Welcome to the fourth quarter 2019 Arista Networks financial results earnings conference call. During the call, all participants will be in a listen-only mode. After the presentation, we will conduct a question-and-answer session. Instructions will be provided at that time. If at any time during the conference you need to reach an operator, please press the star followed by zero. 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. Charles Yager, Director of Product and Investor Advocacy . Sir, you may begin.
Thanks, 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 fourth quarter ending December 31st, 2019. If you would like a copy of this release, you can access it online on 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 2020 fiscal year.
Long-range financial outlooks, industry innovation, our market opportunity, the benefits of recent acquisitions, 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, most recently 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 on 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'll turn it over to Jayshree Ullal.
Thank you, Charles Yager, and welcome to your first Arista Networks earnings call. Thank you everyone for joining us this afternoon for our fourth quarter 2019 earnings call. We delivered a non-GAAP revenue of $562.5 million, with a non-GAAP earnings per share of $2.29. Services contributed approximately 19% of the revenue, consistent with the typically higher renewals at the year-end.
Our non-GAAP gross margins were 65.2%, influenced by a strong performance of the enterprise vertical and associated software attached products. Overall, our 2019 gross margin came in at 64.7%. We registered a record number of million-dollar customers in Q4 as a direct result of our enterprise momentum. By the end of 2019, we have acquired over 6,300 customers cumulatively, with Microsoft at 23% of total revenue and Facebook at 16.6%. In Q4 2019, and actually much throughout the year, cloud titans was the largest vertical.
The enterprise segment is now consistently the second-largest and strongest segment, followed by the financials in third place, Tier 2 cloud specialty providers in fourth, and service providers in fifth place. Both service providers and Tier 2 cloud providers have been slow for us. In terms of geography, 2019 international contribution was 24%, with the Americas at 76%.
In terms of new products in 2019, Arista Networks, as you know, delivered a banner year of disruptive products redefining networking with a highly differentiated software stack, management, and flagship platforms. The 7280 and 7500 Series especially have become the gold standard in 100 Gb spine networking. We also introduced substantial 400 Gb innovations with 10 new platforms. We launched a new portfolio of cognitive campus edge products for wired, Power over Ethernet switches, and wireless, including Wi-Fi 6.
Our inherent software flexibility brings federated management and control planes across multiple merchant silicon data planes, this is one of our key differentiators. We continued our systematic innovations in CloudEOS and CloudVision features. In particular, we have doubled our CloudVision customers, delivering real-time streaming telemetry, availability, scale, automation with change control, as well as third-party interoperability. We are pleased with the increased acceptance of total software bookings, both in subscription and perpetual licenses, now approaching 5% of total revenue annually.
We began Q1 2020 with the close of our third acquisition, Big Switch Networks, an SDN pioneer founded almost a decade ago. This is a strategic step for us in bringing a strong combination of engineering expertise, deeper entry into the network edge and broker market, and increased software multi-cloud visibility. As you know, we've always been focused on software-driven networking as a mission in Arista.
With Big Switch, we are expanding that mission through more deep analytics strengths, complementing our switch-based DANZ, or Data Analyzer platform, with Big Switch's deeper Monitoring Fabric across public, private, and hybrid clouds. Both companies also share a unique visionary status with data center networking in the Gartner Magic Quadrant.
We welcome the BCN team, BSM team into Arista family and really look forward to a strengthened partnership not only with them, but with Dell Technologies, as well as the approximately 300 customers we're getting to know better. While the Big Switch acquisition is not material, we do expect this to contribute to our software strengths and bookings. Our goal is to become accretive in calendar 2021. As we wrap 2019, I really do believe that Arista Networks is in the clear forefront of making cloud area networking more and more mainstream.
We see that networking and the future of it is not silos or boxes with multiple operating systems and spaghetti OSes, but a new uniform software-driven place in the cloud architecture. We are proud of our market leadership with the number one spot in 100 Gb E switching and ready for 400 Gb migration ahead of us.
We are experiencing trials that are going well with several customers. We are poised to achieve our first $100 million target for the first four quarters of campus revenue as well. Our total market trend continues to be an important investment area for us as we have doubled the sales and systems engineering capacity over the past two years.
With the promotion of Chris Schmidt to Senior Vice President of Worldwide Sales and Ashwin Kohli to Senior Vice President of Customer Engineering, reporting to Anshul Sadana, their transition to their new roles has been seamless following the departure of Manny Rivelo to a P/E firm in conjunction with his executive advisory role to us. Both Chris Schmidt and Ashwin Ashwin Kohli have long tenure with us and epitomize the Arista Networks ways.
As I've traveled the world, including hundreds of customers we touch, it is clear to me that we're winning strategic enterprise franchises across tech, media, banking, healthcare, insurance, and retail sectors, to name a few. Large enterprises are increasingly frustrated and anxiously seeking better quality and architectural leverage to cloud-based principles. Our programmability with software and quality with the lowest critical vulnerabilities in the networking industry is a refreshing and welcome change for them.
Arista is gaining strategic relevance, almost doubling our million-dollar customers within the past three years. I expect this momentum to continue and to happen across all the non-cloud verticals. With that, I'll turn it over to Ita Brennan for more financial specifics.
Thanks, Jayshree Ullal. Good afternoon. This analysis of our Q4 and full year 2019 results and our guidance for Q1 2020 is on non-GAAP, and excludes all non-cash stock-based compensation impacts, certain acquisition-related charges, and other non-recurring items. A full reconciliation of our selected GAAP to non-GAAP results is provided in our earnings release. Total revenues in Q4 were $562.5 million, down 7% year-over-year and above the midpoint of our guidance of $540 million-$560 million.
Service revenues represented approximately 19% of total revenue, up from 15.2% last quarter, reflecting typical fourth quarter service renewal seasonality in conjunction with a lower product revenue number. International revenues for the quarter came in at $137.7 million or 25% of total revenue, up from 19% in the third quarter. Looking to the year, international revenue accounted for 24% of total revenue, down from 28% in the prior year.
The shift in geographical mix for the year was largely driven by heavier U.S. deployments by our cloud titan customers. Overall gross margin in Q4 was 65.2%, above the upper end of our guidance range of 63%-65%, and up from 64.4% last quarter, which reflected a lighter cloud titan contribution in the period, combined with mix performance from our enterprise and financial verticals.
Operating expenses for the quarter were $154.3 million or 27.9% of revenue, down from last quarter of $163 million. R&D spending came in at $96.2 million or 17.4% of revenue, down from $105.3 million last quarter. This decline largely reflected lower engineering and prototype costs in the period. Sales and marketing expense was $46.4 million or 8.4% of revenue, with increased headcount somewhat offset by reductions in other sales costs.
Our G&A costs were consistent with last quarter at approximately $12 million or 2.1% of revenue. Our operating income this quarter was $205.8 million or 37.3% of revenue. Other income and expense for the quarter was a favorable $11.2 million, and our effective tax rate was approximately 15.5%. This lower tax rate reflected the release of some uncertain tax position-related reserves following final agreement with the relevant tax authorities.
Please note, however, that we do expect to see some upward pressure on the effective tax rate over time as various tax jurisdictions continue to modify their tax rules. This results in net income for the quarter of $183.4 million or 33.2% of revenue. Our diluted share number for the quarter was 80.26 million shares, resulting in diluted earnings per share number for the quarter of $2.29, up 1.8% from the prior year. Now turning to the balance sheet.
Cash, cash equivalents, and investments ended the quarter at approximately $2.7 billion. We repurchased $51.5 million of our common stock during the quarter at a weighted average price of $189 per share. This brings our total repurchases for the year to $260 million over three quarters. As a reminder, our board of directors has authorized a three-year, $1 billion stock repurchase program commencing in Q2 2019.
The program allows us to repurchase shares of our common stock opportunistically and is funded from operating cash flows. We generated $227 million of cash from operations in the fourth quarter, reflecting strong net income performance and a decrease in working capital requirements of approximately $115.4 million. DSO came in at 65 days, up from 63 days in Q3, reflecting the timing of billings in the period. Inventory turns were 2.9 times, down from 3.1 times in the last quarter.
Inventory increased to $244 million in the quarter, up from $240 million in the prior period. Our total deferred revenue balance was $575 million, up from $529 million in Q3. As a reminder, our deferred revenue balance is now almost exclusively services related, with any significant product deferred revenue balance having been recognized in the income statement in the first half of the year.
As Jayshree Ullal mentioned, we had two greater than 10% customers of the year. Microsoft is 23% and Facebook is 15.6%. If you exclude the recognition of product deferred revenue referenced above, Facebook would have represented approximately 12% of revenue for the year. Accounts payable days were 44 days, up from 31 days in Q3, reflecting timing of inventory receipts and payments. Capital expenditures for the quarter were $2.4 million. Turning to our outlook for the first quarter and beyond.
While we're not in a position at this point to provide full year guidance, we wanted to reiterate the various twists and takes discussed on our last call. 2019 has been a challenging year for our cloud business, with significant volatility and an overall muted demand picture. As we look forward to 2020, we believe this trend continues, with demand from this part of the business being flat to down on a year-over-year basis.
This trend, combined with a tough year-over-year revenue comparison due to the recognition of $118 million of product deferred revenue in the first half of 2019, will likely result in a meaningful decline in cloud revenue for 2020. Enterprise and financials are expected to grow healthily, but are not yet large enough to fully offset the expected revenue decline from cloud. Service Provider and specialty cloud will likely remain sluggish for the year.
On the gross margin front, we would reiterate our overall gross margin outlook of 63%-65%, with customer mix being the key driver. Focusing specifically on Q1, we expect to trend lower in this range, given a lighter revenue number and the typical first quarter weighting towards cloud. We continue to manage investments in the business carefully, with targeted growth in sales and R&D headcount, balancing the need to expand our market coverage with prudent financial management.
We announced the acquisition of Big Switch Networks earlier today. This represents an immaterial transaction which brings us some additional software capabilities and a strong engineering talent pool. From a financial perspective, this is a software subscription business with up-front licensed revenue recognition and a fair amount of services deferred revenue. We're beginning the business and accounting integration now, and the acquisition will be recorded in our financials for the first quarter.
We have included a small revenue contribution and two months of expenses for Big Switch in our guidance for the first quarter. We will provide additional clarity on the go forward income statement impacts once we've completed the purchase accounting analysis. Our guidance for Q1 does not reflect any impact from the ongoing coronavirus outbreak in China. While we do not have a significant direct manufacturing footprint in China, there may be some indirect supply chain impacts. We will look to monitor and attempt to mitigate these as the situation unfolds.
With all of this as a backdrop, our guidance for the first quarter, which includes the non-GAAP results and excluding any non-cash stock-based compensation impacts and other non-recurring items, is as follows. Revenues of approximately $522 million-$532 million. Gross margins of approximately 63%. Operating margins of approximately 24%. Our tax rate is expected to be approximately 21%, with diluted shares of approximately 80.5 million shares. I will now turn the call back to Charles Yager. Charles?
Thank you, Ita Brennan. We are now going to move to the Q&A portion of the Arista earnings call. Due to time constraints, I'd like to request that everyone please limit themselves to a single question. Thank you for your understanding.
We will now begin the Q&A portion of the Arista earnings call. In order to ask a question during this time, simply press star, then the number one on your telephone keypad. If you'd like to withdraw your question, press the pound key. We ask that you pick up your handset before asking questions in order to ensure optimal sound quality. Your first question comes from Pierre Ferragu with New Street Research. Your line is open.
Hi. Thanks for accepting my question. I was wondering on the day two days for your cloud pricing trends. We understand like flat to slightly down year-on-year, which means that we'll have a bit of a recovery during the year. I think it's surprising that it's contrasting a lot with what we've heard from other suppliers on the computer side, on the memory side.
We've seen much better numbers in the first half of last year, and much more positive comments about the first half of this year. You don't seem to be on a similar cycle. Things seem to be coming back slower for you guys. Any way you could help us understand that would be very helpful.
Sure, Pierre Ferragu. I will take part of the question, and Anshul Sadana can elaborate. The cloud CapEx spending has never correlated one to one with Arista's network numbers. The reason is many, because as you know, it's a very small part of their total CapEx. It's almost negligible, the network piece. Often they make the compute and the storage decisions first, and the network comes later. There's a lag. Our visibility right now, as I've often told you, is one to two quarters, and we can't say much about the whole year, let alone Q2 or second half.
While we are making predictions based on our 2019, both the absorption of deferred revenue and the current understanding of their plans, things might change. We're reflecting what we know best as of now, especially when you look at our year-over-year comps for 2018 and 2019. Anshul Sadana said I did good enough. She now has nothing more to add.
Thanks, Jayshree Ullal.
You're welcome.
Your next question comes from Meta Marshall with Morgan Stanley. Your line is open.
Great. Thanks. I just wanted to know if you could kind of give a little bit of an update on the Campus business traction, and just whether you had seen any kind of pause in purchasing activity due to macro. Thanks.
Well, thank you, Meta Marshall, and thank you. I think it's the first call for you with Arista Networks, so welcome. As you know, our Campus business is in its very early stages. We only started shipping in Q3 2019. We've had two full quarters now of Campus, and I'm very encouraged by the enthusiastic response from both our existing customers as well as new customer acquisition.
As I look at our million-dollar customers that I was talking about, and the record number we have achieved, many of them have also been Campus. So I see that, at least from our little oasis or island, we're not really a macro bellwether. We're more of a technology bellwether, and the technology on both on our PoE and Wi-Fi has been very well accepted, along with our clients and CloudVision, and Q4 was a very good indication of that.
Great. Thanks.
Your next question comes from Erik Suppiger with JMP Securities. Your line is open.
Yeah. Thanks for taking the question. One, what are the priorities in terms of your integration with Big Switch? What are you going to be doing? Which products are you going to be focusing on? Secondly, is there anything that we should be concerned about with China? Did you see any evidence of impact from the virus issues going on there?
I'll take the first question. Anshul Sadana, maybe you, as you know with your COO title, you know more of what's going on there. On the priorities for the Big Switch. Well, as I said, first of all, we're incredibly impressed by the engineering team. We have selected the best of the best, and they're very much part of the acquisition has closed, and I want to give a big shout-out to Marc Taxay and Ita Brennan for the hard work that went into that. It's a 10-year-old company. There's a lot of detail.
At the end of the close of the acquisition, we have absorbed approximately 75 employees, a large percentage of them in engineering. There are two product lines that Big Switch carries, the BMS, which is the monitoring product, which as I explained, is a perfect complement to DANZ.
DANZ is Arista's in-line integrated switching product. DMF is just icing on the cake. It has deeper visibility, service nodes, recorder nodes, the monitoring fabric, et cetera. BCF, which is meant for more of a converged fabric, and we fully expect that to be a very important piece, especially with our converged infrastructure partners like Dell Technologies. Both products will be fully supported and carried through, but with different channels.
To your question on China and supply chain with respect to coronavirus, as Meta Marshall mentioned, it's really comes down to secondary effects of our manufacturer suppliers or their sub-suppliers and their raw materials coming from China. So far, we haven't seen any big impact because it's just been a few weeks, really a few days after the end of the Chinese New Year.
Our manufacturers are saying they are okay for the short term. However, if the situation continues for a long period, and there's really a supply issue of raw materials and over an extended time, then it can have impact in the future. It's too early to say anything right now, and we expect more answers from the Chinese suppliers, the factories, through the rest of the month and even in March. Again, too early, but for now, we're okay.
Very good. Thank you.
Thank you, Anshul.
Thank you.
Your next question comes from Rod Hall with Goldman Sachs. Your line is open.
Hi, guys. Thanks for the question. Girls. Girls and guys is probably the right order for that, actually. I wanted to ask a couple of verticals questions. Meta Marshall, you had commented that cloud, you expect a meaningful decline, and I don't know if you could give us any more feeling for what you mean by that. We were thinking a kind of low double digits sort of decline this year, but I don't know if that sounds like more than that.
Then the other, on the positive side of this equation, the service provider commentary, you said sluggish. We were thinking that'd be declining this year, but I just want to clarify, are you expecting a little bit of growth there or stabilization there, kind of flattish revenue there?
A bigger picture question, Jayshree Ullal, is have you guys considered disclosing these verticals or some subset of them? Is that something we could maybe look forward to at some point in the future, at least on an annual basis, so we can all track kind of what's happening under the covers in the verticals?
Okay. Rod Hall, I'll try and answer as many of your questions. The last one first. The answer is we try to do our best by ranking them, and I guess you're looking for more granularity, so we'll definitely take it under advisement, like we do anything you suggest. Maybe that's a data point for Analyst Day.
Going back to your cloud decline and service provider, half the down I think definitely means half the double digit down. We don't exactly know how much now because we're in the first quarter. I think Anshul Sadana and the team will get much more visibility in maybe Analyst Day or second half, because that's when we really get a good sense. Q1 is such a seasonal quarter, we don't have enough data. It's more an extension of Q4.
Ask us this question again in May or June, and I think we'd have a lot more to say. Service providers, yeah, I think when Ita Brennan actually explained the definition of sluggish to me, which is it's not that bad, and I think the worst was last year. It's bottomed out and it can only get slightly better is our feeling. We're not holding our breath or anything, but we believe it can't get worse.
When it's sluggish, would it be some range around zero then, Jayshree Ullal?
0+ or minus. Yeah.
Negative numbers?
Not minus.
Not declining further. That's what I think.
Above zero.
Our hope for right. I mean, it was declining a fair amount in 2019, and we'd like to think that it starts to recover a little bit from there, but sluggish kind of implies slowly.
Okay. It sounds better than I thought, so appreciate that.
Yeah.
Okay. Thank you.
Your next question comes from John Marchetti with Stifel. Your line is open.
Thanks very much. Jayshree Ullal , as we're getting closer to this, reaching the end of this sort of first $100 million on the campus side, and you've added a couple of pieces there, I'm just curious as we look out the next couple of years, how maybe we should think about that business trending up towards a little bit more of a meaningful contribution to the overall growth rate.
Ita Brennan, just real quick on the tax rate, I just wanted to make sure that I heard it correctly, that at least starting in 2020, we should expect these lower rates to continue and that it may tick back up over time, but that we're at least expecting that lower tax rate to continue over the near term. Thanks.
Yeah, let me take the tax one first, maybe. The Q4 tax rate of 15%, that's a one-off, right? It's some very specific reserves that were released in the period. As we go forward, we've been thinking about the structural tax rate somewhere around 21%. That's what we guided for Q1. There's probably more upward pressure than downward pressure on that over time.
I think, 21%- 21.5%, it's in that range somewhere. I think for now, you're in the 21% range for Q1, and then maybe you see a couple of tens of basis points kind of increase over time, but it'll take time. That was more a kind of a longer-term structural rate statement. The 21%+ or minus hasn't really changed in the current quarter.
John Marchetti, to take your question on the campus, I think being a new kid on the block when there's been a fairly mature market and a $10 billion TAM, largely Cisco and maybe HPE Aruba. Just for us to enter in and be taken seriously is a good sort of business. I have to say, the enthusiasm for Arista, the software, the cognitive CloudVision has been very well received.
I think delivering the first four quarters of $100 million will really establish a baseline for more growth. As I said at the last analyst meeting, I'd be very disappointed if we didn't have that more growth translate into doubling and doubling again. My team would be looking to take the $100 million to $200 million and then to $400 million and then some. More on Analyst Day maybe on exactly how we do that in the details.
Great. Callers, as a reminder, please keep one question per call on the floor. Thank you very much. Let's go to the next caller.
Your next question comes from Jason Ader with William Blair. Your line is open.
Thank you. Jayshree Ullal, you guys have prided yourself from day one on a single OS. Now with Big Switch, you're going to have a second OS, so I'm just wondering how we should be thinking about the future of Big Switch's OS.
Yeah, no, that's a really good question. We've actually done three acquisitions and had to deal with the OS types before. Jason, before I answer your Big Switch question, let's take the other two. Metamako, very FPGA centric. It was really for the high-frequency trading market, low latency. The OS did not play a big role, and when it does, it'll be EOS, right?
Mojo, really a radio management Wi-Fi. We immediately integrated into CloudVision, and again, the OS was less important than the CloudVision integration to bring wired and wireless together. Big Switch, we fully expect that same CloudVision integration with DANZ and DMF, the monitoring fabric. The OS will be somewhat transparent and the unification of inline DANZ and a monitoring fabric will be much more important.
The Big Cloud Fabric is the unique product, we believe the go-to-market channel there is not necessarily a typical Arista customer who's looking for EOS and CloudVision, but a technology partner like Dell Technologies or Nutanix that wants to integrate this with their servers and storage. Very much like SONiC or F5, this is a converged infrastructure solution.
Is it right to think about that serving a part of the market that you probably never would have served and therefore you can reconcile it with the overall strategy?
That's correct. Again, it's early days and we're still learning. Both Doug Murray and Kyle, the co-founder, are teaching us as we speak. If you look at their over 300 customers, the overlap with our customers is only one-third. 60%, 65% of the customers are new to us.
Great. Thank you, Jason Ader.
Your next question comes from Brian Young with Deutsche Bank. Your line is open.
Hi. Thanks for taking the question. I wanted to ask about your 400G opportunity. Can you talk about expected market share for 400G, maybe just at a high level, especially versus your peers? I think it's fair to say that you are dominant in 100G at hyperscale clouds. Is the expectation to win the majority of 400G deals, or are you taking a more conservative view where your peers might see sizable wins as well? Thank you.
Well, I think our peers should have seen sizable wins on 100G too. I don't know why they missed them both. It's very unnatural for an incumbent to lose a speed transition. The fact that Arista became number one has just never been done in the networking industry before. Now, what we see the 400G overall timeline is we have been consistent on that. We are seeing early trials. We expect mainstream deployments this year, particularly in the second half.
Obviously the cloud will play a big role, but it'll also be some of the high-end enterprises and service providers as well. Given the trials we're seeing, we don't expect material revenue of 400G until the second half of this year and really 2021. The other thing I'd point out is nobody just builds 400G in isolation. It always happens in conjunction, and they're always mixing and matching 100G and 400G. Anshul Sadana, you want to add to that?
Yes, we do comment here. Brian Young, I know everyone's focused on 400G. I do want to remind you, and everyone else, 100G is actually growing in 2020, not being replaced just yet. There are many other equations coming soon as well, whether it's eight by 50 going to four by 100 or eight by 100 as well. It's not just a single transition that our customers are planning or working with us on. There are multiple speed transitional roadmap items being discussed.
Lastly, there is some dependency on the DCI or backhaul networks on availability of DR optics at scale, and that's still expected to be second half, maybe even end of the year. Hence any material requirements in that space will be in 2021.
Anshul Sadana, just to reiterate your point, the TAM for 100G is $4.5 billion. I don't know the TAM for 400G will even hit a few $100 million this year.
Yes. TAM is not a time.
Yeah. TAM is not a time. Okay. Good one-liner.
Thanks, Brian Young.
Your next question comes from Ittai Kidron with Oppenheimer. Your line is open.
Hi, Ittai Kidron.
Yes. Hi, ladies. Two questions from me. I know Charles Yager tries to limit me to one. I'll try to sneak one in.
Hi, Ittai Kidron.
Yes. Regarding the acquisition of Big Switch, I do want to understand the relationship with Dell, and also if there's a revenue concentration for Big Switch. If I remember correctly, Microsoft is a big customer for them, if you could discuss that. On Dell, they also have other companies they work with in this space, Cumulus, if I remember correctly. Help me understand how you think the nature of that relationship is going to look like going forward.
That's a good question. First of all, we didn't see any significant revenue concentration and certainly not Microsoft. That could have been a past statistic, but not true now. There's obviously some big customers, and they have some top 10, but not specifically one that is a 10% concentration. Coming back to your question on how do we see this, we see the partnership with Dell getting stronger.
How Big Switch was selling was really software only disaggregated from hardware, and a lot of the hardware was Dell switches. We continue to plan to offer that model and strengthen our partnership with Dell and make that stronger. We are not changing the sales motion, and it complements what Arista's doing in the high-end enterprise and cloud titans very well. No change.
Very good.
Thank you, Ittai Kidron.
Thank you.
Your next question comes from Amit Daryanani with Evercore ISI. Your line is open.
Hi. This is Irvin Liu, dialing in for Amit Daryanani . I also had a question about your campus switching business, which continues to do very well for you, but can you perhaps help us understand what the margin profile for this business looks like versus your router portfolio?
Oh, okay. I would generally say the margin is about the same, but if you said versus our router portfolio as opposed to versus our switching portfolio.
The broader portfolio. Sorry.
Okay. The broader. Okay.
Yes, broader.
Yeah. It's the same. More pricing pressure in the campus always because the market is defined that way, but the margins aren't significantly different.
Thank you.
Thanks, Irvin Liu.
Your next question comes from Ben Bollin with Cleveland Research. Your line is open.
Good afternoon. Thank you for taking the question. I wanted to go back to 400G. I was hoping you could talk a little bit about how you think the builds for 400G differ from what you saw in 100G. Specifically, I'm interested in any thoughts you have on how your partners approach their own OS development efforts, any switch standardization efforts. Just any high-level thoughts there. Then how do you think the margin opportunity for 400G, how does that compare to what you saw with 100G?
Sure. Ben Bollin, in terms of the 400G architecture and how our largest customers are leading the way there, the entire architecture needs to contemplate a server flow that's 400G being able to go through the network. You have to operate it end to end. You cannot just operate in silos and be done with it.
For large-scale architectures, upgrading the backbone and the DCI networks is almost a necessary step one before you can do 400G in the broader leaf-spine designs. That's dependent on DR, other optics, and so on, but that testing is going on already in the lab trials that we are involved in today. For AI, which are closed clusters, 400G is already starting to see a little bit of deployment, but these are still very small scale, simply a 32 by 400 type of design in a mini leaf-spine.
There, we're working with our customers very well with co-development. It may not be just the OS. That problem I think we've already addressed with our best customers and partners. Actually co-development with the NIC and the FPGA and the GPU and so on, that is also happening already.
Your next question comes from Sami Badri with Credit Suisse. Your line is open.
Hi. Thank you. I had a question for you regarding Big Switch and just the operating margin drag that the Big Switch acquisition is going to create in 2020, and maybe I was hoping for some specifics on 2021 accretion that you mentioned. Is this a one Q2Q 2021 accretion, or is this a back half of 2021 accretion that you are anticipating?
I'll answer the second question, and then, Ita Brennan, if you could I wouldn't call it too much of a drag, but maybe. In terms of when it'll be accretive, the forecast we have challenged the team with is to be accretive by the end of 2021. Obviously, I'd like to see it sooner.
Yeah. I think, once we get out of Q1 was always going to be a quarter that was hyped from an operating margin perspective, just because where the revenue came in at and the mix of business that we have, with cloud being heavier. You are seeing a lower operating margin in Q1. I think once we get out of Q1, we had talked on the last call about a 35% operating margin being the target, and I think we can absorb Big Switch within that. Sorry. As we move through the year, we'll have more to say on the top line on the next call once we sort out some of the purchase accounting and other stuff that we need to work through.
Got it. Just actually a clarification, is the $100 million run rate for campus, does that include any Big Switch contribution, or is that just the stuff pre-Big Switch?
No. Big Switch was not there in December 2019, right? I was talking about Q3 and Q4, we're well on our way to $100 million. I just want to clarify, it's not run rate, it's revenue. It's real revenue.
Thanks, Jayshree Ullal.
Your next question comes from Tejas Venkatesh from UBS. Your line is open.
Thank you. It looks like Microsoft was only down 5% in 2019, a bit better than what you were expecting. Given that in the past, you provided early color on what Microsoft could be as a percentage of revenue, I was hoping, Jayshree Ullal, you could do that for 2020.
Tejas Venkatesh, I'd be a fool to try. Nice try, Tejas, I'd be a fool to do that right now, wouldn't you say, after the surprises we had last year? I'm teasing you.
Well, this one went better. The second thing I wanted to ask about was, is your cloud now less than 10% of revenue, and is the visibility sort of improving given that it was such a drag in 2019? Thank you.
That's a good question, Tejas Venkatesh. None of our five verticals that I report are less than 10% revenue.
Great.
Your next question comes from Aaron Rakers with Wells Fargo. Your line is open.
Thanks for taking the question. I kind of want to build on that last question. As you look forward, I think last quarter you alluded to that one of your large cloud titan customers were actually just flat out pausing with regard to their spending dynamic as it relates to maybe a server cycle of variable consideration. As you think about your outlook today, how would you characterize if whether or not that's changed at all? Has the slowdown become more pervasive across your cloud titan customers? Just any kind of update on how you roll up that cloud titan forecast this year relative to what you thought kind of the last quarter.
Yeah. That's a good question, Aaron Rakers. First of all, just to reiterate, all the surprises we received from that specific cloud titan remain and are being factored into the 2020 forecast. Those numbers will be lower this year, right? As for the other cloud titans, each one's unique, so it's not necessarily feeding into the others. That is specific to that one. Each one has their architecture, their timelines, their migrations, their specific CapEx plans. I wouldn't roll one into the other.
Okay. Thank you.
Your next question comes from Paul Silverstein with Cowen and Company. Your line is open.
Great. Jayshree Ullal, I've got a similar question, but from a different angle with respect to cloud. During the actual quarters, Jayshree Ullal spoke about a traumatic pause, and I think you also mentioned the fact that it wasn't just a question of when they would return and to what magnitude when they did. The question I now have for you is, I assume your definition of your visibility for Microsoft and Cisco is not what it was at its peak, far from it.
What visibility, how would you characterize that visibility today looking downstream, not just the next quarter or two, but further out? Do you have any visibility as to what those customers will look like a year from now? It's obviously about a quarterly period, but obviously speaking from the annual standpoint, what the contributions of it.
Hi, Paul Silverstein. If you go back consistently to our last three years, I think we've always said we don't have more than a one to two quarter visibility on any of our cloud cycles, and that hasn't changed. I wouldn't be able to give you annual visibility of what the spend this year versus next year would be. I do think we get greater visibility in the second half of this year on how the next year will look like, but in terms of broad trends. That answer hasn't changed despite the pause and puts and takes.
All right.
If I can add to that, we don't have visibility into their surprises. No, fair enough. On the enterprise, just very quickly. Jayshree Ullal, you kept referencing the million-dollar-plus deals that almost doubled over the past three years. Can you tell us how many million-dollar deals you have in enterprise?
Hundreds. Many hundreds.
All right.
No, it can't be many hundreds, because you'd be well beyond your $100 million target.
I'm talking about campus specifically.
Oh, you're only talking about campus? Oh, I see. I don't have that answer, but I was talking about overall enterprise, including campus.
Okay. You don't have the answer.
Yeah.
Thank you.
Your next question comes from Jeffrey Kvaal with Nomura Instinet. Your line is open.
Yes, thanks very much. We have been hearing from some of the larger OEMs that the server chipset availability is a little tighter than they might have hoped. I wouldn't have necessarily given verification for Arista Networks a year ago. Given sort of the tenor of the conversation we had last quarter, I'm wondering if that is something that we should be monitoring just in case some of your other web scale customers can't get all the servers they want, or some of the Tier 2 players can't get the servers they want, and that may then impinge upon your switch sales to them.
No, Jeffrey Kvaal, it doesn't really impact us as much on a short-term variation over there. Especially the cloud companies, they decouple supply chain planning for these issues and have one or two months of gaps anyway. If there's a fluctuation of one or two months, it doesn't really impact us.
Okay, thanks, Anshul Sadana. As a clarification, I think in the past you've talked about the likelihood of coming back to year-over-year revenue growth in the fourth quarter of 2020. Is that still sort of a reasonable place for us to stick a yard mark?
Yep, Jeffrey Kvaal, that's our hope.
Okay. Thank you.
Thank you. Your next question comes from Alex Kurtz at Piper Jaffray Capital Markets. Your line is open.
Thanks. I just want to clarify the comment about the sluggishness for cloud service provider and service provider. That sluggish comment, Jayshree Ullal, and you, that goes for both verticals?
Yes. Combined.
Okay. Combined. Okay, thank you. Jayshree Ullal, last quarter, you outlined changes in how you thought the cloud service provider segment was considering on-prem versus on-prem infrastructure spend, and maybe some of them moving back to cloud. Can you give us an update on how you see that vertical? I know you outlined it here in the growth projection, but that was a change from prior view. I was just wondering if there's-
That's correct.
update on that segment specifically.
Nothing's changed significantly. Some of the Tier 2 cloud providers have resumed some small spends, and some of them are still evaluating.
All right. Okay, no longer, you don't see it as a near-term secular opportunity right now from what you can see?
No, not in the first half. I think there's not a drag. I mean, not a drag, but not an amazing upside either.
Okay. All right.
Your next question comes from Tal Liani with Bank of America. Your line is open.
Hi, guys. Just two questions. First, if I remove Amazon, Microsoft, which I have perfect numbers for.
Which one? I'm confused. Stop.
Eventually I'm going to get to the right answer, yeah. If I remove Microsoft, which I have perfect numbers for both years, and then I assume Facebook at 9% last year, because that's the highest number below 10%, versus with this year, the growth is only 6%. If I assume for Facebook 5%, which again, it wasn't a 10% customer, so it has to be below that.
If I assume anything below, the growth is even lower than 6%. The question is, why aren't you growing faster with everyone else? Forget Facebook and Microsoft. Why don't we see faster growth like we used to do before? We've always been doing this exercise without Microsoft, and the growth was always very strong. The second question, not related.
Hold on. Can you hold? Tal Liani.
Hold on.
Tal Liani, we're still processing your first question.
Okay.
We're still trying to understand how you computed this. Did you compute deferred revenue in your question?
No, I did not.
The issue going from 2019 to 2020 is that you have this deferred revenue, $118 million, which is like 5% of revenue approximately, right? Close. Right? You have to backfill effectively in 2020, right? That kind of puts a drag on the growth rate before you start. Right? You have to adjust for that. Even with that, what we're saying is that cloud vertical will be kind of flat to down from a demand perspective, right?
By the way, is the deferred related to Microsoft or is the deferred related to everyone else?
If you trace back in time, it was non-Microsoft, obviously it's a large number. I mean, you can figure out that it is with the Facebook impact, right? We talked about that as 16.6% was deferred, probably 12% of revenue was as deferred, right.
Okay.
Your next question comes from Simon Leopold. Sorry.
No, it's fine. We'll come back to you if we have time.
Yeah. Okay. Perfect.
Your next question comes from Simon Leopold with Raymond James. Your line is open.
Great. Thank you. I got confused, not only from Paul's question there. Wanted to ask you about the commentary several of the hyperscale providers made about extending the useful lives of their servers. Wondering how that translates into your business for intra-data center switching, and if this was an aspect that you were aware of when you had provided your forecast last quarter or whether the commentary we heard during this earning season was also new to you. Thank you.
No, Simon Leopold, when we did our earnings call last quarter, we very much stated what we have seen from at least one of the cloud titans, where they were delaying their refresh, and now you're seeing the market that some of the others are participating within this as well. We're not seeing this with the other Arista customers so far, even in the cloud space.
For us, it's limited to one customer. We don't deal with the other large cloud company Arista do as much, so we're not as exposed there. It's not a market wide trend. It's very specific to their architecture, the next generation they're on, and the sort of type of offload they're looking for to decide which CPU generation to select.
Is there a way or some math to figure out how to translate if, say, they extend the life by one year, so instead of three-year replacements, they get a four-year? Is there some arithmetic or rule of thumb to help us think about how to quantify the impact for you?
We don't have it. I'm quite sure someone on this call has a model around that. No, that's not something we try to forecast.
Okay.
Mostly a delay of a year in the spend.
All right. Thank you for taking the question.
Thank you, Simon Leopold.
Your next question comes from Jim Suva with Citi Investment Research. Your line is open.
Thanks, everyone. I sincerely just have one question because I'm just a very simple guy and not as smart as others. Whether it be Jayshree Ullal or Ita Brennan or Charles Yager, a quarter or two ago, you talked about applied cloud titans skipping a refresh cycle or elongating their purchasing. Some of the commentary after that was, well, maybe they're using white box or maybe they found a better compute standard or a way to fit more through compression or duplexing or some other standard.
Now that we've had several months behind us, can you give us any visibility of, do you feel more confident that it truly is just a delay? Are they looking at other solutions? Or just kind of revisit that topic. Thank you.
Yeah. Jim Suva, I think all the theories of those answers are not true. I think the customer's been pretty straightforward with us that they have always been using Arista Networks as well as some internal development, and we've been working with them on the internal development. We feel very comfortable that their forecast has changed, and they continue to be an important partner with us.
Okay. Thank you so much. It's appreciated.
Thank you, Jim Suva.
Thank you.
Your next question comes from Samik Chatterjee with JP Morgan. Your line is open.
Hi. Thanks. This is my first question. I just want to, given your commentary on the slower spending you're seeing from your customers, you obviously have a Q1 seasonality that's weaker than what we've seen historically. As we look through the rest of the year, is it fair to assume, given the visibility you have right now, that the seasonality through the remainder of the year will be weaker than what we saw in normal years like 2018, for example? Are you still just comfortable reiterating the full year earnings guide that you gave last time, which was mid-single digit line revenues?
Okay. Samik Chatterjee, I'm a little confused by your question. Are you talking about cloud customers or customers at large?
Cloud customers.
Cloud customers. Okay. Now that makes sense. Yeah. Look, we feel the same way as we did last time, which is, we had a lot of weakness in the spending, and it's reflected in our Q1. I think we will know better about the rest of the year when we get to the second half more. As it stands, nothing's changed. The CapEx that they will spend, Arista Networks feels in a very strong position to compete, differentiate, and get the business. We're not losing market share. We're winning the sockets. The rate of spend and adoption, we do believe, will be a flatter down year this year.
Okay. Thank you.
Your next question comes from James Fish with Piper Sandler. Your line is open.
Hey, team. Happy almost Valentine's Day.
Hey, James Fish.
Can you guys go over the linearity of the mix of customers over the course of the year, and if it was back half loaded at all? Just wanted to be clear, is Cloud Titan going to be flat to? Flat to down double digits on top of the $118 million deferred headwind, is the flat to down double digits inclusive of the headwind?
The meaningful revenue decline and the double digit are all referencing revenue, right. Saying basically the revenue numbers will decline meaningfully, right. That includes the $118 million, right. When we talked about demand and we said that we expect it to be flat to down, that's not referencing that double-digit number. That's a commentary that we will play out as we see where it goes, but it's not trying to say that it's a double-digit number.
To answer your question on Q4 linearity, I'm just looking at the chart my boss is giving me. It was pretty linear across the three months. We had a record $1 million of customers in Q4 and good spend across our top 50 customers, nothing unusual except more campus.
Okay, thanks.
Your next question comes from Tim Long with Barclays. Your line is open.
Hi, Tim Long. Tim, are you there?
Tim Long?
Let's go to the next one.
Tim, call back if
Your next question comes from Ryan Koontz with Goldman Sachs Securities. Your line is open.
Hi, thanks. Quick question. I was wondering if you could speak to your outlook for international. Hyperscale is kind of a big piece of that shipping to their international destinations, but if you could speak to kind of any updates in strategy or channel development there that's going to supercharge that business for you? Thanks.
Anshul Sadana and the team have invested pretty significantly in international, especially, of course, the developed countries in Europe and Asia Pac. We had a strong quarter. We're starting to see some important customer wins in the enterprise and even some small service providers. We obviously are much more channel-led in international. That's always been the case. That's been a strong area of experience for us. I, Anshul Sadana, you want to add to that? I think that's a strong area of growth.
No, actually, both on headcount growth, we actually have a significant number of heads added to international. On the channel side, the channel development plan we have is a separate one for U.S., so there's a separate one for EMEA and APAC as well, with a different ecosystem, different set of partners and so on, and that's coming along reasonably well. As Jayshree Ullal mentioned, the international locations are mostly for physical channels, but now we're working with them to make it channel-led as well. If we remove cloud and so on, the rest of the organic international business is doing well.
One other, just one quick add to what Anshul Sadana said. Generally, our customer logos are higher internationally than in the U.S., though a purchase of sale. Out of the 6,300 cumulative customers, half are international.
Great. Thank you.
Your next question comes from Tim Long with Barclays. Your line is open.
Okay. Sorry about that.
Hey, Tim Long.
Hi, how you doing? I just wanted to ask kind of two related questions. First, could you talk a little bit, there's been a lot of chatter about silicon diversification in switching areas. Can you give us your views, obviously, with the big customer making an announcement and starting to talk about some traction there. What do you think the impact is there? Related, what are you seeing just overall on white box these days? Are you seeing any change to who's using it or how they're using it? Thank you.
Okay. Well, as you know, Arista Networks has always been a big proponent of merchant silicon and ASICs, notwithstanding this new announcement, have been around 30 years. ASICs by themselves are not new. We do see three dimensions. First of all is from a best-of-breed system standpoint, we couldn't be more pleased with the silicon we receive from Broadcom, both on the Trident Tomahawk side and Jericho.
We have very high confidence. It's a case where you can't just build one point product, you have to have a full roadmap, and we've always been ahead of vendor-specific ASICs, and we believe that'll continue. In terms of software, silicon by itself is not so interesting if you can't build a system. Obviously EOS, we feel is the most competitive software differentiated across many merchant silicon devices.
I think we supported it over 18 silicon families, maybe more than that. In general, we feel like that leaves only one other thing, which is Cisco selling chips. Obviously, just like others, we're going to be competing with Intel and Broadcom on that one.
Your last question comes from Tal Liani with Bank of America. Your line is open.
Tal Liani , you have one last question.
Thanks very much. By the way, I just want to comment something. We'll take it offline, but we shouldn't remove the curve, and I can explain later why. The number is the number. You can remove both sides. In any case, I wanted to ask about the gross margin. Why is it declining sequentially next quarter? I don't think anyone asked this question.
No, they didn't. It's because our customer mix, right? We were heavy Enterprise and Financials in Q4, we're a heavier cloud mix in Q1. Q1 always has a heavier cloud mix because Enterprise just takes time to ramp into those quarters.
Got it. Thank you. Great.
All right.
Okay. Well, this concludes the Arista Networks Q4 2019 earnings call. Thank you for joining us today. Please note that moving forward, our earnings calls will move to Tuesdays, starting with the Q1 2020 call, which will take place on Tuesday, May 5th. Lastly, we have posted a presentation which provides additional information on our fiscal results, which you can access on our investor section of our website.
Thank you for joining, ladies and gentlemen. That concludes today's call. You may now disconnect.