Arista: Technology And Management, A Winning Combination In Cloud Data Centers

| About: Arista Networks, (ANET)
This article is now exclusive for PRO subscribers.


Arista was founded by two of the leading figures in the Silicon Valley world and those men still hold 30% of the outstanding shares.

Arista's technology approach in which it differentiates its switches by the use of its proprietary EOS software is considered by many observers to set the company's offering apart.

The company has enjoyed substantial and profitable growth for several years and while crossing the billion dollar revenue milepost, its operating margin is 25%.

Arista shares have been mired at modest valuations since the company was sued by Cisco Systems for patent infringements.

A couple of weeks ago, I wrote an article about Cisco (NASDAQ:CSCO) and its role in the SDN space. SDN is a concept with a long history that is just now finding broad acceptance. Part of the promise of SDN is that it would be the harbinger of an open networking world and that it would significantly ameliorate network bottlenecks. But along the way, Cisco has seemingly hijacked some of the concepts of the SDN paradigm. Cisco's version of SDN is totally proprietary and really does little to mitigate costs in a modern data center. Good for Cisco, not so great for users!

Despite Cisco's seemingly successful defense of its heartland in terms of selling more and more network switches, the company has basically been in a very slow growth mode for some years. These days, Cisco has more than 20 different operating systems deployed across multiple platforms and multiple generations of hardware. Whether it is a strategic decision on the part of Cisco or just a byproduct of nature, Cisco's ability to really innovate is hobbled by the requirements of supporting so many different operating systems.

And yet the world is rapidly changing as the scale of data centers reaches proportions undreamed of just a few years ago. There are articles that tell of Microsoft (NASDAQ:MSFT) building data centers that are populated by more than 1 million servers and many of these servers have their own combinations of virtual machines as well. And there are far more instructions that have to be processed. Just consider what happens when a Facebook (NASDAQ:FB) user wants to update his/her homepage. It might seem as though something as basic as that would be a simple problem in terms of network capacity but, of course, it isn't. Each FB update typically generates literally hundreds of ancillary transactions in its wake.

Hyper-scale users like FB and Google (NASDAQ:GOOG) (NASDAQ:GOOGL) and Amazon (NASDAQ:AMZN) simply could not and cannot sustain the cost of traditional network architectures. Currently, Gartner says that network overhead, i.e. the switches, the servers and the like represent more than 50% of the cost of modern networks. Further, given the number of transactions generated by many hyper-scale users, using a legacy architecture with what are described as "over-subscribed" switches is not really an option either.

Cisco may have been successful in staving off some type of calamitous assault on its market share, but it is clearly not capturing all of the available growth in modern data centers. During the course of my research into Cisco's relationship to the SDN paradigm, I came across a name that was new to me and perhaps to many readers, Arista Networks (NYSE:ANET). Gartner had written a glowing evaluation of the company's products and prospects and so, I thought a deeper dive might be in order. If Cisco's relation to SDN is primarily defensive, Arista seems to have developed a strategy that embraces the new architecture and appears to afford users a clearer path toward building the type of efficient networks that are required because of the current demands of hyper-scalers.

Arista is not without its controversies to be sure - anytime your principle competitor is Cisco, investment controversies are almost certain to arise. Arista is the defendant in a significant patent infringement case brought by Cisco, that could take up to 50% of the company's current cash balance of $700 million. Arista generates a substantial amount of cash flow. Last year, its free cash flow grew by more than 50% to $180 million. I think some scenarios that have been published regarding the company's liability and how it might impact operations are more than a bit overblown. I will discuss the case below although I make no representation that I am a patent attorney or have any particular expertise.

Arista shares are currently trading toward the bottom of their 52-week range, primarily because the concerns primarily regarding the lawsuit have sparked investor fears. Usually, companies that are achieving 25%+ revenue growth and are said to be leaders in cloud-based networking sell for valuations that make little sense to many readers. Here, using the basic valuation metric of price to sales of about 4x ratio seems to compensate for significant fears and risks.

Arista does have broad analyst coverage. 27 analysts provide estimates and the average of their forecasts calls 27% revenue growth in the current year. Arista is a company of some size and a relatively long operating history and revenues are likely to cross the $1 billion mark this year. That being said, there have been relatively few articles on Seeking Alpha discussing the name. About half of the company's shares are held by insiders and most of the balance (42%) is owned by institutions. And its short interest is 23% of the float. If the investment thesis is accurate, then the shares have significant runway. I will try to touch on a few highlights of what the company does, how it does it, and why that is better. I will avoid commenting on the specifics of what makes up Arista's product family. It will not help most readers in evaluating the shares and the material is incredibly dense. That being said, the new 7500 switch/router that the company announced the other day is another clear disrupter. The capabilities of the new product can replace the routing layer in many networks. There are major implications in terms of performance and the elimination of network bottlenecks and it will certainly enlarge the size of Arista's TAM. Arista these days really does have the best and the brightest in the networking space. It is always interesting to consider a new name.

OK - What is Arista's secret sauce? This is a company that makes large Ethernet switches - there are plenty of competitors and potential competitors.

Yes, there are. This is the hardware business after all and vendors have been making Ethernet switches for eons. But not all Ethernet switches are created equal. In the latest Gartner review of the space, the consultant's write that "unlike in the past, vendor differentiation is shifting toward software - including management, automation and orchestration - compared with hardware." Gartner's Research Director in this space, Andrew Lerner, writes that "Arista has a scalable architecture, tight integration with a variety of orchestration and SDN products and flexible software." Mr. Lerner also writes that "Arista has taken a very open and agnostic approach that's cost-effective so it's a very compelling story for the enterprise… that doesn't lock you into a specific architecture." Of equal interest is the fact that Arista jumped several places from the 2014 Gartner survey to the current position it enjoys as the absolute leader of the 11 vendors.

The secret sauce that Arista has developed is its EOS operating system that runs on its switches. To the extent that Arista is said to have the best software out there and that software has become the most important differentiator, it is hardly surprising that Arista is by far the fastest growing vendor in the space. They just seem to have what most users say they want: flexibility, scalability, reasonable price points and an open/agnostic platform that works with devices from most other vendors.

While there will always be a market for whatever it is that Cisco sells, and in point of fact Cisco is ranked #2 in the leader quadrant by Gartner, many users are going to reject Cisco's ACI with its relatively rigid architecture and relatively high price points. That isn't to say that Cisco will not be successful with ACI regardless of how closed it may be. Cisco may lose market share to Arista. That has been going on for a few years, but it has a huge amount of market share that it could lose before anything became too noticeable because few investors are looking for Cisco to provide much growth.

At the other end of the market, there is a category called White Box Switching. White Box Switching is the ability to use off-the-shelf, generic components in the forwarding plane of an SDN. Gartner's report said that particularly in the market for hyper-scale data centers, white box switching is becoming more acceptable given that proof points as to its efficacy are proliferating. Some white box switches come pre-loaded with minimal software while many are simply "bare metal" devices. There seem to be two camps regarding white box switching. One camp says that generic switches, because of their lower cost will find a broad market and will hurt those vendors who offer a more proprietary hardware/software platform. On the other hand, some observers argue that users are going to want a network composed of vendors who are responsible for overall performance and are willing to pay extra for proprietary software. It seems likely that the benefits of white box switches are not distributed evenly. Most SMB's will never be able to reap the advantages of using white box switches when compared to hyperscale users such as Facebook and Microsoft. I don't purport to have second sight regarding these trends but white box switches are hardly a new phenomenon. I think they are getting more publicity now than in the past because they are part of the data centers that the 7 large hyper-scale vendors are putting together. That is not an insubstantial market to be sure, but again according to Arista's market research, the real sweet spot these days and projected through 2020 is for all other cloud deployments + service providers outside of the top tier. In any event, all is hardly lost for Arista even within the cloud titan (their words not mine) space.

At the moment, according to ANET's own market research, Cisco holds a market share of around 61% compared to ANET's 12% market share in the classification called high speed data center. As the market itself is said to be growing in the mid to high teens, it seems quite likely that white box vendors will continue to increase and take share without significantly impinging on the ability of Arista to continue its growth trajectory. Indeed, Arista's growth in what it calls the Cloud Titan market has been exceptionally strong even through the end of last year. Microsoft represented 12% of Arista revenues which is down in percentage and up in absolute dollars. I think that the most likely scenario is that some of the Cloud Titans will introduce more and more white box switches into more of their data centers but will also continue to expand their deployments of Arista switches in a co-existence fashion. So white box phenomenon or not, Arista is getting its share of switch deployments in the public cloud.

I think that overall, part of Arista's success is having a product that hits most of the right buttons in terms of its appeal to users and part of its success is based on addressing the fastest growth component of the data center Ethernet switch market in terms of the service providers and the enterprise. But the TAM of the data center Ethernet switch market is huge and Arista has 12% of it suggesting it has plenty of leeway in which to grow.

I think it is self-evident that the moat here is Arista's EOS. It is simply far more difficult to recreate software than to go down the hardware cost/performance curve. I also think that it is evident that within the high speed Ethernet switch market the hyper-scale vendors are growing but the tier just below that of the hyper-scale users is growing most rapidly. while demand for public cloud providers continues to expand, the expansion beyond the hyper-scale users. It is the hyper-scale users who might try to recreate the functionality of Arista's EOS and some will doubtless go down that path. But for even slightly smaller users, even contemplating trying to build an EOS is something quite beyond the pale.

To use an analogy that comes to mind, NetApp (NASDAQ:NTAP) first developed its OnTap operating system as long ago as 2000. OnTap in its various implementations was the single bit of technology that distinguished the company from all of the other storage vendors that abounded in the first decade of this century. Everyone knew that OnTap allowed users of those years to do the same things that operators of data centers want to do today. And yet it took more than a dozen years until a company with 4 times the market share of NetApp, EMC was able to catch up and offer its own operating system. I don't suggest that the analogy is perfect and some will argue, I imagine, that either the white box phenomenon or some other unexpected competition will prevent Arista from continuing to take market share. But there really is nothing visible at this point and some competitors such as HPE (NYSE:HPE) that actually are thought of as competitors, are actually partnering with Arista and OEMing their switches for use in a unified solution that will be offered by SIs and other value added resellers.

Arista has had a consistent history of significantly surpassing earnings estimates and not by a small amount. In its last reported quarter, the beat was 31%. Operating expense growth overall was 38% year on year although a quarter of that growth came from the litigation expense that Arista is incurring at this time. Most of the beat was simply related to stronger than planned revenue growth. Revenues had been forecast to be $240 million and they actually were $254 million. With many opex categories more or less fixed in the short term, most of the additional gross margin dropped to the bottom line.

Like some vendors in the networking space, this company does not believe that it has visibility to forecast results for more than one quarter. The company has a history of some seasonality with its Q1 quarter typically down from the prior year Q4. The company forecast that this Q1 would follow a similar pattern and would show growth of 33% down quite a bit from the growth of 42% seen in Q4. In 2015, Q1 revenues increased 3% from the 2014 Q4. So, forecasting that Q1 revenues will decline by 7% sequentially seems to be an exercise in conservatism that has been typical of Arista. Management commentary during its last conference call suggests that it sees macro conditions as favorable for demand and that it has not seen any decline in the enterprise space that Cisco spoke about when it presented its results for what was fiscal Q2 in the middle of February. There is no indication that either the competitive situation or the demand picture has changed materially. It would not be surprising given the difference between the prior seasonal pattern and the results forecast for Q1 '16 to see a significant upside.

The consensus estimate for the full year is for revenue growth of 27%, far below the 42% growth achieved last year and for EPS growth of 13%. Try as I might, I find it more than a bit difficult to believe that non-GAAP margins will decline significantly during the course of the current year, and I also find it really tough to believe that a company that will just cross the $1 billion revenue mark in the current year is likely to see a significant revenue growth slowdown. As mentioned above, this company has a relatively long history of beats and raises. The current $2.76 EPS projection is up from $2.60 90 days ago and last year the company beat its projections every quarter by amounts ranging from 35% to 11%. I think that it is more than likely that a similar pattern will set up in 2016 and that strictly looked at from a business perspective, the potential for a significant and continuous series of quarters that are both beats and raises is the set up for significant share price appreciation.

While the overhang of the Cisco lawsuit is no trivial matter, that overhang has persisted now for some period of time and the fact is that the company both on its latest call and before spoke to having workarounds should the lawsuit result in an adverse ruling. Given Q4 results that were achieved under the litigation cloud, and given company assurances of a practical and timely workaround, I think that the lawsuit contingency would primarily be on the damages to which Arista could be liable. I think analyst opinions from several major brokerages advising investors to steer clear of this name until after the settlement of the litigation are quite misplaced.

I understand the EOS differentiator and that software rather than hardware is driving more and more cloud networking decisions - but how did Arista get to where it is today? It couldn't have just been that!

It really is not a useful exercise to discuss the details of this company's product line-up. It is really hard finding clever and interesting things to say about the specifics of large Ethernet switches. Perhaps I ought to seek assistance from the people on Saturday Night Live because on my own, I am just stumped.

There really have been two basic factors driving the success of Arista besides the EOS software. One of these is partnerships. This company has had a consistent strategy of not attempting to develop all the components of a complete data center solution itself. In that regard, it casts itself as the "anti-Cisco" which these days competes for all most all of the necessary components of modern large scale data centers.

Arista partners with all of the major security vendors including Checkpoint and Palo Alto. It partners with the storage vendors; it partners with Dell even though Dell is perhaps a leader in white box switches. One of its newest and apparently most successful partnerships is with HPE. The offering utilizes Hewlett's converged architecture and storage in conjunction with Arista's programmable network platforms to deliver optimized data center solutions. At the moment, Arista has no fewer than 50 technology partners and it is the openness and flexibility of its operating system that allow it to integrate so easily with so many partners. Many, many IT vendors talk about partnerships which are announced and then disappear almost like the dew on a summer morning. It is partnerships that have enabled this company to grow rapidly with a very controlled sales and marketing spend.

The other major component of the company's success of its technology strategy which involves leveraging commodity silicon in conjunction with its EOS. I don't think it is necessary or useful to discuss the details of the strategy and in any event, those kinds of overarching pronouncements are often just that, pronouncements not realized in practice. Over the years, many companies have articulated similar technology strategies but far fewer have been successful in achieving the goal of using commodity silicon to drive costs down the curve while maintaining functional differentiation.

So far, at least, the company has been doing better than many would have forecast in getting the benefits from its technology strategy. The company has maintained gross margins just below 65% and net margins last year surpassed the company goal at 26%. Stock-based comp, at just over 5% of revenues, is quite modest especially given the company's emphasis on attracting "star" quality engineers from Cisco. Again, while the company's technology strategy is far from unique, its ability to execute that strategy and to enjoy the benefits from using commodity silicon as opposed to custom chips, is hard to execute. This company seems to have been able to pull it off to this point and its emphasis on EOS is providing the differentiation it needs to achieve decent margins and rapid top line growth.

Why oh Why do I have to discuss patent litigation?

Patent suits are hardly unique in the IT space. By its nature, technology invites claims as to what is unique and patentable and what is in the public domain. I really don't purport to be a world expert on the topic, but I think that if one simply goes through the background of what this matter is all about, there are some tentative investment conclusions that can be drawn.

Arista started as a company in 2004. The founders were Andy Bechtolsheim and David Cheriton. For those unfamiliar with the names, Bechtolsheim was a founder of Sun and its chief hardware designer in that company's glory days. Cheriton is one of the leading technology intellectuals in the Silicon Valley where he teaches at Stanford and has become very wealthy investing in names such as Google and VMware. But a significant portion of his wealth came from selling Granite Systems to Cisco in 1996. He is said to be a renowned expert on distributed systems and networking. Jayshree Ullal was recruited by the founders of Arista in 2008 to come and run Arista. Her background: she was Cisco's VP of Data Center & Switching reporting directly to John Chambers. Her career at Cisco spanned more than 15 years and she was a talent whose loss Cisco and John Chambers found hard to swallow.

Not terribly surprisingly, Ms. Ullal, over time, recruited loads of star talent from Cisco. Mid-level engineers at Cisco these days are very unlikely to get rich and many of them had worked for Ms. Ullal. Cisco management hates and fears Arista and its key employees at least as much as a spurned lover might hate an ex-paramour. Some kind of litigation was probably inevitable.

The lawsuit was filed in December 2014 in the last months of John Chambers' tenure as Cisco CEO. The lawsuit alleges that Arista had stolen its intellectual property and copied what is called the Command Line Interface commands. The commands in question are certainly embodied in Arista's products. The issue is whether or not the commands are within the public domain. When the lawsuit started, it covered 12 Cisco features covered by 14 Cisco patents. According to Cisco, the patents were granted to former Cisco employees who now work at Arista.

There appears to be a convergence of opinion that Cisco filed its infringement suit when it did because of Arista's success. It is not really possible for this writer to say, even after 15 months, if Arista is using some Cisco features illegally. The mechanics of the suit relate to attempts by Cisco to bar importation of Arista switches that are said to infringe on patents involved in creating the CLI. So, perhaps the biggest concern to both investors and Arista customers is that Arista will be cut off from the supply of products it sells and that Cisco could claim damages from all of the Arista customers using switches that incorporate the stolen technology.

In early February, the International Trade Commission made an initial determination that Arista infringed on 3 Cisco patents and did not infringe on 3 others. At about the same time, Arista filed a counterclaim against Cisco alleging antitrust violations under the Sherman Act. The heart of that allegation is that "Cisco uses a bait and switch strategy" where Cisco claims that its command line interface (CLI) is an industry standard and then attempts to penalize competitors who use it. Arista also claims that Cisco raises the prices of what is called Cisco's SMARTnet service and support offering for customers who deploy competitive networking equipment alongside their Cisco infrastructure. It is almost certain that the latter allegation is true. The question is whether or not discriminatory pricing alone is enough to sustain an anti-trust complaint.

At this point, Arista has said on numerous occasions and before many different audiences that it has developed workarounds that will not infringe on the Cisco patents in question. Ms. Ullal has stated that the workarounds are ready to ship and customers will not notice any difference in performance or functionality of the new switches. In some cases, Arista has apparently started shipping switches with the workarounds incorporated.

This is still a lengthy process which is nowhere near the finish line. I think it is self-evident from Arista's performance to conclude that its customers are not too concerned about the litigation. Billion dollar companies growing 40% must have customers who think the litigation will have no impact on their purchases. On the other hand, investors are quite concerned and have valued the shares at modest levels based mainly on concerns about what might happen if the case is decided against Arista.

I have no reason to disbelieve that Arista has not developed workarounds and thus, regardless of the outcome of the suit, its business will continue unabated. If that is really the case, then it is all about damages. The courts have thrown out some of Cisco's damage arguments to this point. The company has about $700 million in cash and generates lots of cash each year. I would be surprised if any damage award might be so high as to endanger the financial stability of this company or deter prospective customers from continuing to buy Arista products. And that is about all I can usefully say about this litigation other than to add that most of these conflicts are settled before they are decided and that John Chambers, who probably was angrier about the personalities involved in this matter than anyone else is now retired and Chuck Robbins is running the company.

How to figure the appropriate valuation for a company whose franchise is under legal challenge.

Arista only became a public company in June 2014 so there is not much valuation history from before the lawsuit was filed. It obviously commanded significantly higher valuations before the filing of the lawsuit than it has since. Its all-time high was $93 and that occurred at the end of September 2014. In the wake of the lawsuit, the shares reached a trough of just over $56/share back on February 8th of last year. During the tech panic earlier this year, the shares reached a low of $53 before rebounding modestly to about $61.

The shares have been subject to wild gyrations depending on analyst commentary either about the company's competitive positioning or about the state of the lawsuit. In the middle of March, brokerage firm Jefferies sent out a note to the effect that suggested that the company's hyper-scale customers are migrating away from the kind of branded hardware that is sold by Arista. I have discussed that trend before. It really doesn't worry me that much. There will be mixed networks at hyper-scale customers for many years to come. I think that it is rather illogical for the Jefferies analyst to have lowered his 2016 revenue forecast for Arista in the wake of a lively discussion regarding the specifics of the migration held on the last conference call. The CEO said what she said and one can either believe her or not. I tend to believe her.

Arista has a current enterprise value of $4.3 billion. Revenues this year are estimated to be a bit greater than $1 billion. The company's P/E is 22X this year's estimate. In 2015, the company generated $180 million of free cash flow. Using estimates for consensus, revenue growth and consistent cash flow margin yields free cash flow of $230 million. Fee cash flow of $230 million provides free cash flow yield of greater than 5%. Obviously, with those valuations, the company's shares are not deep value. But does anyone imagine that absent the lawsuit a company growing as fast as this, with sustained and relatively high levels of profitability and clearly a leader in what it describes as "software cloud networking" wouldn't command much higher valuation. Who really can tell - double, I think so. Maybe triple. Trying to answer hypotheticals is simply not something at which I am any good. No omniscience here, I am afraid - I leave all that to the politicians.

My contention, for what it is worth, is that an investor is being paid an enormous sum to wait out a lawsuit whose settlement, one way or the other, is highly unlikely to damage this company except from a liquidity standpoint. That really isn't that hard to bear when a company generates a sizeable amount of free cash-flow as Arista does. It is a bet - of that I am aware - and some readers are going to throw up their hands and declare they don't want to be involved in a name with an overhang - but Arista really does have the leading solutions in its space and that has nothing to do with the CLI. The company has loads of partners and it will continue to thrive in the future.

Some Final Thoughts

Arista is a relatively large company that may not be known by many Seeking Alpha readers. It history as a public company dates back little more than 2 years. The company was founded by two of the best known and most widely respected names in the IT space and they still own more than 20 million out of the 68 million shares currently outstanding. Other insiders including the company's CEO and its CTO own an additional 12 million shares. The company is run by the best and brightest networking minds in the world and I think it is reasonable to assume that in its specific field the company will remain the technology leader for years to come.

The company makes high capacity Ethernet switches but its secret sauce is really its enterprise operating system that allows users high levels of flexibility and the ability to readily change configurations and to more readily manage networks. As such, Arista products enjoy some significant TCO advantages. The company sells into 4 major verticals including hyper-scale cloud companies such as Microsoft, it sells to a host of service providers and it sells to other tech companies. In the past, Microsoft had accounted for as much as 25% of revenues, but as the company has grown that percentage is likely to be less than 10% this year.

Management has taken a consistently conservative approach to guidance and it seems likely that such is the case for estimates regarding 2016 as well.

Clearly the 600-pound gorilla in the Arista room is the patent infringement lawsuit filed by Cisco. I can really not offer any dispositive commentary as to the suit's merits or demerits. I think I can say that Arista has and will have workarounds such that its operations will not be materially impacted.

In my opinion, the shares are selling for far less than would be the case, absent the overhang from the litigation. The short interest, at 23% of the float is quite high and the shares are subject to significant volatility as both the lawsuit and concerns about its strategy surface from time to time. I think the preponderance of the data points suggest that Arista shares are significantly undervalued.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.