of this is the tail end of the year, really an impact on 2016. That's important to remember because it takes customers time to test, qualify. We don't look at products as just silicon and hardware. We also have a tremendous amount of software releases coming. We tend to go at the rate of 3 or 4 software releases a quarter.

A year, a year.

Our last release -- sorry, a year. And our last release just had 120 new features across the leaf and spine and in many cases, core routing as well. So very comfortable that not only are we expanding our performance and scale, but also our feature velocity and programmability in many of these cases. So -- and yes, we will look to add silicon diversity, but we're not ready to talk about chips that are not available.

Your next question comes from the line of Kulbinder Garcha of Crédit Suisse.

I just have a question really around customer concentration and the verticals and geographies. Just Jayshree, can you just clarify? Did you say that Americas grew 80% and international grew 20%? Did I hear that right? I'd say that broke up in the call. Am I...

No, no, no. I said the Americas contribution was 80% and international was 20%. And I was clarifying that that's the organic international. If you take the expansion of our cloud customers who deployed internationally, then it was 70%, 30%.

And then basically, in terms of customer concentration in the quarter, just how many 10% customers were they? And are all 4 verticals roughly still running at 25%? And is that something that you can hold as you go through this year, do you think?

Yes. So first of all, we don't break out customer concentration on a quarterly perspective and -- but I'll reiterate that the only customer concentration that we have announced and we still would talk about is Microsoft. So they remain the only one, if we were to have broken it out. The cloud Titans contributed very strongly this quarter. And so they were more than 25%. All 4 verticals did contribute. But in order of contribution, I would say, cloud first; service providers and Tier 2, second; high-tech enterprise, third; and financials, fourth.

And just one follow-up on that. And if that's the case, your gross margin still exceeded. And so is there something about the gross margin structure of the business that is presently continuously surprising because I would have thought the mix was that high towards cloud. It would have brought you more towards perhaps the lower end of your gross margin rate, but obviously, it hasn't.

No. Look, I think we -- you should continue to expect to see our gross margins trend downwards, not only because of the cloud concentrations, but even as mainstream customers buy high volume, that they're going to get a cloud type of discount. So I wouldn't want to leave you with the message that our high-gross margins this quarter is a trend. I will continue to reiterate that our long-term view is 60% to 65%, and then our guidance for next quarter is 63% to 65%. So we're executing very well. We're managing our discounts with great discipline. But we continue to favor high discounts and low gross margin for large customers in general, including the cloud.

Your next question comes from the line of Simona Jankowski of Goldman Sachs.

Jayshree, I wanted to clarify first on the EOS as a service product. Of the customers you've attracted so far, have they been new customers to Arista who are deploying that? Or has it been existing customers who are choosing to use that payment structure? And then I also wanted to ask you about the strategic significance of your tie-up with Infinera.

Okay. So first of all, it's a very new introduction. We just introduced that in March. So it's still too early to call in terms of a lot of new customer wins. They've been mostly been in single digits. But of the ones I recall, all of them have been new.

Okay. And then on Infinera?

I think Infinera is an example of an important strategic interoperability and solutions tie. It highlights, Simona, something you know very, very well and you've discussed, which is we can't be everywhere and we work closely with our technology partners to provide the best of capabilities in each one of us. So what we are really participating in Arista is in the data center and intra data center opportunity and obviously, a lot with third-party optics. But what we like about Infinera and particularly with their Cloud Xpress product line is we can now interoperate and offer solutions inter-data center across much longer-haul and distances. So there are obviously more operating at Layer 1, and we're doing still more at Layer 1, 2, 3 and 4.

Your next question comes from the line of Erik Suppiger of JMP.

On the litigation, you were about $7 million in the quarter, and that's going to $10 million to $12 million. Do you think that $10 million to $12 million is a reasonable base to go forward in terms of the litigation cost on a quarterly basis?

It's hard to predict litigation. I will tell you this is a complex case and expensive. I'm going to ask Marc Taxay if he can answer that question. We feel good about $10 million to $12 million for Q2, but now we are at the peak of our activity, aren't we?

I think that's right. I think Q3 actually -- Q3 of this year actually is probably going to be a little higher. The expectation will be just because the hearings in the ITC will be occurring then. But otherwise, I think, what we've got there is pretty consistent.

That sounds like a norm, but there will be some peaks. Okay?

Okay. And then one other quick one. Cisco was talking yesterday about some good growth in their IPFIX customer adoption. What are you seeing on the SDN front? How is that market evolving from your perspective?

I have to be honest here. It almost feels like we're in 2 different planets. I haven't seen any of the IPFIX customers, so they must be very, very much Cisco-specific applications. The 9K, which is also considered to be a strong contributor, my guess is it's 3% of their total switching revenues. So again, it's pretty insignificant. And our real focus is on building open programmable systems where customers are going from yesterday's enterprise applications to tomorrow's modern cloud killer application. So I think we have to make a distinction between IPFIX being bundled to customers and what's really in production. According to leading analysts, there's very few in production.

Your next question comes from the line of Subu Subrahmanyan of Juda Group.

My question was on adjacencies. Jayshree, you talked about this briefly, but when you think about WAN opportunities, be it a more routing-like functionality, maybe optical integration into packet platforms, can you talk about where you are in kind of the process of providing that functionality and when that can start to become meaningful from a revenue perspective?

Thanks, Subu. First of all, I want to highlight that one of the beauties of adjacencies is we want to be able to develop the ones that our software and our Merchant Silicon is naturally adhering to rather than go off and create a new one. So as we look at adjacencies, they have to be part and parcel of our engineering strength and our go-to market strength. That's a key tenet for us, that we're not going into a completely different market or segment or a decision-maker or buyer. So when you look at the kinds of adjacencies that we've already done, I want to talk about the ones we've already done, there's 2 that come to mind. One is we're already supplying through merchant optics a variety of short-range, long-range, single-mode, multimode, universal optics that provide a huge range of interconnect options for our leaf and spine solutions. And as you know, we were the pioneer of leaf and spine and cloud networking back in 2010. A second example of where they're already in adjacencies will be data analytics. A set of features we introduced over a year ago called DANZ data analysis is now really a nice option where customers don't have to buy a dedicated switch and a dedicated appliance for their data collection and monitoring and analysis. And this is a big deal for big data and structured and unstructured data. And it ends up being just a software option rather than dedicated hardware. And then there's what's you bring up. I think the advent and embracing of large multi-cast and route scale to [indiscernible] is going to be a natural feeder at the signs for us to do more routing. And there's 2 aspects to that. One is [indiscernible] is starting to have larger and larger routes. You're going from 60,000 routes to potentially 100,000 routes, to even 1 million. And the second is our software is getting more and more capable in this area. And this is why EOS has moved from -- what was it, Andy, when you founded the company? 1 million lines of code to over 8 million? And this is what routing and scaling protocols and programmability and working with all of our partners is doing for us. So I hope I answered some aspects of your question, but there's a natural adjacency there and I'm excited about the increase in TAM that could do for us in the next few years.

Your next question comes from the line of Scott Thompson of Wedbush Securities.

A question around the Savvis deal. It sounded as if that was strong this quarter. Should we expect that to continue? Could this be a trend where service providers come on strong for a few quarters? Or is it maybe just a sign that things are getting a little more competitive and you guys are mixing it up there? Can you comment a little bit on service provider and Savvis in particular?

Well, I could definitely comment on service provider, but I didn't mention any one by name, Scott, so...

Okay. Go ahead and comment on service provider. I guess I heard something that was mispronounced or something.

Okay. Maybe my accent. Service providers generally come in 2 or 3 flavors for us. There's the Tier 1 service providers. And as you can imagine, they're the most difficult to engage with and penetrate. And you can imagine, we've been working with them in large trials. And we tend to work with them on a project basis. And this represents a tremendous opportunity for us. But today, it's very early days for us. Then there's Tier 2 service providers and Internet-hosting providers and content providers. And this is where Arista has actually shined. We've done very, very well because all of the media streaming and content generation resembles in many ways and then the metro junction resembles in many ways distributed aspects of the cloud typing market. So everything from the scale, the features and the demand they put on us are somewhat similar to that. So I would say our success to date has been in those 3 segments, the Internet-hosting providers, the Tier 2 cable providers and the content providers. And I would say we've got a lot of room and scope there, and we've got a lot of room internationally, and we've got a lot of room in Tier 1 service providers. One of our areas of great focus, in addition to financials, is really a go-to-market and ahead of service providers, particularly in North America right now. So stay tuned and maybe we'll be able to share more wins, but we feel quite good about that.

Your next question comes from the line of Mark Sue from RBC Capital Markets.

Maybe the economics and the rationale of as you push more into market adjacencies is the goal to expand your wallet share with existing customers. Are they asking you to move into these areas? Or is it more a broader vertical expansion from your part? I'm just trying to get a sense of how we should track your early success as you spread your wings further.

Right. I think -- thanks, Mark. I think, first, it's important to reiterate that we've got a long ways to go in our primary market, which is the data center, which is at $10 billion going to $15 billion and within our 4 verticals. Our share, as I shared with you in 2014, is in the 7% to 9%. So we're looking to grow that into the double digits and more double digits and more double digits. No triple digits, for sure. So that's important, and that's where our primary focus is. In terms of adjacencies, the point I'm really making is the adjacencies are coming to us, both in terms of technology and as you rightly point out, in terms of customer demand. They are saying, hey, if we're using an Arista's spine, why can't we make the spine do other things, whether it's a data center interconnect or DANZ features or more optics or more routing. So we're just -- the spine -- and Arista is the only product who has a real spine. Anybody who says they're a spine is not. Many of them are spineless products. And so these have, in my view, the most capacity scale and capability and software advantages to offer our existing 4 verticals the best adjacencies.

That's helpful. And Jayshree, maybe just the Cisco lawsuit, is it having any impact on your rate of bookings? Any color there would be helpful.

Yes, yes. That's a good question. We said this in the last quarter, now reiterate again. I want to be really clear on that. The impact of the lawsuit with our customers has not been there. We do have to explain it. There's no doubt that we're spending a lot of time and money on the lawsuit litigation process itself, but our customers have been very savvy and understanding.

Your next question comes from the line of Ehud Gelblum of Citigroup.

So a couple of things. First of all, just a clarification. When you said, Jayshree, in an earlier question that you don't see IPFIX, I thought IPFIX for the most part it was proprietary and running on Nexus. And so anyone is going to actually consider IPFIX. By definition, it almost has a sort of a pure Cisco deployment to begin with, like they made that -- made the decision already. So I just want to confirm or clarify that it will be odd for you to actually see an IPFIX consideration because you wouldn't have invited that switch party, anyway. I just want to confirm if that's true. And that will lead to my other question about overlay controllers. How often -- obviously, you're partnered with NSX. You're partnered with Nuage. How often are you seeing deployments, if you got the overlay controllers, versus how often are you seeing bare switch deployments? And is that ratio kind of changing or kind of trending over time? And then I have a follow-up.

Yes -- no, they're both good question, really. On the IPFIX, yes, that's exactly right. Maybe we are on 2 different virtual machines and planets. If IPFIX is being installed with catalysts and existing Nexus, there's a good reason we don't see it. But we do see the Nexus 9K, so that's why I'm a little surprised we see that, but we don't see the IPFIX. So I think it really goes down to there's a set of customers, especially our 4 verticals, that really demand open standard innovation. And I'm not really looking for vendor lock-in. And like you pointed out, there's probably a set of customers they're not in, which is more the traditional enterprise that is designed to be end-to-end Cisco, if you will. And so to answer your second question on overlay controllers, as you know, for the 2 years that we've all talked about SDN, Arista's been adamant that we're having to focus on building a programmable underlay software stack. It's taken us the better part of 10 years to do that. It's programmable at, at least 6 levels, kernel, data plane, management plane, control plane, application plane and customer scripting plane and virtualization plane. But our goal has always been BYOC, bring your own controller. And one of our favorite partners is VMware, both with the vSphere vCenter and NSX. We see a lot of customer activity together. It's been early days. It's really been only -- we announced this relationship on -- specifically on NSX last August, so it's been about 6 months. And we are very pleased with the momentum and progress there. We do work with other controllers, and we also connect natively very many times into OpenFlow or OpenStack. So we don't preclude any of those options because we have all the right northbound and southbound APIs. We also have many savvy customers who want to play with our SDK and customize it too. So all of the above options are very, very valid, but it all starts by, "You can't do much with an overlay controller if you don't have a good software underlay."

Right. And finally, your inventory was up $10 million sequentially. Last quarter, it was up $20 million sequentially and you made the comment that it was up so much because "supporting revenue growth." The fact it was only up $10 million versus last quarter's $20 million, how should we read into that? And does it look like perhaps there was -- was some of the strength this quarter due to builds and -- that happened late last quarter, that you got inventory in place for but couldn't quite deliver by the end of the quarter? And so did that help boost this quarter? And are we kind of back to some sort of a normal trend? Is that the best way to read into...

Thanks for helping answer the question and the answers for these. I'm going to ask my CFO to take that one. Andy?

Yes. Obviously, we had an increase in deferred revenue. It's a significant increase in deferred revenue this quarter, which goes hand-in-hand with an increase in inventory.

Your next question comes from the line of Simon Leopold of -- from Raymond James.

A quick clarification that I -- hopefully is easy and then a broader question. I think Jayshree, you talked about a split between domestic and international of 80%, 20%. In the past, you've given us, I think, the split, Americas, EMEA and Asia Pac. I'm wondering if we could get sort of a like-for-like comparison on the regionals. And what I was hoping to ask you about was earlier, you talked about the 4 key verticals and I wanted to really understand the mirror image of that in terms of understanding the progress. How much business have you done outside of the historic 4 key verticals? What's been the progress? What measurements can we take in terms of understanding the movement into the other markets that you've talked about in the past?

Right. Okay. I'm going to take your last question first because otherwise, I have trouble remembering. So if you don't mind. And so our 4 verticals, Simon, really account for more than 90%. That's why they're our top 4 verticals in a huge way. Because even when we do some enterprises, generally, they tend to be high-tech enterprise and we put them into high-tech enterprise buckets. So at this point, I would look at us as very much driven by these main 4 verticals. And there's not a lot we're expanding yet or going into beyond these 4 verticals at the moment. And there's so much to do in these 4 verticals. It's something I'd reiterate. Just to clarify the geography, I hope I didn't confuse you, all. Actual Q1 2015 numbers is Americas at 71% -- or 70% and EMEA and Asia Pac at 30%. And what I tried to say in that 30% is that a contribution to that EMEA and Asia Pac of 30% came from the U.S. cloud Titans deploying there. So if I took those out, it would be 80%, 20%. And beyond that, we really don't break out the EMEA and Asia Pac separately. So it's 70%, 30% or 80%, 20% depending on how you count organic and inorganic.

Your next question comes from the line of Paul Silverstein of Cowen.

This is Fahad in for Paul. If you could provide the total customer count for this quarter?

I think we said it, over 3,200.

Over 3,200, sorry. And lastly, on the competitive front, can you comment about the increasing threat from top-of-the-rack vendors entering into the leaf part of your business? Do you see an increasing competitive platform in white boxes?

No change for us. I know we've discussed white boxes quite a bit. The heart of a white box is the software. And anybody can build a box, but nobody can build a software. And I think there's really 3 types of software stacks: the classic, what I call Blob-OS; and the Cloud OS, which is what we're building; and then the build-your-own-stack, which you can do with many operating systems. But a majority of the customers don't build their own PC or server, let alone their network, which is so mission-critical. So a lot of discussion, but no change in competitive dynamics.

Your next question comes from the line of Tal Liani from Bank of America Merrill Lynch.

Two questions. One -- the easy one is on service margins. We only have 2 data points, we have last year and this year, on the quarter and it went down from 73% to 80 -- to 64% roughly. Can you discuss the service gross margin dynamics? I don't have anything in between, so I don't know if it went up and down and kind of what are the trends there. Second, it's more kind of a broader question. I read an interesting article this morning about Cisco actually having a white box strategy, taking the 3K, stripping out the software and letting Amazon put in their own software on top of it. Whether it's right or wrong, I have no idea, but the hardware is basically the same. Generically, it's the same hardware based on same ASIC that everyone else is using. So the question I have is how difficult is it for your key customer, your large customer, Microsoft, how easy or how difficult is it for them to switch out the vendor, to replace you with someone else? Are they -- can you speak about not the hardware part, but rather the software part? Can you speak about how you're entrenched into their network, into their system? Is it your software or their software, et cetera, et cetera?

Okay. So your first question was on service margins and why have they come down year-over-year. As you know, our revenues increased significantly year-over-year. It's an area we're making significant investment, both in people and in Software-as-a-Service investment. So I think 73% was too high and 64% is more real. So that's the first question. In terms of the strategy on leaf and spine and how sticky we are with customers. The spine tends to be a 3- to 5-year decision. It's extremely sticky, both in hardware and software. It's not the kind of thing they easily move from one to the other. And clearly, that's where a lot of our strength comes from. Leaf, in general, tends to be a more disposable unit. Usually, those switches and the silicon churn is every, what do you say, Andy, 18 months, 2 years?

18 months.

Yes. So the product cycle is at least 2 to 3 years. So in every cloud provider, in general, the lifespan of a leaf product is 2 to 3 years. The lifespan of a spine product can be 5 to 7 years. And so our stickiness is shorter in a leaf than in a spine.

Got it. And Jayshree, just one follow-up. Your position, you're entrenched into the 25-gig and 50-gig cycles for next year, you're one of the leaders in the space. I remember that from the beginning early on. The question is where can you innovate there? Meaning, if -- is it just about your operating system and is it -- and I'm asking the question in a dumb way, but is it that important or can it be replicated with other operating systems the same kind of value you bring in?

You actually ask the most insightful questions that I didn't answer in the previous time you asked it, which is anybody can churn chips and build boxes in Asia or here and they can be white or bright or blue. But Arista's -- and our competitors' value-add is really the software and the -- building high-quality rich protocol, stable, single-image software that solves your OpEx and CapEx problems is a nontrivial effort. It's taken us several hundreds of engineers. It's taken us 2004 to 2015, so over 10 years. And there's a real case, and many customers do ask us, would we port our EOS to other platforms? And as I've often said and I've said very openly, to me, that's not a white box, that's a software strategy that we don't preclude. And that's the real differentiator, whether it's on 25-, 40-, 50- or 100-gig. Executing on the hardware, you have to do every year. Executing on the software is a multi-year, multi-decade effort.

And your next question comes from the line of Sanjiv Wadhwani from Stifel.

Sorry, there's noise in the background, I'm at the airport. But 2 questions, Jayshree. On the international-domestic mix, understanding that they go 70%, 30% because the [indiscernible] is building outside of the U.S. Is that something that we should continue to see? I mean, are we going to see more 70%, 30% mix versus sort of the 80%, 20% you've had in the most recent quarters? And then second question, you kind of the partially answered it with Tal, but understanding that your software is a differentiation. However, there was one other key element with white box that is the support. And are you starting to see more white box in RFPs or more penetration of bright box? And again, I understand the software aspect, but the support aspect seems to be addressed by the bright boxes?

Okay. So the 80%, 20% has been pretty steady for the last 4 quarters on our domestic and international mix. We would like to see it improve, and that's why we're making significant investments on international and I'd certainly like to see it go from 80%, 20% to 75% to 85% for starters and then 70%, 30%. The international mix we had this quarter was really due to the fact that many of our cloud providers, who we've been selling to for a long time, first started their deployments in the U.S. and now in their second and third year are adding international expansion. And so that, I believe, contributed this quarter and can contribute to subsequent quarters because more and more of their success in the U.S. will require them to think of how they will do this outside the U.S., since many of them are worldwide organizations. And that's why from a measuring point of view, I want to make sure that Arista measures both the organic and the inorganic international success. So I view this as more cloud success than international success, to be very clear.

And on the bright box, Jayshree?

Yes, on the bright box, honestly, I have to tell you, Sanjiv, the discussion on it is far greater in the news and press and print than it is with our customers. Our customers are really not looking to be switch developers. They want a high-quality switch where they can reduce their CapEx and OpEx. And to the extent our software and our technology can tune both of those, there's not that many questions on how do they build a white box, there's very many more questions on how can they make SDN and cloud technologies work for them to reduce their OpEx.

This concludes the Arista Q1 2015 earnings call -- actually, hold on. We've got one more question, who just popped in. Alex.

Your next question comes from the line of Alex Henderson from Needham.

Jay, I was wondering if you could talk a little bit about what the rate of new feature additions on your software platform look like versus what you see from Cisco versus what you see from the white box competitors? What's the rate of velocity that you're seeing in your feature adds?

Very high. I think that's the beauty of our software architecture, right, everything from the published subscribed model to the high availability. We're able to check in code. We're able to test code. We're able to do automation and diagnostics. And we're able to define a feature, implement a feature and get a feature to a customer in 3 to 6 months that could otherwise take 2 to 3 years in classic traditional software releases. So we see that as a 2-part advantage. One is in our approach to how we write code, but the other is the underlying architecture. It's so programmable that we can compile and test and put it out very, very quickly and with very, very high quality and with very, very little effect on the other software agents because they're still isolated from a software containment and software [indiscernible] point of view. So this is really our hallmark. And this is why when you see the rate at which we've had features, it's faster by 3x to 5x more than other vendors and participants there. And this is because the company -- look, 2004 to basically 2008, 4 or 5 years, just building the software architecture. So it's the hallmark of the company and one we are proud of.

Could you tell me, do you see, as you track the actual number of feature sets that are added, that...

Yes, so like I said, like I said, we did 120 features in our last release. If I were to compare those to traditional companies, it would be 1/2 that or 1/4 of that or sometimes even 1/10 that.

How about versus Cumulus?

It's difficult to answer that question to a specific vendor, so I won't try.

Okay. And then second question, do you think you have a time-to-market advantage as a result of the hypervisor on your switch when the next-generation 25-gig chip comes out from Broadcom?

So again, one of the things -- and I'm actually going to go back to your question on how about white box features. I was able to get an answer. Thank you, Marc. Generally, white boxes add x features, legacy adds 2x and Arista adds 5x. That's a good range of contrast. So -- and going back to your question now.

Time-to-market on the Tomahawk.

Time-to-market...

Let me actually answer that. So the key -- one of the key attributes we have in the company is the single-image EOS across all of our platforms, which really include, of course, the next-generation chipsets as well, which means that customers who receive the next-generation chip technology will be able to run exactly the same familiar and fully proven EOS release that they're running today. So we don't expect any hiccup or any other delays in rolling out the next silicon because, again, the software is exactly the same as we're shipping today.

Okay. This time, this does conclude the Arista Q1 2015 Earnings Call. I also want to mention that we have posted a presentation which provides additional perspective on our Q1 2015 fiscal results, which you can access on the Investor section of our website. Thank you to everyone for joining us today.

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