|Shares Out. (in M):||71||P/E||41||43|
|Market Cap (in M):||5,555||P/FCF||0||0|
|Net Debt (in M):||-400||EBIT||0||0|
|Borrow Cost:||General Collateral|
Arista (“ANET”) is a short. The Company sells into a rapidly growing market but will be progressively sandwiched between increasing competition at both the high and low ends. At the high end sophisticated customers are moving to “whitebox” switches with off the shelf hardware and customized software. At the enterprise level, Cisco has deemed data center switching a core market for them and they are competing accordingly. And at the low end they are facing low cost options from software focused new entrants. We have also seen new product introductions by several companies across the spectrum and erosion of ANET’s margins is inevitable. Additionally, realizable TAM is smaller than bulls appreciate which means opportunity set is also smaller. With respect to valuation, the market bakes in extremely optimistic (and unlikely) market share assumptions to justify current valuation-even in a reasonable best case scenario for the Company the shares likely have downside. Finally, the Company is engaged in two substantial lawsuits which could materially impair the Company’s franchise and also seems to be detering incremental customers. With the Company priced for perfection, risk/reward to the short is compelling.
|Market Cap||$5,555||Net Income||$26.1||$52.6||$106.2||$170.9||$157.4||$171.6|
|EV||$5,155||EPS ex stock comp||$1.54||$1.90||$1.81||$2.14|
|EPS ex stock comp and legal||$0.97||$1.55||$2.39||$2.17||$2.34|
|P/E EPS ex stock comp||50.5x||41.1x||43.2x||36.4x|
|P/E ex stock comp and legal||50.2x||32.6x||35.9x||33.3x|
ANET is a supplier of 10/40/100GB ethernet switches for Data Center (“DC”) environments. They sell their switches into four key verticals: cloud titans/internet content providers (MSFT for their Azure cloud product most prominently), service providers, large financials (high frequency traders are a large component of this vertical) and high tech. Arista differentiates itself via its Extensible Operating System (“EOS”) which is a modular and programmable OS built on a Linux platform that allows users to customize the switch and create applications that more narrowly address the customers needs. This presents an alternative to traditionally closed architectures like CSCO whose switches embed their own (less flexible) OS. Note that ANET also uses a “merchant silicon” model whereby they buy chips off the shelf and layer on their own OS which is a departure from the more traditional model where CSCO would design their own OS and chips since they believe it allows them an opportunity to differentiate themselves via hardware as well (although CSCO has also moved towards offering merchant silicon options lately). ANET’s core customers are sophisticated companies with the in-house engineering capability to fully utilize the customizability of their OS. They also tend to be companies for whom switching is a core part of their business model and where maximizing performance is critical.
The Company was founded in 2004 by Andy Bechtolscheim and David Cheriton (co-founders at Sun) and Ken Duda and is run by a suite of former CSCO executives.
Key thesis points:
· TAM is smaller than bulls believe
· New (and intensifying) competition from whitebox threat at high end and CSCO at both high end and enterprise levels and commoditized players like Cumulus on low end. CSCO has deep pockets and doesn’t cede share easily. Dell also becoming a more dangerous competitor.
· Margins will compress as they saturate core TAM and need to address adjacencies to grow
· Valuation implications dramatically unrealistic
· Legal threat overhang with two pending lawsuits that could materially impair their franchise (and at the very least will result in real, cash legal expenses in interim)
Evidence on key thesis points:
· TAM is smaller than bulls believe.
o The Company (and investors) cite ANET’s TAM as the $8Bn/year Data Center Ethernet Switching market but a large portion of the DC market isn’t really addressable.
o VAR indicates that because ANET’s differentiation is its customizability, their core TAM is really firms with in-house engineering talent to utilize that flexibility (“the only people who really need ANET are those deploying full scale clouds with 24/7 throughput”). That eliminates most unsophisticated organizations.
§ General purpose/campus switching customers don’t need ANET’s customizability (and probably lack the engineering expertise to exploit it).
§ Blade switches are used to communicate between line cards in the same server which isn’t something ANET switches are geared towards.
o Similarly, VAR suggests that organizations that are deeply penetrated w CSCO (which is many rank and file enterprises) are unlikely to utilize ANET for incremental switches since the headache of managing two systems is not worthwhile.
§ CSCO’s incumbency advantage as well as the fact that they offer a full suite of services (storage, firewalls etc) means there are parts of switch TAM that ANET won’t be able to penetrate.
· Competitive dynamics intensifying across several vectors.
o At high end, internet content providers (“ICPs” which represent ~25% of ANET sales) like GOOG, MSFT and FB are increasingly embracing whitebox switching solutions. These companies have the engineering talent to develop switch operating systems that are optimized specifically for their needs (and they obviously save money relative to buying ANETs OS).
§ Note MSFT is ANET’s largest customer which utilizes ANET switches their Azure cloud environment and they have recently moved to develop their own Linux-based switch called Azure Could Switch to allow for the flexibility and scalability they need.
· While the shift of networks into the cloud was initially positive for ANET since it empowered sophisticated providers, in the long run it may ultimately make them more vulnerable to whiteboxing since the cloud networking providers are adept enough to design their own OS.
· To that point, whiteboxing represents ~7% of 10G port shipments v 23% of 40G port shipments which speaks to the fact that customers with more robust needs also tend to be more sophisticated (and therefor capable of whiteboxing)
§ Whiteboxing currently represents ~8% of data center switching market but that figure is growing with a possible inflection point in uptake in the near/medium term. Per GS Survey this figure is expected to grow:
o CSCO is a deep pocketed competitor and they have declared DC switching to be a “core market” for them (on a call introducing a new line of switches intended to target some ANET products, John Chambers called the products “Arista killers”).
§ CSCO has a history of crushing peers and as the biggest and best capitalized Company in the space, they are a dangerous competitor.
§ VAR contacts responded that “nobody gets fired for buying CSCO. An IT person needs to have a very good reason for taking a chance on ANET.”
§ In response to ANET’s success with merchant silicon, CSCO now offers a merchant silicon option in their Nexus 9000 line (actually using the same chip as in ANET’s comparable product).
§ CSCO management has talked about successfully winning back major Web 2.0 customers with this product and VAR indicates they have also been pricing this product at 50+% discounts with aggressive bundling strategies.
§ CSCO doesn’t cede share easily (and VAR indicates that CSCO has been offering more aggressive price concessions in certain scenarios). In the broader switching market, no competitor has ever gotten more than 11% share (HP got to ~11% of general purpose switch market but CSCO dominates that with ~65% share)
o The Company is also facing increased pressure from new entrants.
§ JNPR recently entered into DC switching market with their QFX10000 family which is ramping in Q3-15-Q116 and allows the Co to address a segment of the market that they hadn’t previously addressed.
· VAR with switching customers indicate JNPR now forcing their way into the conversation on some new RFPs.
§ In the ultra-high performance segment (100G) HP and BRCD have recently announced plans to introduce a new product in that premium segment that had essentially been a 2 player market (CSCO/ANET)
o Market becoming increasingly commoditized
§ Dell recently introduced an innovative new offering that VAR suggests is gaining traction. Dell will sell a commoditized Dell switch and allow customers to run one of several OS’s on it. So a customer could run a Dell OS, a JNPR OS, or one of several other different options. Dell is embracing the commoditization of switching hardware which plays to their strength of scale and could be problematic for ANET as this would introduce an option ~half the cost of ANETs switches.
§ Smaller (but well capitalized) players like Cumulus will offer customers Linux based switches. Companies like this are designed to enable whiteboxing by lowering procurement costs and offering their software separate from any hardware.
· Cumulus switches priced ~75% below ANET
§ There is also a distinct trend towards Software Defined Networking (“SDN”) which effectively disaggregates software and hardware in switching. At a very basic level this approach entails using “dumb” switches that only have very basic software installed and having many of the configuration decisions and instructions decided centrally and continuously pushing out instructions to the “zombie” switches.
· Disaggregating hardware and software should lead to increased competition and lower priced switches.
· VAR with major switch purchasers characterized “SDN as one of the 3 biggest trends in the networking space.”
o VMWare and HP are the Company’s two most important channel partners. HP has plans to introduce their own switches and VMware was recently acquired by Dell who also makes their own switches.
§ VAR indicates Dell’s sales efforts in switches have gotten more aggressive past few quarters which suggests they are focused on this market. Reducing ANET’s ability to leverage VMW channels would be a material problem for ANET.
· Margins will compress as they saturate core TAM and need to address adjacencies to grow
o ANET has converted the sophisticated players who know what they are doing-they now need to move beyond low hanging fruit which is going to be harder. They are trying to accomplish this by expanding internationally which will bring continued margin pressure
§ Ramping up international growth will not be cheap. For context, ANET reports ~60 reseller relationships worldwide (v almost 800 domestically). CSCO has nearly 70,000 global resellers. Expanding their reselling infrastructure internationally is going to lead to increased costs and margin compression.
§ In Q3 ANET reported 22% of revenue from international markets but international revenue growth slowed to 6.5% sequentially v 13% for the domestic business.
o Co is effectively overearning v their self-determined model as well as peers with GMs ~66% vs model of 60-65% (which is where peers also are as per below).
· Valuation implications unrealistic
o ANET has ~8% share of data-center switching market and a market cap equal to ~60% of that market.
o Utilizing market share data as discussed earlier as well as ANET’s own margin goals, we can approximate market share required to justify current share price. Assuming a 20x multiple and the top of ANET’s margin range, the Co would need to capture +1/3 of its TAM which is an improbable task.
§ This would represent a near quadrupling of ANETs market share despite a market that is getting more competitive and one in which they have lost their first mover advantage-this is highly improbable.
|Realizable TAM (MM)||$3,003||$3,484||$3,770||$4,082||$4,329|
|EBIT Multiple||17.5x||17.5x||17.5x||17.5x||17.5x||Implies ~25x P/E|
|Required Earnings Power/Share||$4.46||$4.46||$4.46||$4.46||$4.46|
|Required Earnings Power (MM$)||$305||$319||$323||$327||$340|
|Operating Margin||23%||23%||23%||23%||23%||At top of cos 18-23% target|
|Implied Market Share||40%||37%||35%||34%|
· Legal threat overhang with two pending lawsuits that could materially impair their franchise (and at the very least will result in real, cash legal expenses in interim). Losing these cases could be catastrophic for the Company (and prelimary rulings in some have been decidedly against the Company)
o ANET being sued by co-founder (and second largest shareholder) David Cheriton for breach of contract alleging that ANET is illegally using technology that had been licensed to them by Optumsoft (which is a private Company he founded).
§ Basic issue here is a software development tool called TACC that ANET licensed from Optumsoft. The license was royalty-free but Optumsoft maintained the ownership of “improvements, corrections or modifications (to it)” as well as “any derivative works therof” and Optumsoft argues that ANET’s EOS should be characterized as an improvement to the original TACC code.
o ANET is being sued by CSCO for several counts of patent infringement and copyright violations. Note a large portion of ANET’s founding team/C-suite came from CSCO (founders, CEO, CTO, CDO and several SVPs). Also note that VAR with customers indicates that ANETs manuals use the exact same language/terminology (including typos).
o There are two parallel actions here-one at a District Court which seeks monetary damages and another at ITC which seeks to bar ANET from importing its products into the US (where 85+% revenue is earned). This dual track is suggestive of CSCOs “all guns blazing” approach. Also, note that CSCO doesn’t have a reputation for filing frivolous lawsuits. Note the claim for monetary damages has a 6-year lookback so would be +$2Bn but, as importantly, would impair ANET’s business going forward.
§ After hearing arguments, ITC staff presented their position that ANET infringed on 3 of the 5 patents being litigated in one track of ITC litigation. While the staff recommendation has no formal impact on the ruling, they do tend to influence the judge’s ultimate decision.
§ CSCO’s claim essentially asserts that ANET has essentially replicated their own command-line interface (“CLI”) which is used by customers to communicate with the switches as well as manage and configure them. They also allege violation of patents that, at a basic layman level, improve performance of a network and increase resiliency of switching devise.
§ CSCO’s CLI is well regarded and given CSCO’s size is something that many engineers are familiar with. ANET’s very similar CLI makes switching costs lower since it doesn’t require an engineer to learn a new coding language.
§ ANETs founders were named on two of CSCOs asserted patents which means it will be very difficult for ANET to deny they did not know about CSCOs patents which means that the likelihood of “willful infringement” is raised and under patent law, a willful infringer can be subject to treble damages.
o Company will spend +$40MM in legal expenses this year (after which these costs should moderate as trials move closer to verdicts). While they obviously suggest pro-formaing this out, this is a material cash expense.
o Cases at ITC expected to be complete by Q3/4 2016. District Court cases proceed throughout ’16.
· Consistent insider selling. There have been no material insider buys in past 12 months
· Their focus on the OS and highly customizable software means enables them to be a structural winner
· They are able to successfully rebuff competition from whitebox threat
· CSCO (or another party) acquires them
Signposts/things to monitor:
· Share shift (especially relative to CSCO and among marquee customers)
· Margin progression in the face of presumed sales ramp
· Success or failure of whitebox products
· New product announcements and changes in pricing
|Company Name||TEV/EBITDA LTM - Latest||LTM Levered Free Cash Flow Margin %||Est. Annual EPS Growth - 1 Yr %||Est. Annual Revenue Growth - 1 Yr %||Est. Annual EBITDA Growth - 1 Yr %||NTM Forward P/E||NTM TEV/Forward Total Revenue|
|Palo Alto Networks||NM||26.07%||107.2||43.6||109.5||95.92x||11.37x|
Low hanging fruit has been picked in terms of market penetration. IDC's moving to whitebox combined with more competition at lower end as well as increased marketing/legal spend will lead to earnings shortfalls which will hurt the stock given juicy multiple.
|Subject||The outside view|
|Entry||01/05/2016 09:53 AM|
Why do you think Bechtolsheim, a founder and significant shareholder hasn't sold a single share? Especially in light of the fact that he was the guy who headed up the group at Cisco (disputed patents) that left to form the core at Arista.
Do you think your 2016 estimate for revenue growth is modest given that Q3 revenue was $220 million with 40% YoY growth?
Why didn't you talk about the opportunity in routing ($4 billion TAM) where they are introducing a product (one of their customers has already programmed EOS to perform routing) and monitoring ($3 billion TAM)?
Microsoft has already shrunk from a 22% customer to a 9% customer so is this really a significant headwind? Especially given that they are still a customer and expect Azure to grow 150% in the next 2.5 years but have largely pre-built for current growth and will need to reinvest (see Azure margin progression)?
Why hasn't whitebox become a more significant product already with only about $200 million in revenue? Does any of the competition offer and integrated solution as ANET does with VMW/ Nicira and others? Isn't F5 at more risk of being disrupted (similar margins to ANET) given this integration? Could you make an argument that Google, Amazon and Facebook might have a bigger payoff implementing EOS and focusing on their core competencies rather than commiting internal resources to proprietary networking solution?
|Subject||Re: The outside view|
|Entry||01/07/2016 07:17 PM|
Good write-up, bdad. Afgtt2008 asks a lot of good questions, many of which I was going to ask.
At a high level, the hyper-scale cloud titans (GOOG, FB, MSFT, etc.) are the most meaningful portion of ANET's revenue, and all have voiced plans to significantly grow their data center switching capex to facilitate bandwidth/ workload capacity build-outs. I have done a significant amount of work trying to mosaic how these cloud titans are planning to ramp up/down their usage of ANET's EOS vs. the whitebox + internal software engineering alternative. All mention the appeal of the eventual cost savings they hope to realize with whitebox solutions and the more integrated/faster-lead time development process that utilizing an in-house engineering team will enable. However, they also have said they cannot justify moving to white-box solutions overnight -- they cite massive operational risk as the primary barrier.
Data center switching downtime can be incredibly costly, and the trouble-shooting of networking hardware issues is incredibly complex. White-box vendors typically do not provide the networking support staff necessary to trouble-shoot potential DC switching issues at customers that use their white-box hardware, especially integration issues with the cloud titan's internally-developed software, which the third-party white-box support staff would likely not be familiar with given the cloud titans' secretive nature.
My take-away from conversations with numerous consultants, switching vendors, and VARs was: yes, white-box is a real threat to ANET/CSCO, but most cloud titans' data center switching capex is not yet burdensome enough to justify the cost savings (and commensurate operational risk) that migrating to a white-box vendor would bring them.
Absent a near-term catalytic comment or mosaic on these cloud-titans migrating off of ANET to CSCO or a white-box vendor, I don't understand how you can get comfortable shorting ANET right now. Further, with ~20% short interest and the buyside bogey of ~$3 2016 EPS x 30 P/E = $90 target price still alive, it seems there's a real risk of a squeeze, and the near-term risk/reward seems to skew more upward than downward. Maybe it will be impossible to know what the cloud titans will do until it's done and already reflected in ANET's stock price, but I have found enough disconfirming evidence to hold off shorting ANET at this point.
|Subject||Re: Re: The outside view|
|Entry||01/09/2016 06:26 PM|
I second (third) the concerns. Of course, stock already down from $78 to $68, albeit in a bad market.
The writeup said something like "no one ever got fired for buying Cisco" which sounds like "no one ever got fired for buying IBM"
Also, at one point it says it´s easier to consoldiate around one vendor for ease of use, at another point it says ANET is similar to operate vs. Cisco (in reality many of the switches are similar to operate).
I´m familiar with the industry, having commented on the Cisco writeups, and think Cisco is a lumbering giant well past its prime while ANET is much more of a startup. I saw Andy B. at an ANET conference booth a few years back - imagine the chairman manning an industry conference booth? They seem much more driven. Cisco has so, so much more fat and is (often) very high priced that I would much rather be long a potential Cisco killer than short.
|Subject||Long thoughts on a short idea|
|Entry||09/22/2016 01:13 PM|
With the introduction of the 7500R and it's variants combined with the move to 40 G & 100 G switches isn't there a reasonable chance that growth accelerates over the next year? I realize the in network opportunity for routing is small at ~ $500 million but with Netflix as a reference client you're bound to get accelerated uptake. Then you get data centre interconnect expanding the routing TAM to $2 billion early 2017 and service provider WAN routing late 2017 moving the TAM to > $4 billion.