Welcome to the third 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.
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 third quarter ended September 30, 2019. 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 fourth quarter of the 2019 fiscal year, longer-term financial outlooks, industry innovation, our market opportunity, the benefits of recent acquisitions, and the impact of litigations, 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. You should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. Please note 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 will turn the call over to Jayshree.
Thank you, Charles. Thank you, everyone, for joining us this afternoon for our third quarter 2019 earnings call. Our profitability growth combination was once again demonstrated with a non-GAAP revenue of $654.4 million, with a non-GAAP earnings per share that grew to a record $2.69. Services contributed approximately 15% of revenue. We delivered non-GAAP gross margins of 64.4% influenced by a solid performance from our Cloud Titan and enterprise verticals. In terms of customer trends, we registered a record number of new customers in Q3 and continue to drive this new customer logo expansion at the rate of one to two per day throughout the quarter. For calendar 2019, we do expect to have two customers that will be greater than 10% of our revenue, Microsoft and Facebook.
In Q3, the Cloud Titan vertical segment remained our largest one. The modern enterprise segment is now consistently becoming our second largest, with financials in third place and service provider and tier 2 specialty cloud providers coming in at fourth and fifth place. In terms of geography in Q3, the international contribution was 19%, with the Americas at 81%. In terms of new products, we introduced important enhancements to our CloudVision platform, dubbed CloudVision 2019. Arista's CloudVision is bringing cloud principles to network operators across Places in the Cloud, or PICs as we call it. The largest cloud providers in the world have driven advancements in telemetry and automated network operations that improve many of the same network operations tasks for the enterprise. CloudVision ups the ante to deliver these analytic and telemetric capabilities to organizations in the enterprise of many sizes.
Key highlights of CloudVision 2019 include dynamic scale, elastic agility, deep visibility, and open integration where we can derive visibility metrics from SDK and SNMP-capable platforms, including managing third-party devices to bring multi-vendor capabilities across the entire enterprise. I would like to offer some further color on Q4 2019 guidance given our significant drop. After we experienced the pause of a specific Cloud Titan's orders in Q2 2019, we were expecting a recovery in second half 2019 for Cloud Titan spend. In fact, Q3 2019 is good evidence of that. However, we were recently informed of a shift in procurement strategy with a material reduction in demand from a second Cloud Titan, reducing their forecast dramatically from original projections for both Q4 2019 and for calendar 2020. Naturally, this type of volatility brings a sudden and severe impact to our Q4 guidance.
Given the tepid forecast and volatility of this cloud segment, we believe the Cloud Titan forecast should be modeled as flat to down in calendar 2020. I do want to take an opportunity to reiterate that our market share for both 100G and overall high-performance switching remains solid and strong. We are proud of our strength in the enterprise and financial segment with growing success in our very first quarter of shipping Cognitive Campus portfolio products, which is now on track for $100 million in the first full year of shipments. With that, I'd like to turn it over to Ita for more specific financial metrics.
Thanks, Jayshree. Good afternoon. This analysis of our Q3 results and our guidance for Q4 2019 is based 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 Q3 were $654.4 million, up 16% year-over-year and above the midpoint of our guidance of $647 million-$657 million. Service revenues remained strong, representing approximately 15.2% of revenue, down from 15.6% last quarter, reflecting typical seasonality of service renewals. International revenues for the quarter came in at $122.1 million, or 19% of total revenue, down from 27% in the prior period. This volatility in geographical mix was largely driven by a shift towards U.S. deployments in our Cloud Titan business. Overall gross margin in Q3 was 64.4%, above the midpoint of our guidance of 63%-65%, and down slightly from 64.7% last quarter. This reflected a healthy Cloud Titan contribution, combined with good performance from our enterprise and financial verticals. Operating expenses for the quarter were $163 million, or 24.9% of revenue, up slightly from last quarter at $158.7 million.
R&D spending came in at $105.3 million, or 16.1% of revenue, up from $101.7 million last quarter. This reflected headcount growth and slightly higher levels of product-related NRE and prototype spending in the period. Sales and marketing expense was $46.8 million, or 7.1% of revenue, up from last quarter with increased headcount somewhat offset by reductions in other sales costs. Our G&A costs were consistent with last quarter at approximately $11 million, or 1.7% of revenue. Our operating income for the quarter was $258.2 million, or 39.4% of revenue. Other income and expense for the quarter was a favorable $14.9 million, and our effective tax rate was approximately 20.5%. This resulted in net income for the quarter of $217.1 million, or 33.2% of revenue.
Our diluted share number for the quarter was 80.75 million shares, resulting in a diluted earnings per share number for the quarter of $2.69, up 27.5% from the prior year. Now turning to the balance sheet. Cash, cash equivalents, and investments entered the quarter at approximately $2.4 billion. We repurchased $115 million of our common stock during the quarter at a weighted average price of $224 per share. As a reminder, our board of directors has authorized a three-year, $1 billion stock repurchase program commencing in Q2 2019. This program allows us to repurchase shares of our common stock opportunistically and will be funded with operating cash flows. We generated $269 million of cash from operations in the third quarter, reflecting strong net income performance and a decrease in working capital requirements of approximately $25 million.
DSOs came in at 63 days, up from 51 days in Q2, reflecting the timing of billings in the period. Inventory turns were 3.1 times, up from 2.4 last quarter. Inventory decreased to $239.8 million in the quarter, down from $314.2 million in the prior period. Our total deferred revenue balance was $529 million, up from $502.2 million in Q2. As a reminder, our deferred revenue balance is now almost exclusively services-related, with any significant product deferred revenue amounts having been recognized for the income statement in the first half of the year. Accounts payable days were 31 days, down from 37 days in Q2, reflecting the timing of inventory receipts and payments. Capital expenditures for the quarter were $4.7 million. Turning to our outlook for the fourth quarter and beyond. We continue to experience significant volatility of demand from our cloud business.
We saw a strong recovery from the customer who had paused activity in the second quarter, only to be surprised by a dramatic reduction in forecast for Q4 and 2020 from another key Titan. All indications are these actions do not represent a loss of positioning or share for Arista at these customers, but will likely result in demand from this part of the business being slapped down on a year-over-year basis for the remainder of 2019 and into 2020. While we are not at this point in a position to provide overall guidance for 2020, we did want to make the following points. Firstly, a recap on deferred revenue and its impact on 2019 results. As outlined on prior calls, we recognized $80 million and $38 million of non-Microsoft product deferred revenue in Q1 and Q2 2019 respectively.
These amounts represented product sales, ships, and bills in the prior year for which revenue was deferred pending customer acceptance of legal redesigns and features. While not impacting our 2020 cash metrics, this still set up some tough comps for year-over-year revenue growth, particularly in the first quarter of 2020. At this point, we believe this trend, combined with typical Q1 seasonality and the recent updates to cloud forecast described above, may result in revenues for the first quarter of 2020 that are approximately 5% below Q4 '19 levels. On the gross margin front, we would reiterate our overall gross margin outlook of 63%-65%, with customer mix being the key driver. We'll 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.
Finally, you should expect to see us continue to execute against the stock repurchase mandate in an opportunistic manner. With all of this as a backdrop, our guidance for the fourth quarter, which is based on non-GAAP results and excludes any non-cash stock-based compensation impacts and other non-recurring items, is as follows. Revenues of approximately $540 million-$560 million, gross margin of approximately 63%-65%, operating margin of approximately 36%. Our effective tax rate is expected to be approximately 20.5%, with diluted shares of approximately 80.3 million.
I will now turn the call back to Charles. Charles?
Thank you, Ita. We're 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.
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 1 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 the line of Simon Leopold with Raymond James. Your line is open.
Well, thank you very much for taking a question. Appreciate the added disclosures and details you've given us on this call, so thanks for that. I wanted to maybe get a better understanding of the 2020 commentary, given that at least looking at CapEx as an indicator or revenue for business lines like Azure and AWS seem to be encouraging, suggesting 2020 could be a better year in terms of the CapEx forecast going back double digits for the web scale guys. Just wondering how you think we should square your more cautious tone on the cloud relative to looks like better capital spending trends and healthy revenue trends from the web titans. Thank you.
Thank you, Simon. Well, as I was trying to explain, our Q4 forecast is actually quite consistent with many of the cloud CapEx reported in recent calls, which is overall flat to down. The overall trend for Q4 is down, and we're projecting that same flat to down trend for CapEx next year for the overall Cloud Titan spend. One of the things that, as you know, you may have remembered this, we were not tracking from a networking point of view in prior years, cloud CapEx nearly as well. There's not a one-to-one correlation. In some cases, what we're seeing in the cloud CapEx is a redistribution to infrastructure, not to networking.
If you look at the two reasons why we believe it will slow down also in 2020, it's because many of the Cloud Titan customers are extending their use of server assets and delaying the network purchase longer and buying other infrastructure or investing in other aspects. The second is the 400G adoption. We had predicted initially that deployments could start as early as second half this year. We are shipping 400GbE products for initial trials this year, but the initial deployments have shifted by more than a year to second half 2020, and we think mainstream production will be 2021. The change in customers extending their investments and the deployment of 400G is causing us to be more muted about 2020.
Thank you very much for that.
Thanks, Simon.
Your next question comes from Tim Long with Barclays. Your line is open.
Thank you. If I could just follow up on the Cloud Titans again. Maybe just a two-parter. Number one, could you talk a little bit about the customer that had the recent sudden change, and it seems it has a long tail to it as well. Any visibility into why they're doing it as their business is changing, or is it just, as you said, just the spending is changing? Secondly, do you think this is a trend that highlights even more than flat to down risks for some of the other large customers? Thank you.
Okay. Specific to the Cloud Titan whose forecast reduced dramatically, I think there were two main reasons, I'm going to ask Anshul, our cloud expert and COO, to elaborate. First is, they are managing the CapEx for networking and modulating their inventory and shifting to more of a just-in-time type forecast. Typically, they gave us two quarter visibility, sometimes even three and four, now they're moving much more to a real-time forecast at quarterly intervals. The second is this particular Cloud Titan is extending their server assets by more than a year. Once the server assets get extended, that is significantly delaying the network spend, too. Anshul, do you want to add to that?
No, Jayshree, that's absolutely right. The server refresh delay is a specific decision for this one Cloud Titan. They didn't see enough ROI in doing the refresh just right now, so they might wait a generation. Hence the impact we are seeing as well, because they won't upgrade the network if they're not upgrading the server.
To answer the other part of your question, Tim, on other Cloud Titans, we think some will be stronger, some will be flat, some will be weaker, but if we average all of that, we see flat to down.
Okay. All right, thank you.
Okay.
Your next question comes from Ittai Kidron with Oppenheimer. Your line is open.
Hi. I guess I want to follow up then on your recent explanation here, Jayshree. The CapEx moving to just in time, that shouldn't affect your business. You might not have visibility, but that still means business needs to come as long as they keep building. I guess moving to the server refresh cycle, is that where the bulk of your business with that Cloud Titan was in just refresh, upgrade of existing platforms? There was no new build with this customer?
There's always multiple levels of connectivity. Bulk of it is obviously the first layer. You got to build servers for us to put a network. The second layer is usually regional spine and data center.
Okay. Ladies and gentlemen, please stand by. We're currently experiencing technical difficulties. The conference will resume momentarily. Thank you for your patience.
Hello?
The next question, please.
Hello? Hello?
We lost the phone.
We're back.
Did you hear that, Josh?
Yeah, if you could repeat it, I think everybody got disconnected in the middle. I hope it wasn't something I said.
No. Now I have to remember your question to repeat the answer.
The question is, again, reiterating the question regarding the nature of your business with them. Is it just tied to server refreshes? There are no new greenfield build with them?
Yeah, no. Clearly, it all starts with, if we don't have servers and storage, we can't connect with the network. The nature of our use cases starts with server spend correlates to network spend, which in turn creates layers of additional spines, which can be the aggregation or the regional spines as well. That's the symptom, and the cause is more networking spend. Now, that doesn't mean they don't spend on new data centers. I think the CapEx of many of the Cloud Titans, including the one we're discussing, reflects that they will spend in a healthy fashion on the infrastructure for new data centers. To correlate that back to networking will take time, because first they have to buy the new servers, and then they have to buy the network, which could go well into late 2020 or 2021, most likely.
Very good. Maybe as a follow-up, Microsoft just won the JEDI contract. What does that mean to you? How do you look at that and what it could do for your business?
We're very pleased with that. As you can imagine, Anshul and the team worked very hard on several certifications and partnerships with our Cloud Titan vendors. Having said that, the first thing that happens with these large contracts is they get contested. While the award may be given, we think it'll be time for us to see material benefit. May take 6 to 12 months.
Very good. Good luck.
Thank you.
Thank you, Ittai.
Your next question comes from Samik Chatterjee with JP Morgan. Your line is open.
Hi. Thanks for taking my question. Just moving beyond your commentary about the volatility in spending from the Cloud Titans, I just wanted to ask about the tier 2 cloud providers. It sounds like you have more visibility or more stability in terms of what you're seeing in terms of spending patterns from them. Is that kind of fair? If you can kind of elaborate on what you're seeing on that side, does this drive you to focus more on that segment going forward?
Samik, thank you. While majority of our guidance was due to the specific Cloud Titans, you might have noticed in my commentary that both the service providers and the specialty tier 2 cloud providers came in at 4th and 5th place. I think this is the 1st time the specialty tier 2 cloud providers have been dead last. In my view, this segment is weak and the results have been mixed. I think the new tier 2 companies, specialty cloud companies that started growing very well for us in 2017 and 2018, are now having to review their investments and decide from a matter of economics, which ones make more sense. Do they rely on their own cloud or go to the public Cloud Titans? Some of the tier 2 companies are finding it difficult to compete. Some are continuing with the strategy.
It is a mixed bag for us, and especially in Q4, we don't expect much success from this category.
Okay, thank you.
Thank you.
Your next question comes from Alex Henderson with Needham & Company. Your line is open.
Great. Thank you very much. I was hoping we could spend a little bit of time, relative to this cloud issue, to what extent you're confident that there is no competitive incursion here that's causing it, and that, in fact, you have sustained share at that customer. How can we judge that? How do you get your arms around clarity around that point?
Alex, that's a very good question. From our perspective, the competitive dynamics have not changed in the cloud or in general. We always have aggressive competition, and we will continue to see aggression there. What gives us confidence that the Cloud Titans are delaying their spend or distributing their CapEx differently is, you know, we always pride ourselves in a close partnership and relationship with Cloud Titans. Generally, especially in the case of Facebook and Microsoft, they've been not only a vendor-customer relationship, but really a co-development that requires the kind of partnership which is engineering to engineering. It's not just business. When you look at that, there's no evidence that competitively or white box-wise, there's been any change. There's been a process change. There's better inventory management, there's better procurement optimization, et cetera.
You can always expect these cloud customers of ours to want to be multi-sourced. It isn't any different than we've seen in the past in behavior, in relationships. In our innovation, we have 10 400G products, and a lot of them are in trials. The relationship and the technology partnership couldn't be better. Anshul, do you want to add to that?
Sure. Thanks, Jayshree. Alex, we work very closely with these customers to a point where we are working on the 2021 roadmap along with these customers right now, and are quite well aware of the changes they're making to the architecture as well, and have very direct feedback from customers as well, that there is no alternate that's displacing us. It's simply the demand has gone down, and we are very confident of our share when that demand comes back as well, since we collaborate with these customers. We are not worried about it, and the customers are being pretty direct as well. This is not our share going to someone else, their demand reduced.
Okay. Thank you very much.
Thanks, Alex.
Your next question comes from James Faucett with Morgan Stanley. Your line is open.
Thank you very much. Adding my follow-ups to the other questions that have already been asked. On this change in architecture and strategy and what they're doing with their servers, is this related to how they're implementing servers and networking, obviously, by extension, so that we're looking at a permanent lengthening of replacement cycles? Or is this somehow just related to the current cycle? I guess I'm trying to get a sense for what the, even as the customers come back, what the opportunity is and how we should think about the frequency that they'll need to come back and add additional capacity or upgrade networking equipment, et cetera.
Yeah. No, good question, James. You may know that server cycles tend to go in 18 months to three years, and generally, they get upgraded in that type of timeframe. In this particular case, because of the server vendor and architecture there's no change in server architecture. I have to emphasize that. They are choosing to delay their new server deployments by at least a year. It's no more, no less, no change in architecture, but really a delay of server spend, which is causing a delay in network spend.
Correct. That has some short-term impact on the IO needs from these servers. If there's no new server, they may not need as much new IO that they were planning on. In the long run, these things do balance out. It's just a near-term, I'd estimate, so 1-year type of cycle until they do start the refresh.
Your next question comes from Aaron Rakers with Wells Fargo. Your line is open.
Thanks for taking the question. Maybe I'll shift gears a little bit. As you think about the model and the growth rates that you've outlined looking into next year, I'm just kind of curious, do you take a more active stance in kind of protecting the margin profile of the company? How do I think about just the investments that the company's previously kind of alluded to that would be required to really position yourself for a campus ramp as we move into next year? Just any kind of commentary on how you've seen campus thus far?
I think I'll take the model question, then Jayshree can take the campus question. I think as we think about the business as we go forward, I think we believe we can operate healthily in the plus or minus 35% operating margin model that we've talked about for some time, as we've talked about as part of the long-term model. We're guiding 36% for Q4, even with the kind of reductions in revenue, right? We do have some flexibility there. I think you have to probably expect that you won't see the 39% and 40% type operating margin numbers that we've been putting up more recently. We will continue to make investments that'll be targeted.
You'll see us continue to invest on the sales and marketing side because we believe that's important as we go forward, and some headcount, et cetera, so R&D as well. We think we can still do it within that envelope.
Yeah. Aaron, regarding campus, as I said in my opening remarks, we are marching well to the $100 million in shipments for the first four quarters. Q3 was our first quarter. What surprised me pleasantly on our campus acceptance is half our customers were existing, but half were new. If you had asked me to forecast that, I wouldn't have betted that. I would have thought 80% would be existing. We're really getting a lot of interest in our campus.
In fact, I would say one of the strong reasons our enterprise segment is number two is not only because of the data center, but small numbers on the campus in Q3, but I believe the two will influence each other, that we will become more relevant in the enterprise because of both campus and enterprise and the architectural shift that we can guide to public workloads versus private. I think when I look at why, we're very differentiated. The word cognitive to us is really architectural, both on our Wi-Fi and PoE switches and the spine with CloudVision. Customers are really appreciating our differentiation on flow analysis, on security, on bringing an integrated, cognitive, secure, software-driven integration together, much like we did with the data center. We like our early progress and execution there.
Your next question comes from Alex Kurtz with KeyBanc Capital Markets. Your line is open.
Yeah, thanks. I have more of a clarifications than a question. I just want to make sure we all understand that the account that's driving this downside here is not your historically largest customer. The second part of that is, given the disruption that you saw from Microsoft earlier in the year, I guess, what's the context of their spend level as they go into Q4 and into 2020?
Yeah. Just to clarify, Alex, it's a second Cloud Titan. It's not the one we mentioned the last time that had a Q2-
Right
2019. Specific to Microsoft, as we are projecting, we fully expect them to be a north of 10% customer concentration for 2019, and we expect to have a second new Cloud Titan customer, which will be Facebook.
Just into 2020 around Microsoft, just given the disruption we saw this year, Jayshree, any early read on kind of returning to more normalized spend in 2020 with them?
[Munjal], you want to say a few words?
Sure. Alex, so far, we don't have a long-term guidance from Microsoft.
Yep
There is nothing different than they are on a usual spend pattern. We have not.
Okay.
given any other guidance.
No blip, no pause.
That's correct.
Not yet.
Yeah.
Yeah.
Okay, thank you.
Thank you, Alex.
Your next question comes from Tejas Venkatesh with UBS. Your line is open.
Thank you. As you reflect on the fundamental technology drivers of bandwidth growth in the cloud that contributed to very strong growth over the years, has anything fundamentally changed? I ask because this year we've had sort of two different cloud vendors have some sort of hiccup. One was probably a public cloud vendor, the other a content cloud vendor. Completely different drivers, and yet you're seeing a pause. Is there any sort of technology fundamental change?
Tejas, I don't see any fundamental change. I think our strategy and TAM is valid. I think their strategy that they want to invest has been very strong the last four or five years. Perhaps the only change I would allude to is they're adding more process, more optimization, more care, and feed into their forecasting, more discipline, more hygiene. I don't see any other change. I think they continue to invest for scale, and as you know, they're all doing very well. That doesn't mean they will spend equally well.
Thanks.
Anshul.
Tejas, the pause that we mentioned in Q2 was very tied to some internal financial planning for the customer and inventory planning. It was nothing to do with the architecture or the bandwidth and so on. The second instance you're seeing right now with the other Cloud Titan is tied to their server refresh. When you model these out long term, there is no change in their growth expectations of traffic and networking needs, both from a bandwidth standpoint as well as backbone and traffic engineering needs. With video storage and now AI workloads growing, there is always going to be more and more need for networking. We're not seeing that trend change. Obviously, we have to wait for the customer to come back until they start the refresh.
Thank you. As a quick follow-up, any change in enterprise spending? I realize your share of that market is lower, but are you seeing any deal elongation and so forth at all?
It's too early. As you say, we are not the bellwether on enterprise. I think we understand the cloud much better. Because we're a new entrant and we have new products, so we're probably not the best indicator of change.
Thank you, Tejas. Next question.
Your next question comes from Jim Suva with Citi. Your line is open.
Thank you. I have just one question. The change in the procurement strategy of this Cloud Titan, why won't it spread to both other Cloud Titans and maybe even the 2nd-tier type Cloud Titans? Is there the risk of that or any visibility of why this challenge won't spread? Thank you.
Thank you, Jim. I'll comment, and I'll have Anshul elaborate. I think the short answer is no, we don't see a lot of risk of that because anyway, our visibility was 2 quarters. With this particular Cloud Titan, they've optimized it to 1 quarter. If we were always relying on 1 and 2-year forecasts, that would be a bigger dramatic change to our belief system and how we plan with them. Since it's always been 1 or 2 quarters and a further refinement on this particular Cloud Titan customer to 1 quarter only, we don't see a big change or big shift.
Great. This is their own internal process and planning on how they plan networking purchases with respect to these data center facilities going live. They are optimizing their processes and their org and fixing issues they might have had in the past. This does not apply to any of our other Cloud Titans. They're very specific to organizational issues in a company, not an industry trend or a technology trend.
Thanks, Jim.
Your next question comes from Rod Hall with Goldman Sachs. Your line is open.
Yeah. Hi, guys. Thanks for the question. I want to just check what you're thinking on growth in 2020. I'm just playing around with the verticals here and thinking about what maybe you're implying with this. I wonder if maybe you could put us in some sort of a ballpark for overall revenue growth, and then talk to us about why the down eight or so that you're implying in the guidance doesn't kind of materialize through a better part of next year, so you end up with even lower revenue growth. Maybe that's what you're already thinking. That's one thing, if you could just put us in some kind of a revenue growth ballpark for next year. The other thing I wanted to ask is enterprise spending is clearly very weak, and a lot of the rest of this growth depends on enterprise.
I wonder if you could just update us on what you're seeing there, how much risk you think there is to your enterprise numbers as we look into 2020, or at least the early part of it with the slowdown that we're observing. Thanks.
Yeah, Rod, I'll take the first part of that and then maybe hand off to Jayshree on the last part. I think from a model perspective, we're not yet trying to call a 2020 growth rate for the overall business. I think we've tried to put some pointers out there and make sure everybody is aligned on some of the impacts from deferred, et cetera. I think as we enter the year, we will have a tough comp for the first quarter in particular, and the second quarter as well, to some extent, because of the deferred. We've talked about the cloud vertical being flat to down. What we've seen pretty consistently is that the enterprise part of the business, the financial verticals have been growing well and have been offsetting that, although not entirely. We hope that will continue, that we continue to see that.
That's an offset to the other part of the business, which is really the cloud and service provider pieces, which have been muted as we've gone through this year. I think we've pretty much guided the Q1 number only because we want to make sure we reflect the deferred correctly and that we reflect the seasonality of Q1 correctly in your model.
Okay. Then Enterprise, if you want to, can you tell us.
Yeah. On Enterprise, I do believe from a demand and TAM perspective, we have a lot of opportunity for execution. We could do very well both in Enterprise and Financial. If we get affected by macro situations, that affects everyone, not just us, so we would be influenced by that. Barring any macro situation, we feel very good about our execution to date and going into 2020. That'll hopefully offset some of the flatness in the cloud that's here.
Just coming back on the numbers, just to see whether.
Go ahead, Brad.
Okay. Well, I was just going to say, it looks to me like it could easily be mid-single digits growth next year, and I don't know if that's a crazy number from your point of view, or is that a plausible scenario?
I don't think it's a crazy number. We'd have to grow Enterprise into significant double digits. I think at this point, we're not feeling strongly optimistic about the mid-teens growth that we projected at the Analyst Day because of how poorly cloud is doing. Right?
Right.
Everything else in the Enterprise and Financials is doing well. How about you give us, looking into Q4, Q1, and then we'll come back to you. We don't know.
Okay. Appreciate it, Jayshree. Thank you.
Thank you, Rod.
Your next question comes from Amit Daryanani with Evercore. Your line is open.
Yep. Thanks a lot for taking my question, guys. I guess a question of verification. Just to ensure the entire $130 million revenue miss versus the street at least, was that all attributed to this Cloud Titan customer shifting patterns, or was there something else? That's one part. Secondly, maybe we just touch on what are you seeing on the Enterprise side, and do you think Mojo could be a meaningful driver for revenue growth as you get into calendar 2020?
Yeah. To quickly answer your question, Amit, the specific Cloud Titan was the absolutely the largest part of the gap in guidance. There were declines, as you know, we've had a deteriorating declining performance in service provider. New to the mix was a deteriorating and declining performance in tier 2 specialty cloud too. It was a combination of all three, with majority being one specific Cloud Titan. Specific to Mojo's very much factored into our campus numbers, and we think the whole wired, wireless, cognitive, edge is really getting ignited with the Mojo product. We have completed our integration of CloudVision and Wi-Fi. Our distributed cloud-managed Mojo is better than many of the standard controller offerings and legacy offerings in the marketplace. We believe our campus is doing well, and a good contribution from that is Mojo.
Your next question comes from Brian Yun with Deutsche Bank. Your line is open.
Hi. Can you talk about what's changed on the 400G side? It sounds like the anticipated deployments for 400G have been pushed out one year from your initial estimates. Kind of interested to get your view on what you think is causing those delayed deployments.
I think when we first began our 400G foray, you may have remembered Andy Bechtolsheim spoke about it, we were always concerned, not whether we would have products and differentiated ones, which we do, whether the optics was ready and the ecosystem was ready. We're shipping 10 types of products now. We thought the ecosystem would be ready by now, and the optics has pretty much moved a year. Correct me if I'm wrong, Mansour. By virtue, you cannot build 400G products in anything more than trials if you don't have good optics to connect to it. By virtue of the 400G optics moving out, we believe most of the initial deployments will move from second half this year, which is what we thought before, to second half 2020.
Which means production installations is when you go from thousands of ports to million ports, will really be 2021. I want to be clear that we are shipping 400G products. We're very proud of them. As always, we're ahead of the industry.
Your next question comes from Paul Silverstein with Cowen & Company. Your line is open.
Jayshree, I hate to be the umpteenth person to ask you about Cloud Titan, but I will be. Two related questions, if I might. One, I just want to make sure I've heard you correctly first before the question, which is you said two 10% Cloud Titan customers, Microsoft and Facebook, for 2019. Was that the statement in terms of the time period?
Yes, that's right.
Okay.
That's correct.
You also said that the particular Cloud Titan in question that's the problem has shifted to one quarter from a two-quarter forecast. Correct?
Our forecasts were anywhere from two to four quarters, and they've now shifted specifically to one quarter.
All right. Here are the two questions. One, with respect to the softness in 2020, putting aside 4Q19. With respect to 2020, you made the point previously that when Microsoft, when they went cold turkey previously, you said the real question isn't them coming back, but to what degree. When you talk about the softness, the significant decline in 2020
I presume they've given you that insight, notwithstanding the shift to a one-quarter forecast away from the previous two- to four-quarter forecast. It sounds like they've given you visibility into next year.
That's right.
How much softness are you talking about, are you expecting at this point? The other related question, somewhat different, but-
If you can allow me to answer that, Paul, and then we can get to the other questions.
Sure.
This particular Cloud Titan has not only given us a dramatic reduction for Q4 2019, but has given us a dramatic reduction for much of 2020. Unlike the other Cloud Titans, where there's a pause and they come back, and it's more consistent, we fully expect this particular Cloud Titan to reduce next year significantly.
Okay. All right. On the broader question, 400G, correct me if I'm wrong, but I think you've made this point publicly in the past that with respect to your business of selling intra-data center switches for leaf-spine and top-of-rack, there's a 12.8 terabit upgrade cycle driven by new Broadcom silicon Tomahawk and Trident, as well as the Innovium equivalent. In that, if 400G optics aren't ready, the various Cloud Titans will buy the far higher capacity switches that you and Cisco and Juniper are introducing, and they'll just deploy them in high density 100G configurations until the 400G is ready and/or is at the right price points. It sounds like you've changed your thinking on that.
There's no change in thinking that Jericho and Tomahawk Trident will be used with higher capacity in 100G configurations. We'll continue to see incremental deployments of that. No change there. There won't be a wholesale change from 100G to 400G in the spine until 2021.
Well, Jayshree, I can't ask you-
100G will be alive and well.
We can't ask you to speak for your customers. That would be unfair. Does that imply that there's been, and maybe this is stating the obvious, but does it not imply that there's been a change in demand on the part of Cloud Titans in general to the extent that if the demand were there, again, instead of deploying 400G, they would just deploy a lot more 100G, and the impact to you as a switch vendor would be relatively nominal. You won't see it, we won't see it. It clearly appears that there is a general softness in demand.
Paul, to your overall theme, we do very well and have good products that are 12.8 Tbps, 128 by 100G switches. That point is well covered, and customers do buy that as necessary for the architecture. The dramatic change for one of the Cloud Titans is really coming from them delaying their server refresh, which delays a network change that they otherwise would have done, which would have added more capacity. It gets delayed until they start that refresh because these architectures go hand in hand. You don't touch the network and the servers independently. It goes together in a cluster and in your whole architecture. On the 400G side, the industry delays are in general because they will change the entire ecosystem.
Many of the optics companies forgot about backwards compatibility, which doesn't work for the cloud companies because 400G has to work with 2 by 100 on the other side and so on. Otherwise, you can't upgrade a large network. Everyone's going through those motions to get the entire ecosystem ready, which will take at least a year before it starts getting deployed in production.
Thanks, Paul. Next question, please.
Your next question comes from Jeffrey Kvaal with Nomura Instinet. Your line is open.
Yes. Thank you. I guess I completely understand the downtick in cloud spending over the course of the next few quarters, and I guess now we will have to start thinking about what your comments mean about when we might come through the other side of that. It sounded as though 18-24 months after a server refresh is over is the typical timeframe. If you pushed it out six months for whatever reason, that would strike me as being maybe in the fourth quarter of 2020, but more likely 2021. Does that math work out?
Yeah, Jeff. I think the reason we are saying the Cloud Titan forecast will be slacked down in 2020 is because any projections of optimism and growth, the earliest possible is Q4 2020. We really think the impact is 2021 to get beyond the slacked down forecast. You're right, a server cycle that's delayed by a year means it's a year from now, which means any impact to IO is a year from now. Any kind of production deployment is over a year from now. I think 12 months is a good rule of thumb as a minimum.
Thanks, Jeff. Next question, please.
Okay.
Your next question comes from Sami Badri with Credit Suisse. Your line is open.
Hi. Thank you. I'm just trying to square away some of the commentary here. To give you some context before my next question is that some of the third-party multi-tenant and colocation providers have reported some of their strongest backlog quarters, all for data center capacity commencing in Europe in 2020. The fair majority of those big bookings are actually being driven by Cloud Titans. As these Cloud Titans expand internationally, is there a difference in topology or network architecture that is taking place because of the way they're building or the way they're connecting that means that possibly alternative vendors would be more favorable? Are we looking at something that is completely uniquely different than we've ever seen before?
Sami, typically the Cloud Titans deploy us globally. As you know, we have a lot of international data centers that they have built with us. As they expand to additional data centers, the only difference is not architecture, but size of the data center. Depending on whether they're going into highly populated densities or smaller, they will rinse and repeat the same architecture, but the scale and size may vary. They really want it to be one uniform architecture, ideally with one consistent software and one vendor. It is very rare to see that they would repeat it with a different vendor. I don't see any uniqueness in the international data centers except size.
Got it. Thank you for that color. Just, sorry, one follow-up on the.
Can we do the next? Sorry, Sami, we're running out of time here. Can we go to the next question, please?
Your next question comes from Erik Suppiger with JMP Securities. Your line is open.
Yeah. Thank you for taking the question. I just want to understand if this issue with the Titans is beyond this one customer as you get into 2020, because it looks like you're reducing the outlook for Q4 by, call it, 20-ish% of revenue. I'm not sure if we should carry that through, and it seems like one customer, who presumably is not 10%, why would we be cutting it that much? Is this more endemic across the broader Titan universe, or how should we be thinking about this?
I think what we're saying is our numbers are very large with the Cloud Titans, and we will do very well with some, the smaller ones. We may be flattish to down with the larger ones, depending on their spend. When you cumulatively add all of this, it's going to be flatter down in revenue. I don't think it means it's not a trend for overall the next three to five years. It's a projection of what specifically will happen in 2020, especially because they're going to milk their servers, leverage their existing infrastructure, keep adding 100G to that, not quite move to 400G. We see 2020 as a transition year with our Cloud Titans in terms of high performance as well. I wouldn't read anything more to the forecast except for 2020, because things could pick up later on.
Thank you, Erik.
Your next question comes from Tal Liani with Bank of America Merrill Lynch. Your line is open.
Hi, guys. I want to understand just one thing most of the questions were asked with. I want to understand, if you look at 2020 and you remove this one customer, because next quarter, you're going to have a down year, revenue down year-over-year. I think you said also 2020. Now I'm trying to understand, if I remove this one customer that was really bad this quarter and provided this $120 million, $130 million shortfall, then what's the underlying growth of everything else? Is 2020 the story of one bad customer who is kind of rethinking strategy, or is 2020 going to be down even if you remove that particular customer?
Oh, loaded question. I think the way to think of this is we have 3 types of Cloud Titans. Some large ones that'll remain flat, some that'll go down, and some that'll go up. The aggregate of that is flat to down. Archana, do you want to add to that, or did I get that right?
No, no, that's right.
Right. I understand, that's a very general answer. At the end of the day, there is one big customer who is down $120 million off $670. That's a giant number. We have to remove that to understand what's happening with the rest of it. When you do the math, is the rest of it still up or it's still going to be down?
First of all, remember, the rest of it, we don't have hundreds of customers here, we have a handful. The Cloud Titan segment.
There's still $550, right. The per quarter, it's still $550. It's still meaningful. Very meaningful.
Yeah. I think, Tal, we have to step back a little bit. It's not like we have perfect visibility of what we think is going to happen in 2020 yet, right? I think what we're saying is, look, we have this issue with this particular customer, which is a large customer and has some significant impact. We have the rest of cloud, which has been unpredictable this year. That's the reality of life. We'll see how it plays out next year. For sure, we're not aggressively forecasting that right now. Right? We have service provider, which hasn't been performing well, and we just saw the specialty cloud vertical kind of go to the bottom of the list, right? Offsetting that has been some good traction in enterprise and financials, and we saw them grow well in Q3, right?
It's not enough to offset completely, which is why we say the flat to down is significant for the rest of the business. Thanks, Tal. Next question please.
Your next question comes from George Notter with Jefferies. Your line is open.
Hi, guys. Thanks very much. I know that you guys are making quite a bit of progress on the routing side. I know there were a number of larger cloud provider customers that were looking at your routing products. Is that an opportunity for you to offset some of the softness you're seeing on the cloud side as incremental success with routing, and any insights there would be great. One other clarification. Ita, if you could just repeat the deferred revenue metrics that you referenced earlier, that would be helpful. Thanks.
Yeah, George, you're very right. Despite all the servers, et cetera, these tend to be obviously stronger in value and fewer in number, but we're doing very well with virtually all the cloud providers, Cloud Titans on routing, and we're doing very well in general in routing. It's improving for us.
On our landscape with tier 2 cloud providers as well, in the routing use case specifically. We do look at that as an opportunity for next year as well.
Just on the deferred, George, this is a continuation from what we saw in the first half of 2019, right? We had recognized $80 million in product deferred in Q1, $30 million in product deferred in Q2, that had basically been shipped and billed in the prior period. From a revenue perspective, that always sets us up with a tough comp coming into Q1. From a cash perspective, it's different, obviously, because that deferred revenue didn't contribute cash for this year. From a revenue perspective, it's obviously gives you a higher bar.
Thanks.
Thank you.
Your next question comes from Hendi Susanto with Gabelli. Your line is open.
Thank you for taking my questions. I would like to understand more about the Cloud Titan phenomenon of delaying the use of its server assets. Is there a risk that other customers, whether Cloud Titan or not, may behave similarly? Perhaps you can share some technical insight when a large customer can extend the use of server assets and when it cannot.
Hendi, this is All of the Cloud Titan customers have their own architecture and the choice of mix and the 25G, 50G architecture choices they've made. If you look at the industry, the Cloud Titans are actually not aligned to the exact same server CPU upgrade cycle. They are offset from each other in a tick-tock manner, aligned with the tick-tock updates from the industry as well. This particular decision does seem specific to only one Titan to us. We have not heard this from anyone else, and the others keep on adding capacity as they need to. This one customer, for them, they felt there wasn't enough ROI, so they decided to delay their upgrade. I would just read it the way we have heard it from our customers, and I would not add anything else to that.
Thanks, Hendi.
Thanks.
Your next question comes from John Marchetti with Stifel. Your line is open.
Thanks very much. I just wanted to go back to a comment and make sure that I heard you right, Jayshree. When you were talking about the overall outlook for 2020, ex the cloud business, if I go back to that Analyst Day presentation and how you guys talked about cloud adding a few points of growth to take you up into that upper-teens range. If we strip that out altogether for next year or even assume it's down a little bit, do we still consider the rest of the business, ex that cloud, still being in that low double-digit to mid-teens range? I just wanted to make sure I heard the way that you answered Rod's question.
Right. No, good question, John. I think just going back to that Analyst Day, what Ita said, or Arista said more specifically, was if the cloud did really well, we'd be in the high teens. If the cloud did average, we'd be in the mid-teens. The current projections we're giving to you on the cloud are below that average that we thought was the norm. With the new norm being flat to down, mid-teens is off the table right now for 2020. Ita, you want to add to that?
No, I think that's right. We hadn't contemplated a world at that point where cloud would be flat to down. Right? That wasn't something that we were contemplating. We were kind of thinking between slowish growth and faster growth, right? I don't think anybody that day would've thought that we would be in a world where we would see that part of the business actually be down to, or flat to declining.
Thank you, John. We have time for one more question, operator.
Your last question comes from Andrew [Statham] with Wolfe Research. Your line is open.
Hi, thank you. Recognizing the revenue pressures in 4Q and 2020, a silver lining potentially could be through growth in other segments. Maybe campus starts to ramp over the medium term, enterprise shows strength, and maybe service provider recovers. You would eventually start to have more of a structural business pivot to higher overall gross margin. Is that something that you perhaps just go ahead and embrace with more immediacy and invest faster and even more heavily in non-cloud, maybe at the expense of cloud, than you previously would've anticipated?
You're right to point out that our business hasn't fundamentally changed. Our fundamentals are great. Our gross margins are 63%-65%. Our TAM is valid. We could grow in other places. I don't think we would want to forecast beyond the 63%-65%, because I think we're in the band. Because if our cloud goes down, we should be on the higher end of the band. If our cloud goes up, we'll be at the lower end of the band. We don't see the band itself going up more. That being said, we're absolutely committed to R&D. We're absolutely committed to targeted hiring and investing. We think we can do that, like Ita said, with a 35-ish% operating margin.
No change in strategy in continuing to invest in R&D and sales and marketing, as well as M&As where it makes sense to grow our other businesses and segments.
Thank you, Andrew.
That's helpful. Thank you.
This concludes the Arista Q3 2019 earnings call. Please note that we have posted a presentation which provides additional information on our fiscal results, which you can access on the Investors section of our website.
Thank you for joining, ladies and gentlemen. This concludes today's call. You may now disconnect.