Arista (in Greek means “to be the best”) develops software and hardware switches for cloud networking solutions to address the needs of large-scale cloud service providers and next-generation enterprise data centers. To put this in lay terms, what this means is whenever you visit a website, whether it is downloading a movie from Netflix or checking your Facebook account, there is a data center in the background that is powering it. Inside a data center, which is made up of thousands of servers, there are switches which are the engines that deliver the content, such as pushing the movies from Netflix out of the door or enabling the applications you use on your smartphone. Arista is in the business of providing these hardware switches and software so customers can operate their data center networks and deliver content to their customers. Arista performs an extremely important function because the data center infrastructure is arguably the most mission critical part of a web based business. If the network goes down, you simply are not in business anymore as your customers or employees cannot access your website.
Even though Arista appears to sell hardware switches, it really is a software provider. In fact it is fast becoming the Microsoft of networking software. Today, six of the seven largest cloud services providers (excluding Google) based on revenue such as Amazon, eBay, Facebook, Microsoft, Netflix use Arista.
One of the main reasons we became interested in Arista is that 75 percent of all trades that happen on U.S. exchanges go through an Arista product. Arista is considered the gold standard on Wall Street for trading infrastructure. Some of Arista’s first customers were high frequency traders who demanded high reliability and required trades be completed in nanoseconds and milliseconds rather than seconds. Our thought was that if the most demanding customers on Wall Street were using Arista’s products instead of competitors’ products then Arista’s products must be superior.
Problems Arista Is Solving
Arista was founded in 2004 to address the limitations of legacy networking products and to create a cloud networking platform that is open and programmable. The key word here is “open” because most legacy network products, such as those offered by market share leader Cisco, hold companies hostage by forcing customers to buy proprietary products from them. Arista does not “lock-in” its customers to a proprietary system but instead allows customers to customize the software to control their own data network and to use best-in-class providers for security and other data center solutions. For example, customers can easily integrate third-party applications such as VMware which provides virtualization software that makes data centers more efficient by spreading work around servers. By comparison, products offered by Cisco are not as programmable and as a result are extremely difficult to integrate with third-party applications. This lack of integration forces customers to continue to either rely on time consuming, error-prone, and costly manual processes, or forgo third-party integration entirely.
One of the main benefits of using open standards is it makes Arista less susceptible to technology obsolescence and places it in the position where it benefits from advances in technology. It also gives Arista the advantage of being able to bring products to market in far less time compared to the competition which has to figure out how to configure its legacy products to work with third-party applications.
Arista’s EOS software (Extensible Operating System) solves two fundamental issues that exist in cloud networks: the need for non-stop availability (network outages in the cloud are very expensive in terms of customer impact) and the need for high feature velocity (no one likes a slow connection). In contrast, legacy cloud network providers often encounter the Christmas tree light problem where if one light goes down (in our case one server), the whole Christmas tree light chain goes away (the whole network goes down). Arista’s software is able to isolate the server that fails from the entire network and can even deliver new updates to the entire installed base of servers with minimal disruptions to the network.
Arista helps customers eliminate many costs by automating most of the functions that previously required human interaction. This is also important because human error is the leading cause of data center downtime. For example, it reduces the extremely high costs of maintaining the switching equipment which typically are $3 to $5 to maintain for every dollar in Capex a customer spends. Customers also no longer need to pay people to manually input commands for every switch they install which can translate into savings of $300 per server or a 10 to 30 times reduction in operational costs.
So far Arista’s technological advantages are working as it has been taking market share away from Cisco who historically dominated the data center market, holding a greater than 80 percent share in switches for many years.
Below is a chart showing Arista’s market share compared to Cisco:
2011 2012 2013 2014 1H2015 Artista 3.4% 4.9% 6.7% 9.3% 11.8% Cisco 73.3% 71.4% 70.8% 66.1% 61.5%
Source: Crehan Research Datacenter Switch Market Share Report Q2 2015.
The main competitive advantage Arista has is that its EOS software architecture was built from scratch specifically for businesses operating in the cloud. To build its software Arista spent ten years of research and development writing 10 million lines of code.
Before Arista, no one in the networking industry tried to create software specifically for the cloud networking world. Most competitors kept their legacy software architecture and then kept adding on new features to an existing platform to adapt to cloud computing. Because competitors did not build in programmability for the cloud from the beginning, they are stuck with their legacy systems and are limited in the features they can add. Without building a new foundation, the legacy providers are weakening their product every time they add new layers, just as you would weaken a house by building multiple stories on a foundation that was designed for a single story.
Arista was co-founded by Andy Bechtolsheim (59 years old and owns 19.6 percent of the company) who spent his entire career in the technology industry and started off specializing in hardware for high-speed networking. He was one of the founders of Sun Microsystems and later founded Granite Systems which was acquired by Cisco. Bechtolsheim ended up working as a vice president at Cisco for many years and then left Cisco to start Arista in 2004. From 2004 to 2008, the co-founders worked on building the EOS software from scratch, spending $100 million of their own capital before they shipped out their first product in 2008 to Microsoft. We believe it is rare to find a business where the founders spend their own money perfecting a product for four years before rolling it out. They were in a position to do this because Bechtolsheim was one of the first two people to invest in Google, writing them a $100,000 check in 1998.
Because Arista was founded by engineers with deep technical expertise, they focus on technology and engineering staff, and invest on average 25 percent of revenues in research and development ($150 million in 2014). They spend very little on sales and marketing, preferring to let word of mouth from customers sell their products. Out of 1,000 employees 700 are engineers, with 90 percent of them working on software development and 10 percent working on hardware. This focus makes sense because Arista’s highly technical products are mostly sold to other engineers.
From the beginning the co-founders focused on building a strong and long-lasting employee culture. The goal was to build a workplace where engineers would love to work. So far it appears these efforts have worked as Arista was named as one of the top places to work by Glassdoor and employee attrition has been in the low single digits.
Most of the executive team at both the senior management and vice president levels has worked together for an average of 15 years at Arista and other companies. For example, co-founder Kenneth Duda (43 years old and Chief Technology Officer) worked at Granite Systems with Bechtolsheim. CEO Jayshree Ullal (54 years old and owner of 11.02 percent of the company) previously worked at Cisco alongside the co-founders.
We believe Arista’s software is still in the early stages of adoption and has significant potential for growth as there is currently a massive secular transformation in data centers moving to the cloud and replacing legacy network technologies. We are therefore investing in the early innings of Arista’s growth story. We base this on the expectation that the cloud network data center industry is expected to nearly double in size to $13 billion by 2019, compared to $7 billion in 2013, primarily driven by an increase in global internet traffic (source: Crehan Research).
Arista is currently focused on its existing customer base of large scale Internet companies which includes software as a service companies, cloud hosting service providers, and financial services organizations. We believe Arista’s conservative strategy of focusing near-term growth within its existing customer base increases its odds of reaching its long-term potential. These three verticals alone represent revenue potential of $2 billion, compared to the $830 million in estimated revenues Arista is slated to earn in 2015.
We are confident Arista will be able to achieve this revenue growth because its track record demonstrates that once customers purchase a product they will continue to purchase products. As of December 31, 2014, end customers that have been with Arista for at least 2 years, on average made additional purchases that were approximately 11.9 times more than the initial dollar amount of purchases made in the first two quarters of their purchasing history. Furthermore, the pace of adoption of Arista’s technology is also increasing. As evidence of this, it took four years for Arista to sell one million ports. To sell the fifth million port took one-tenth the time. One of the main reasons we are paying what we believe is a fair price for Arista is because we believe both the durability and pace of growth will be sustainable.
At our average cost basis of $64.43 per share we paid an enterprise value of roughly $3.8 billion. We expect free cash flows to exceed $75 million (excluding the benefit of stock-based compensation to cash flows) in 2015 which means we are paying a 2 percent enterprise value to free cash flow yield. Using another valuation metric based on 2015’s expected revenue run rate of $830 million, we paid 4.6 enterprise value to sales. By traditional value investing metrics Arista is overvalued. We are comfortable paying a premium to that measure of value because we expect growth to materialize for many of the reasons outlined above. Our downside is protected by the strong balance sheet with $567 million in cash and $42 million in debt. We also believe Arista would achieve a high selling price in a private sale as it has essentially become the dominant provider of network software (similar to Microsoft’s domination of operating systems) in what is one of the fastest growing markets in technology.
Reason Stock Price Has Decreased Is Due to Cisco Filing a Lawsuit
The main reason the stock price declined during the year is because in December 2014 Cisco filed two lawsuits in the Northern District of California for alleged patent and copyright infringement. Additionally, Cisco filed a complaint with the International Trade Commission for patent infringement. What Cisco is asserting in its lawsuit is that it basically has a copyright on the computer language Arista used to develop its software. We believe the lawsuit is frivolous as well as a temporary distraction. First off, Cisco’s patents are so outdated that Arista was inspired to create its own intellectual property, and created a new programming model from scratch, writing every line of code themselves. In addition, Cisco is claiming that it has a trademark on the computer language Arista used which was in place before Cisco even existed. This is similar to someone claiming they have a patent on the English language.
We believe the lawsuit is a publicity stunt from Cisco devised to scare customers away from using Arista and to slow down Arista from taking so much market share from Cisco. The reason we believe this is that five days before Arista was formally served with the lawsuit Cisco filed a press release announcing the lawsuit. In other words, a company that talks to the press before filing a lawsuit obviously has more interest in marketing or press coverage than legal matters.
One of the reasons we have faith in Arista’s leaders is they are protecting themselves against an unknown legal outcome by developing alternative solutions to the patents in question to make sure customers aren’t impacted. In other words, they are leaving nothing to chance.
This post has been excerpted from a letter to partners of Compound Money Fund, LP.
This document is for informational purposes only. It is intended only for the person to whom it has been delivered. This document is confidential and may not be distributed without the express consent of Time Value of Money, LP. The information contained herein is subject to change; however, we are under no obligation to amend or supplement this document. This document is not intended to constitute legal, tax, or accounting advice or investment recommendations. This document shall not constitute an offer to sell or the solicitation of an offer to buy nor shall there be any sale of a security in any jurisdiction where such offer, solicitation or sale would be unlawful prior to issuance of an offering memorandum. An investment in Compound Money Fund, LP involves a substantial amount of risk. Investments should only be made by investors who fully understand these risks and can withstand a loss of their entire investment.