Arista Networks, Inc. (ANET) Q4 2015 Earnings Report, Transcript and Summary

Welcome to the Fourth 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 of 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 Ita Brennan, Arista's Chief Financial Officer. This afternoon, Arista Networks issued a press release, announcing the results for its fiscal fourth quarter and year ended December 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 first quarter of the 2016 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 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 in this call are expressed on a non-GAAP basis and have been adjusted to exclude certain charges. We have provided 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 2015 and Q4 earnings call. I am pleased to report that we had a spectacular year with a strong finish. Annual revenue grew 43.4% year-over-year to a record $837.6 million while earnings per share increased 52.8% to $2.44. This was driven by our innovative platforms, differentiated EOS stack and CloudVision orchestration. Services contributed 11.1% of overall sales as cloud demand fueled our growth across our top verticals, especially the cloud titans. From a geographic perspective, our customers in the Americas contributed 76.8% of total revenue in 2015 as our international theaters progressed steadily. We delivered non-GAAP gross margins of 65.3% for the year as we grew profitability in a highly dynamic and competitive industry. We are now in excess of 3,700 cumulative customers, including new customer upticks and expansion from our existing customer base. Microsoft was 12% of our 2015 revenue, symbolic of the continued shift we see from on-premise enterprise to workloads in the public cloud. We remain fully committed to significant R&D and customer support investments, often in conjunction with our key technology partners to bring open, industry-wide cloud conversions. Let me quickly recap some of our 2015 highlights. Early in the year, we announced 26 new leaf and spline [ph] platforms for 10, 25, 40, 50 and 100 gigabit ethernet, all based on our single binary image EOS philosophy, utilizing diverse silicon architectures. In May 2015, we achieved the coveted leaders quadrant with one other vendor in Gartner's Magic Quadrant for data center networking out of the 11 vendors mentioned. We forged a key agreement with HP for conversion for structure of compute and storage integrated by HP OneView and Arista EOS. We introduced a revolutionary platform, CloudVision, for workload provisioning and workflow automation, offering interoperability across overlays like VMware NSX as well as integration into Dell, HP and OpenStack frameworks. Later in the summer, we extended CloudVision to security with Macro-Segmentation Services, or MSS for short. Holistic security is a top-of-mind customer mandate, and we are working well with security visionaries like Palo Alto Networks, Check Point, F5, Fortinet and VMware to realize it. In fall of 2015, we launched our media and entertainment initiative with industry leaders, transforming the once analog world to a highly digitized, software-defined and IP one. In November of last year, we delivered our first spine Internet booking solutions connecting between data centers with Layer 1, Layer 2 and Layer 3 options endorsed by our leading customers: Box, Equinox [ph] and Netflix. Earlier this year, in January of 2016, we announced a foundational Arista EOS infrastructure with NetDB, a network-wide, state-oriented database. NetDB builds upon our core state architecture by adding network-wide actions including state-sharing mechanisms for control, replication, rollback and streaming analytics. We also announced use cases for hybrid cloud, Docker container support and more programmability models with OpenConfig and Go. All of these are compelling examples and alternatives to the traditional world we all live in. Demonstrably, we have had a busy year, growing our employee strength to over 1,200 with worldwide investments across 70 countries and 120 support people. And as we enter the new 2016 fiscal year, we believe we are gaining market share and increased relevance in the cloud networking segment. We are the only authentic state-based cloud stack, differentiated from mega-scale networking. Others may mimic us with marketing, buzzwords and architectures. We simply execute and deliver. We believe our publish-subscribe-notify is simply several years ahead of the industry, and I couldn't be more proud of our customers' steadfast acceptance of Arista as well as our team's results. Now I will turn it over to Ita for Q4 and 2015 financial details.

Thanks, Jayshree, and good afternoon. [Audio Gap] legal costs associated with the ongoing lawsuits and the release of a tax gap reserve described in the prior quarter. A full reconciliation of our selected GAAP to non-GAAP results is provided in our earnings release. Total GAAP revenues in Q4 were $254.4 million, up 41% year-over-year and above our guidance of $238 million to $242 million. Revenues for the year came in at $837.6 million, up 43% from the prior year. As anticipated, our cloud titan vertical demonstrated particular strength in the fourth quarter combined with solid contributions from the other key verticals. Service revenues continued to grow, reaching 11.5% of revenue for the quarter, up from 11.1% in Q3. International revenues came in at $47 million or 19% of total revenue, down from 22% last quarter. We experienced continued growth in EMEA, offset by some lumpiness in APAC. Overall gross margin in Q4 was 64%, in line with our guidance of 62% to 65% and reflecting an anticipated shift in revenue mix toward the cloud titans in the quarter. Operating expenses for the quarter came in at $85.7 million, up from $83.1 million last quarter. Total spending grew 3% on a quarter-over-quarter basis, considerably slower than our 13% topline growth. R&D spending decreased by $1.5 million, with lower-than-anticipated variable and prototype spending. Sales and marketing expense increased to 11.4% of revenue in the quarter, reflecting additional headcount and higher variable compensation expenses in the period. Our operating income for the quarter was $71.3 million or 29% of revenue. Other expense for the quarter was $0.9 million. And our effective tax rate was 18.5%, resulting in net income for the quarter of $57.5 million or 23.4%. This effective tax rate of 18.5% reflects the full benefit of the renewal of the research and development tax credit for 2015. Excluding this effect, our non-GAAP effective tax rate for the quarter would have been 30.4%, resulting in non-GAAP net income of $49.1 million or 20%. Our diluted share number for the quarter was 72.1 million shares, resulting in a diluted earnings per share number of $0.80, up from $0.59 in the prior quarter and up 50% from the prior year. This earnings per share number includes a $0.12 benefit for the renewal of the research and development tax credit for 2015. Legal expenses associated with the ongoing lawsuits came in at $9 million for the quarter, slightly below our outlook of $10 million from the last earnings call. As a reminder, these expenses are excluded from the non-GAAP results discussed above. Now turning to the balance sheet. Cash, cash equivalents and investments ended the quarter at $687.3 million. We generated $117 million of cash from operations in the December quarter, up from $10.6 million in the prior period. DSOs came in at 54 days, down from 68 days in Q3 and 57 in Q2. We have made significant process improvements in this area during the quarter and are pleased to see this reflected in the reduced DSO metrics. Inventory turns were 3.2x, up from 2.6 in Q3. Inventory decreased to $92.1 million in the quarter, down from $110 million in the prior period. This reduction is the result of increased focus on inventory management and supply/demand planning over the last number of quarters. Our deferred revenue balance was $197 million, up from $191 million in Q3. This balance continues to be largely made up of short- and long-term service contracts with some product deferrals related to acceptance terms of future deliverables. Accounts payable days were 45 days, up from 38 days in Q2, reflecting the timing of inventory receipts and payments. Capital expenditures for the quarter were $5.3 million. Now turning to our outlook for the first quarter. We are pleased with our 2015 financial performance, with revenues up 43% and earnings per share up 58% on a year-over-year basis. As we look forward to 2016, we believe that the fundamentals of our business remain strong. Workloads continue to march towards the cloud, benefiting our key customers, who continue to deploy our differentiated product set across their data centers. It is typical for Arista to see activity in the first quarter somewhat muted as customers complete their budgeting and planning processes. Based on our current visibility, we expect revenues for the first quarter of 2016 to be in the range of $232 million to $240 million, representing a slight reduction from Q4 '15 levels but still showing growth on a year-over-year basis of 30% to 34%. Our revenue mix is expected to continue to be more heavily weighted towards our cloud titan customers. And on that basis, we expect gross margins in the 62% to 65% range. With this as backdrop, our guidance for the first quarter, which is based on non-GAAP results and excludes any noncash stock-based compensation expenses and any legal expenses associated with our ongoing lawsuits, is as follows: Revenues of approximately $232 million to $240 million, gross margin of approximately 62% to 65% and operating margins of approximately 26%. Our effective tax rate is expected to be 28% to 30%, with diluted shares of approximately 72.5 million. Please note that based on our current understanding, we expect costs associated with the lawsuit to approximate $9 million for the quarter. With that, I will turn the call back to Chuck. Chuck?

Thank you, Ita. We are now going to move to the Q&A portion of the Arista earnings call. [Operator Instructions]

[Operator Instructions] Your first question comes from the line of Brent Bracelin with Pacific Crest Securities.

Jayshree, I wanted to start out with kind of the cloud vertical. Obviously, you had a strong showing here in Q4 with the cloud titans. You're guiding to another strong quarter in Q1. We've seen other companies on the enterprise side, specifically Cisco and NetApp, start to see a little slowdown in the enterprise side. So one, what's your view on the strength and the visibility and the continued strength in cloud this year? And could you offer any sort of visibility, what you're seeing on the enterprise side as well?

Thank you, Brent. Well, I'm no macro economy expert, so I'll leave that to the pundits. But in our micro cloud world at Arista, we continue to see strength in the marketplace. Cloud titans had a strong quarter in Q4, and we expect to see that in Q1 and beyond as well. I'm especially encouraged by the rapid shift we are witnessing of enterprise workloads to private, public or hybrid cloud. So I think the cloud economy trend is pretty inevitable to me, and I see that in 2 forms. I see that as hybrid cloud implementations in the enterprise and, obviously, public cloud implementations, where Arista provides product today to our cloud titans. All 4 verticals did well, but clearly, the cloud titans did better. But I would say the other 3 were strong and grew well, too.

Your next question comes from the line of Stanley Kovler with Citi.

I just wanted to ask specifically about Microsoft, if it managed to hit 10% of revenue for 2015. And if so, if the math is right, it was down a few million, in line with the guidance you gave heading into the year, flat to down at Microsoft. What's the outlook specifically for Microsoft for 2016? Should we expect it to be 10% or more of your revenue into 2016?

Thank you, Stan. So just, first, a little quick review on history. Microsoft has always been our top customer with greater than 10% concentration. I believe in 2013, it was in the 20s. In 2014, it went down to the teens. And in 2015, we're now 12%. So as a percentage of total revenue, it's steadily going down. As actual numbers, it continues to be important and is actually steadily going flat to up. And we fully expect 2016 to be an important spend year for Microsoft. I think it'll continue the trend down in percentage but remain relevant in actual revenue contribution.

Your next question comes from the line of Mark Moskowitz with Barclays.

Just wanted to get a sense here. Could you talk a little more about the cloud penetration with respect to the cloud titans in terms of not only the cloud providers but also those who are doing their own private or hybrid cloud on their own? If you could talk about how the penetration runway looks there. And then bigger picture, your numbers obviously speak to themselves in terms of the strength you're seeing, but investors are going to start thinking about what's next. And so how should investors start thinking about Arista in terms of moving into routing over the next few years? And what that revenue contribution could look like over time?

Yes, yes. So thank you. I think, Mark, the way to think of this is our cloud penetration has still just begun. We may not be in the first innings, but we're definitely still only in the second or third innings. There's plenty here, right? Everybody's infrastructures just starting to build out. Most of them started in the last couple of years, so plenty of room for growth. At the same time, there's plenty of room for all of these cloud titans to experiment. We worked with Facebook and Wedge and OCP designs. We worked with Microsoft's open switch architecture. We have worked in the past with other cloud titans, who actually developed switches even before Arista was founded. So we believe these can mutually coexist, and the TAM is so large that you're always going to have a permutation and combination of leaves and spines. Arista's increasing focus will be to work more and more closely in building the world-class universal spine. And sometimes, we may work with Arista leaf. And other times, we may work with other leaves in the network. So we believe our penetration still has a lot of room for growth, and we can coexist very much with internal development. In terms of what's next for us, I think it's most important to know, our primary market is the data center, is the cloud. This is a tremendously large TAM, as you know. I think the total available market in 2015 is still in the $8 billion range. And here we are reporting $837 million, so lots of headroom there. Don't want anyone to forget it. And as for routing, routing is not new to us. We've been doing Layer 3 routing for some time. And we fully expect to build our BGP scale, our Internet cable routes and continue to penetrate -- we are very hopeful that the, how do I say this, the old core routing will become the new spine. And more of these functions will be naturally features on our spine. And I think that's a tremendous opportunity for service providers to consolidate their CapEx and for Arista to provide more OpEx savings as well.

Your next question comes from the line of Steve Milunovich with UBS.

Do you have any comment on the case with Cisco? And specifically, since the ITC judge's initial determination came out just a few weeks ago, if the tone of conversation with your clients has changed at all? Are they at all more cautious?

I'll take a bit of it, and then I'd love to have our expert and GC, Marc Taxay, my good friend, comment more. First of all, Arista is always happy to address specifics in the courtroom, not in sensational blog rooms. But having said that, our primary focus, as you rightly point out, Steve, has been our customers. And their support to us has been unwavering. Our goal has been to offer them appropriate design workarounds with nonintrusive upgrades, and they are very comfortable with our plans. And they are very supportive of what we are doing and how we are working with them. Marc, do you want to add some more?

Sure, Jayshree. Just to briefly update everyone on the status of the investigation. As you point out, starting with the 944 investigation, the administrative law judge did an issue -- issued an initial determination in that case on February 2. As a reminder, that case originally had 6 patents in it, 1 was dropped, so there were 5 remaining. Of the 5 remaining patents, 3 of them, the judge found that we infringed and 2 he found that we did not. It's important to note that of the 3 patents that were in question, they actually relate to only 2 features. The first addresses the manner in which our agents communicate state to SysDB, and the second relates to our implementation of private VLANs. We've said this before, and we'll repeat it here, Cisco is not asserted, and the judge didn't find in this case, that EOS contains any Cisco code. It's also important to note that PVLANs is a common feature that's offered by nearly all network switch vendors today. Going forward, we're seeking commission review of the infringement findings in the initial determination. The ITC will consider our arguments as well as Cisco's and issue its final determination on June 2. If the ITC finds a violation and issues remedial orders, that then goes to the U.S. Trade representative, who will consider whether -- to veto those orders during the 60-day presidential review period. That period will end on August 2. We're going to continue to make our case to the ITC, of course. And as a mitigation strategy, we're also going to be in the process of developing technical workarounds for these patents, as Jayshree mentioned. We are very confident in those workarounds. And if the ITC does issue an exclusion order as part of this process, we'll seek to obtain customs' approval that the workarounds are outside the scope of the orders in order for us to continue the uninterrupted importation of products. This will be a time-critical exercise needing strong execution. Of course, Arista will respect the final determination of the ITC. If the final determination is adverse, we'll work in good faith to implement the workaround so that, to the best of our knowledge, all products supplied to customers will be non-infringing. Now with respect to the 945 investigation, that's still under review by the ALJ. The administrative law judge will issue her initial determination on April 26. That decision will then be subject to review by the commission, and we expect for them to issue a final determination and any remedial orders for that case on August 26. That then goes to the presidential review for the 945, and it would end on October 26.

Thank you, Marc. I feel like I have a PhD in ICT.

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

Alex?

Your next question comes from the line of Jess Lubert with Wells Fargo Securities.

I wanted to squeeze 2 in. Maybe just first you can comment on where you are in the development of the workaround and your confidence in your ability to get those built and approved by the end of Q2. And then for Ita, I was hoping to touch upon margins as expenses came in quite a bit lower than we expected, and you're guiding operating margins to higher levels than we've seen you guide for the last several quarters even though there hasn't really been a change to gross margin forecast. So I was just hoping to understand if you feel like 26% or better is likely to be a sustainable operating margin level going forward? If you think you can make all the needed investments to support your growth at those levels, or if we should expect operating margins to dip as we look out towards Q2 and the second half of 2016?

Yes. I mean I think if you look at the approximately 26% operating margin guidance, it's pretty consistent with last quarter as well, right? And I think the way to think about that is as we ramp spending and make the investments that we need to make, we're doing everything that we need to do from an investment perspective. But revenues are growing at a pretty fair clip, right? And we're being very judicious in how we spend money. And sometimes, the revenue is going to grow faster, right. And that will result in some upside to that. I think our guidance takes a view of, here are the possible things that we need to go do in the quarter, and that's kind of reflected in that 26%. I think in terms of a long-term model, we've talked about maybe a 25% operating margin being a reasonable way to think about, and that gives us scope to grow sales and marketing expense as a percentage of revenue a little bit and maintain our investments in R&D. So I think that's how to think about it. I don't know if the 26% is anything new necessarily. And probably, long term, it's like a 25% operating margin model.

Thanks, Jess. No, I think we have been planning and working on what-if scenarios and cases. While we don't believe we infringed these patents, you always have to plan. And we have been planning over the last year. Our engineers have been hard at work. So we feel very comfortable that in the case of 944 that the majority of the workarounds will be available -- are being worked around right now and will be available in Q2. Marc, do you want to add anything more to that?

Yes, I think that's right. I mean our team has been hard at work. We're continuing to evaluate everything, of course. And we're in the process of issuing out a new -- releasing out a new version of the software, which will come out in Q2.

Your next question comes from the line of Ryan Hutchinson with Guggenheim.

So 2 quick questions here, just to round out that conversation on the legal. Ita, did you factor any -- did you change the methodology as it relates to guidance to factor any sort of disruption as it relates to legal? So that's more of a clarification. And then the question really is on another one of your top cloud titan customers, Facebook. They've been pretty vocal about their intentions to build out EU2 with -- 100% powered by Open Compute. What role, if any, will Arista play? And more broadly, how does the relationship potentially change over the course of time?

I'll take the first part of that question. The guidance for Q1 is business as normal, right, with support of our customers and not really assuming any disruption from a legal perspective.

That's correct. That's exactly right. And your second question on Facebook, Ryan, we have said a number of times, we fully expect our cloud titans to also do engineering. It's not the sole responsibility of Arista. But in many cases, this is codevelopment. And in many cases, it's complementary. So I do think Facebook will use a lot of their capabilities. But it's core versus context. It's clearly core to Arista, and we provide a lot of capability. And then we work closely with Facebook's application to give them the right control, so I'm very comfortable that there's room and opportunity for both.

Your next question comes from the line of Sanjiv Wadhwani from Stifel.

I'm going to try to sneak in 2. The first question, Jayshree, on Tomahawk-enabled switches, just curious to see if those are ramping in line or better or maybe even worse than expectations? And then the second question is really on the legal side. Any update on the antitrust claim that you filed against Cisco? And do you see that impacting the ITC or the District Court lawsuits in any shape or form?

Okay. So I'm looking at the Tomahawk data. Good acceptance of the products in general, especially the cloud titans tend to be the earliest adopters of this type of technology. I would say this is a key year for Tomahawk, meaning good acceptance of 10 gig, 40 gig and 100 gig. We still haven't seen 25 and 50 really take off the way we'd want to. So lot's been written about it last year. But I think the year of 25 and 50 gig will really be this year, and so that's critical. Regarding the lawsuit, once again, I'm going to let Marc talk about this. The only comment I want to make is we are -- when a monopolist asserts ownership of open standards, I believe it not only harms Arista but it harms an entire ecosystem and brings innovation to a standstill. So we really want to innovate. We don't want to litigate. We thought long and hard about this. And Marc, maybe you want to comment on why we did it.

Sure. No, I think that's right. And I think we evaluated the situation, and we felt that based upon what we had seen that Cisco was engaging in a series of anti-competitive behaviors. And we initiated the action really to level the playing field so we can win in the marketplace based on the merits of our products. So we did initiate this filing. We filed, in the copyright case, in the district court a motion for a leave to amend our answer to the complaint that was originally filed to include the 2 counterclaims. The first claim alleges that Cisco is engaged in an illegal bait-and-switch with the industry and its customers to protect its monopoly position in switching and routing. We asserted in that, that Cisco has aggressively promoted its CLI as the industry standard to encourage customers to adopt and heavily invest in the CLI without having to worry about vendor lock-in or having to invest in training and scripting for alternative CLI commands. Our claims allege that when Arista achieved success in the marketplace that threatened Cisco's position, Cisco reversed course and now essentially is using those customers' investments against them to achieve vendor lock-in and protect its dominant market share by seeking to preclude competitors from using its industry-standard CLI commands, contrary to its past representations. The second claim alleges that Cisco uses its market position in switching to punish its customers for buying competing products by prohibitively increasing the prices it charges for servicing Cisco switches where the customer has indicated they plan to buy switches from Cisco's competitors. So what we've done is we filed the amended complaint. That's being evaluated by the court now. The court will make a decision as to whether or not we can include this as part of the existing action or whether or not we should be filing a separate action. We'll find out on that shortly.

Thank you, Marc.

Your next question comes from the line of Alex Kurtz with Sterne CRT.

Marc, we're going to keep going at this. In your prepared remarks, there were a couple of really important sentences that I heard about the timing of the workaround, and it sounded like there was some request for an exclusion of the workarounds and -- as far as how the products could be continued to be delivered to customers until the workarounds were assessed by the court, something along those lines. Could you clarify that and maybe go through that again and what that means for revenue as we go through Q2, Q3 and Q4?

I don't think we mentioned anything on our prepared remarks. Can you clarify what you're asking?

Yes, Marc went into some detail in his prepared remarks, Jayshree, about the process of the workarounds as they go through the court. And I just wanted him to expand on that a little bit as far as the timing and what the gates you had to go through to get that approved.

Sure.

Sure. So as we said, we have been developing design arounds. And we're in the process of releasing a new version of EOS that will include that. The process for us would include seeking approval from customs in order for customs to ensure that we can import those products into the U.S., a non-infringing version of those products into the U.S. And so we'll initiate that process once the final determination from the ITC comes in. It's our view, of course, now we're still fighting this case. We're going to continue to fight it, and we've submitted our briefs over to the full commission. And then once the commission, if the commission issues an adverse ruling, at that stage will work with customs to try to go through the process as quickly as possible.

So just to be clear, we won't know anything about this until June 2, as Marc said.

Your next question comes from the line of Rajesh Ghai with Macquarie.

I had a question for Ita and one for Jayshree. Ita, can you talk about the deferred revenue? The number did not go up as fast this quarter. And the question for Jayshree is that the software-only SKU has been out for some time now. Can you talk about the -- what are you seeing in terms of demand for that software-only SKU? Are you seeing any increase in interest from the cloud titans or the web [indiscernible] Customers?

Yes, so just to take the deferred revenue question. I mean, the deferred revenue balance is going to move around quarter-over-quarter, right? The -- our service renewals are for different time periods, large customers, smaller customers and so on. And then even on the product side, it's all about the timing of shipments, right? So if we're shipping stuff at the end of the quarter to a certain customer, then it's going to be higher. If we ship it the first day of the following quarter, it will be lower. So it's not a metric that you're going to see follow the traditional -- what you would typically see as a very predictable cycle through the...

Yes, I think it's also fair to say, it could stay flat, it can go down, it can go up. I wouldn't read too much into that metric. It's not [indiscernible] customer. And Rajesh, regarding your question on software SKUs, we've offered that a lot and talked about it a lot. But at the end of the day, we've seen that more of our customers want to know we have it but prefer to use our hardware and software together. So the actual success and implementation and deployment of this has been small. The discussion of it has been large. There is a disproportionate difference between implementation and discussion.

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

Just a question, talking about geographies, et cetera. You mentioned that Asia saw some lumpiness. And I'm just wondering if you could talk about how we should expect that and those other international geographies to develop? And I guess I'm looking at it in the context of the feedback that we have received from a lot of channel in international is that it seems like Arista is having a hard time completely satisfying the demand in some of those newer geographies. So can you just talk about how we should expect these international geographies to develop? And what your sense is as to underlying demand trends, et cetera?

First of all, James, I think we're doing very well internationally, but it's hidden by the fact that we're doing very well in the United States. So I want to make that point. I'm especially encouraged by our success in Europe. We just brought in a very strong leader with 25, 30 years' experience in networking. And we're doing very strong in the especially developed countries: Nordics, Netherlands, Germany and U.K. In Asia, what happens is Asia is not one country. It's really Australia, Korea, Japan and Southeast Asia, including India, where our greatest focus is. So first of all, we don't -- we are reasonably low in China. And so our numbers kind of tend to jump around quarter-by-quarter, depending on the deal size and what we win. So we had some fairly large customer wins in Q2 and Q3 than we did in Q4. In terms of the difficulty we are having, I think we're still in the very early stages. I don't see difficulty. I just say we started later. So our real first year of international focus, I would say, is 2015. And I'm fully expecting to see more -- and one of the great statistics for me of success in end market is new customers. We are seeing more new customer acceptance in international than we even are in the United States, which, to me, suggests that the next few years will be very promising internationally.

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

Jayshree, can you just clarify the comments you made around the Tomahawk-linked products, how are you expecting them to ramp the success of [indiscernible], you kind of cut out when you were -- at least on my side -- when you were responding there. That's my first clarification. The second was just in terms of verticals. If I think about financials, big 7 cloud -- well, big cloud titans, other service providers, media -- I don't know how you want to split the verticals now versus previous -- how you previously talked about it. But is it still quite balanced when you look at the 2015 numbers for revenue?

Yes. I think I understood your Tomahawk question, but I'll try again. And Kulbinder, hopefully, if I don't answer it, we'll chat more on the call back. But we are pleased with the transition on Tomahawk. We have not seen it become mainstream yet. Trident is still the main technology, Trident I, Trident II, et cetera. And you have to understand, we've really only had 1 quarter of Tomahawk. We introduced the products, but the actual qualification and production and installation has been 1 quarter. So I see this year as the real indicator. And the other reason I think we haven't seen an immediate pickup is customers are happy with our 10, 40, 100 gig current offerings. They're meeting their features. They're meeting their price points. So the reason to switch to Tomahawk would be specifically more need for 25 and 50 gig, which requires more of an ecosystem: you need the host adapters, you need the NICs, you need the compute, you need the storage, you need everyone to come together. So this would be a more crucial year for Tomahawk than last year was. And your second question was on -- I mean...

Revenue breakdown vertical for 2015.

Yes, so verticals haven't changed. We continue to enforce our top 4 verticals. What I can tell you, looking at our 2015 order or priority or success of the verticals, if you will, is number one, was clearly cloud titans. We -- number 2 was service providers and Tier 2 cloud providers. And financial services and high-tech enterprise were tied for third place.

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

I wanted to just get a sense for your plans on growing the sales organization. Where are you in terms of sales force expansion at this point?

We're doing well. We could always be doing better. I think one of our sales models, if you look at it, which is different from others is Mark Smith and Anshul and Mark Foss, they're not focused on coverage. They are focused on our 4 verticals and what we call the Arista 100. We are looking to make an engagement that's both long-term and impactful and strategic, where they really want to transform the data center. So if it's a regular enterprise IT customer, they're probably less interested in Arista. Because they'll just keep buying what they are, and it will be hard to change old habits. And so our sales model is very much targeted on: Go deep rather than go wide, and really focus in a few good geographies and obviously the top 4 verticals than everywhere. And we are doing very well there. We've hired well both in sales and technical SEs [ph].

Can you give us a breakout on the -- on your -- on the size of the sales organization?

I don't think we've done that historically. But if I look from year-over-year, we've definitely increased by double-digit percentages.

Your next question comes from the line of Jeff Kvaal with Nomura.

I have a question, a return to a theme that you just hit upon, Jayshree, which is new customers. Can you help us maybe think through how much of your business each quarter or each year is driven by repeat orders and how much by new customers? And then the second part of that is, are there particular new customers or suites of larger new customers that you are excited about for 2016?

Yes. No, those are both very good questions, Jeff. I -- new customers is something I personally measure a lot, not because they're the biggest contributor to core [ph] quarterly revenue, but it's basically a definition of seeds you sow for the future, right? In terms of actual contribution of revenue, they tend to be, I don't know, single, double-digit percentages. They're not big. And the reason I say that is because the new customer orders aren't big. To understand that, if you look at a typical Arista sales cycle, our first order is a SDN starter kit or a particular project that could be 50K or 100K. So even if we have lots of new customers, that doesn't add up to a lot of millions of revenue. But what you have to look at is not what the first customer is, but what happens 6 to 12 months and 18 months later. That's when we really get them to be -- and this is addressing your second question, you get them to be a relevant, big, million or multimillion dollar customer. So that's the pattern we see. We see a new logo and a small project, if you will. And then within 12 to 18 months, a more substantial win.

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

So I've got 2 questions. They're both on the kind of on the chip side of things. First off, has there been some delay in the timing of getting the Broadwell chip out from Intel? And has that slipped your time line for the full architecture of 25 gig stack rolling out a little bit in terms of timing into the back half? And then the second one, along the same lines, can you give us an update on what you're thinking on the Jericho product coming out and when that might impact the router market? I mean, that's an $8.6 billion TAM that will roll into the switch market at some point when that gets going. Just give us an update on those 2?

Yes. No, thanks, Alex. They're both good questions. I think you're absolutely right to point out that the 25 and 50 gig NIC, and it's not just Intel, it's Intel, Broadcom. I think Mellanox plays in that. And there's a lot of onboard LAN on motherboard-type activity. All of that has not happened in 2015, and we fully expect them to happen in 2016. And I think one of the biggest drivers will actually be storage. As you move to more scale-out, flash-based storage, there will be less and less fiber channel SANs and more and more Tier 2 IP storage SANs. And we think 25 and 50 gig is a pivotal performance to hit, while 1 and 10 gig may be more for compute. So you're right to point out that, that hasn't happened as fully and well and will in 2016. In terms of Jericho, the one thing I would just say in general about the silicon industry, since I came from this industry 30 years ago, is we always tend to overestimate the availability of a chip as opposed to the availability of a system. I think Jericho was announced by Broadcom a year ago, and we are very impressed with the table sizes and capacity and buffering and capability. But announcing a chip is very different than vendors announcing systems. I fully expect that systems will come into play in the summer of '16. And the qualification of those systems, because it's routing and spine, can take longer. So I think 2016 will be a year of trials for Jericho and core routing. And then the multibillion-dollar market you talk about will really be a multiyear execution.

Your next question comes from the line of Hendi Susanto with Gabelli & Company.

First for Jayshree, what is your grand vision on Arista products and solution in the security space after the new release of the CloudVision? And second question will be for Ita, how much CapEx should we expect in 2016?

I'll take the first question, Ita, and then -- so security is definitely a top-of-mind initiative for a lot of our customers, both in the cloud and in the standard enterprise network. Arista does not fancy ourselves as a security expert. We think that's a market in itself, and it just changes every time there's a new threat vector. And so what we want to do is work with the industry-leading pioneers and visionaries there. One of our closest partners is Palo Alto Networks. We are also working with Check Point and F5, Fortinet, VMware. And what we see is it's one thing to actually say, I detect a threat, which our security partners do. It's another thing to say, I avoid permeating a threat. And this is really where Arista comes in. With CloudVision, what we're doing is truly establishing how you can maintain firewalls and rules and policies with our security partners, but really map this from physical to virtual to advanced telemetry and orchestrate through the entire way so that you're not just catching one silo threat but making sure you can mitigate this across an entire network. So if you will -- somebody catches the disease, and we make sure it doesn't spread.

And then just on the CapEx, I mean we've been averaging about $5 million a quarter. Could you see that spike a little bit a couple million here and there if we add some capacity and stuff? Maybe. But it should stay roughly in that range.

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

I have a question on the 100 gig market. Is there any delay to the deployment of 100 gig because of the 25, 50 gig issues you discussed here? I understand that the technology also reduces the price for 100 gig. And the question is, if we should model slower ramp for 100 G as a result of what you're discussing?

Thanks, Tal. No, I don't think so. I think the fueling of the 100 gig, especially in the spine, is driven by aggregate of 10 gig, 40 gig today, and will only get more as 25 and 50. So what we see in our customer base is everybody is preparing for 100 gig spine, independent of whether their leaves are 40 or 50 or 25 or 10. And the real -- obviously, 100 gig in ports is much more smaller than 40 or 10 or 1 right now. But the real issue for 100 gig for us to solve and the industry to solve is making sure we can deliver incredible scale capacity density but also at the right price point, particularly on the optic side. So I think the embedded optics and the cost of that is a bigger issue than the 25 and 50 gig.

Is your comment on margins -- if I rewind back a few quarters ago, you guided that gross margin will decline almost every quarter. I think you even said 100 basis points every quarter for 4 quarters or something. And now we're seeing better margins. So is your comment on margins related to the delay or to the slow ramp of 25 and 50G?

No, not at all. I would say, Tal, that our margins, we did drop to 64% in Q4. And it's much more tied to mix, customer mix and product mix. But we feel pretty good about the acceptance of both 40 gig and 100 gig last year, and we feel quite optimistic on the additional acceptance of all this, this year.

Your next question comes from the line of Rohit Chopra with Buckingham Research.

A couple of quick questions. Just, Jayshree, on the competitive front, have been there any changes in the posture of incumbents as they continue to experience share losses? And then secondarily, have buyers changed any of their buying patterns? Are there larger or smaller deals as we sort of close the quarter, are sales cycles any longer? Have you seen any changes on the buyer front?

That's a good question. Have there been any aggressive competitive changes? I think it's the same as always. We -- when it comes to Arista, we do see that we receive our lion's share or fair share of aggressive competition, and so it continues to be heightened. But I wouldn't say there's any change. I would say it's always been heightened, and it continues to be so. Have we seen any shift in behavior? Not yet, no. We are seeing strength in the cloud. We're not the expert on enterprise, but the enterprises we serve tend to be extremely knowledgeable and capable and maybe, therefore, are a smaller percentage of the broad enterprise market. But in our world, in our micro Arista world, we haven't seen any dramatic changes.

Your next question comes from the line of Vijay Bhagavath with Deutsche Bank.

Jayshree, Ita, my question is around the weakness we and others are noting in enterprise IT spending in particular. So this a bigger-picture question, which is, would you anticipate the big banks, the high-tech customers you sell into, asking for hyperscale price points that currently a Google or an Amazon would demand from their vendors? And if so, how would you manage margins in this more compressed ASP scenario, where enterprises are asking for hyperscale price points?

Vijay, yes, that's a great question. And my answer to pricing is always the same, whether it's a customer or you, which is if you show me the volume, I'll show you the pricing. So if we have large enterprise customers who can drive hyperscale volume, we'd love to achieve those price points. Because it gives us a chance to drive cost reductions and, therefore, price reductions. And so there are some enterprises who can do that, and there are many who can't. So I think it depends on the sector. Ita, do you want to add anything more to that?

Yes. No, I think that's exactly it, right? I mean it's very much a volume-based pricing model for us, right? So yes, if there's volume there, we can drive the pricing and the cost [ph].

Excellent. And a quick follow-up, if I may, which is -- don't want to beat down the Tomahawk thread here. But my question is really around, should we view 100 gig Tomahawk switching as really a second-half-weighted opportunity, because some of the checks we've been getting is mostly early volume, small volume, top-of-rack upgrades going all at some web scale data centers? The spine 100 gig refresh is really back half. How should we think about a second-half-weighted versus kind of an equal weight of a 100 gig Tomahawk refresh?

I think your thought process on this is very good, Vijay. I think the only thing I'd add to what you just said, and I agree with it, is I think the first half will be heavily weighted to planning 100 gig spine and also deploying as many leaves, whether they're 10, 40, 25, 50. People don't think as hard about the deployment of leaves because they view them as disposable items. They have shorter depreciation cycles. And so I see that the 100 gig spine has definitely started already and will continue to increase in the summer and second half. But I don't necessarily see that as Tomahawk, I also see that as spines based on our Arad technology, leading to Jericho in the future.

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

First, a quick clarification if I might. The R&D -- pro-forma R&D in December was a little bit lower than what we were modeling in the prior quarter. Just want to find out if there was anything kind of onetime or timing nature that may be showing up in the March quarter, how to think about R&D. And then the question I wanted to get an update on is, any kind of quantification we can get to help assess progress on your partnerships? You talked today reminding us on the converged partnership with HP. And in the past, you've talked about VMware. I'd like to try to get some way we can assess progress or maybe outlook on partnerships.

Yes, just to take the R&D question, I mean that does move around a little bit just based on variable spending, prototypes, NRE, that type of stuff, right? So there's nothing unusual happening there, just really timing of those 2 activities.

And Simon, this is a question I'd like to answer better in forthcoming quarters. I believe HP, VMware and Palo Alto Networks and also the Docker container partnership we announced are all examples today of really technology marketing and how we work to sell together and, if you will, move from old-school networking to new converged or cloud networking. We haven't done a good job, and I think we owe that to you all, on successes and metrics we've had there. Because I think we've been more in the business development and marketing phase. We want customers to get that, clearly. I've been -- we've -- I'm particularly pleased with the successful combination we're seeing of NSX and Arista. We're seeing more of that now than we did, say, a year ago. But let's say let's take a rain check on that question when I can give you more concrete answers.

Okay. You know I'll hold you to it, too.

I know. I'm afraid of that. I owe you.

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

Jayshree, I have 2 or 3 related questions. They're truly related. First off, I know you are reluctant to give -- or let's just say you're not giving a more precise breakout in terms of the cloud titans versus the other verticals. But given your business model and how important the issue is from all perspectives, I'm hoping that you will give more granularity rather than that they're your largest switches, say [ph]. Can you at least tell us -- if you're not wanting to give us the exact number, can you at least tell us, are they closer to 50% of revenue than 1/3 of revenue?

None of our 4 verticals, Paul, are anywhere close to 50% of our revenue. They average in the teens to the 25%. That's the averages.

So you're -- all right. Well, the...

I answered your question.

You can't -- you -- obviously, you have to have a vertical that's greater than 25%, right? Unless they're all 25%?

No, you can -- that's not always true. You can have one of them peak over 25% in a given quarter. But over a year -- and I want to be clear when I say that, just because we have 4 verticals doesn't mean we don't have anything beyond the 4 verticals either, right, so...

Well, I'm clearly talking in the context of where you guys break out revenue when you're talking of 4 verticals in that context.

I think I've done my best to help you out by saying while we do not want to get in the practice of breaking verticals out, none of them or any of them are anywhere close to approaching 50%.

All right. I appreciate that. You're telling us cloud at most is 25% or thereabouts. All right. Let me move on. I'm not sure if you've done it recently, but you used to tell us your 10 largest customers on an annual basis back in '13, if I have the numbers correct, it was 43%. Can you give us an update on that? And then I've got a pricing question.

No. I think the only 10% concentration customer we reported in '14 and...

I'm not asking the 10% customers. I'm not asking the 10% customers. I'm asking what are your 10 largest customers and the percentage of total revenue.

Oh, I see. Ita, do you want to take that one?

Yes. I mean I think what I would say is it has been pretty stable, actually, right? It has been -- it is a significant part of the business, right? There's no doubt. I don't know if we've given the exact percentage, but it is a significant part of the business, right? And it has been pretty stable, if I look at it over the last 2, 3 years, it has been...

It hasn't changed.

It has not changed as a percentage of the business. The players inside that top 10 move around, and then the ranking moves around. But the actual percentage has been pretty consistent.

All right. One last question on pricing. I may have these numbers wrong. I apologize if I do. But I think I recall per port pricing declined for you by 15% in '14 and 10% in '13. Can you give us an update on that for '15, what average port pricing does...

We'll take this question offline. I think we don't have that level of detail on past years versus now.

Your next question comes from the line of George Notter with Jefferies.

I guess I wanted to go back to the international business. If I look at it in absolute dollar terms, I think it's been roughly flat for about 3 quarters. And obviously, it's a big percentage of the addressable market. Can you just kind of walk through what you're doing sales and marketing-wise to kind of get that business growing again? And is there some intuitive reason why it has been flat these last couple of quarters despite the larger business really growing?

I think just to clarify the numbers, George, I think EMEA has actually been growing quite healthily for us, and it has been keeping pace actually with the overall business. APAC, I think, to Jayshree's point, it's still at the stage where if we win a big opportunity in a particular country, it's going to change that dynamic a lot, right? So I think about it in 2 phases. EMEA is more developed, and it's more consistent. And we are seeing success there. And then APAC is still a little bit more opportunistic, and we've got more work to do there.

Yes. No, I think it was well said, Ita. I think we've got a lot more volatility in APAC, and we've got more work to do there. In EMEA, I think our numbers have been steadily growing. Sometimes our numbers look higher if the placement of our cloud titans is in a particular geography in Europe, which is not really organic growth, but it reflects in our numbers. But I think in general, U.S. and EMEA have been growing, and APAC has been volatile and lumpy.

We have time for one final question, and that question comes from the line of Simona Jankowski with Goldman Sachs.

I wanted to follow-up on the earlier discussions around Facebook and Microsoft, recognizing that there are some internal projects, but that you're still going to continue to sell to them. In aggregate, would you expect your business with them to be up or down this year to the extent you have visibility? And then related to that, would you expect any other customers outside of Microsoft to cross the 10% mark this year?

Simona, I'll take your first question, which is, no, I don't expect any customer to top 10% this year, but I expect them all to be big customers. Specific to Facebook and Microsoft, your question was...

Up or down this year?

It's too early to call. We don't get that kind of visibility for the entire year. Generally, as I've told you, our visibility is greater in 1 or 2 quarters than the entire year. And we feel good about the 1 or 2 quarters.

This concludes the Arista Q4 2015 Earnings Call. I also want to mention that we have posted a presentation which provides additional perspective on our 2015 fiscal results, which you can access on the Investors 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.