ANET – The Value Investors

      Comments Off on ANET – The Value Investors
ANET – The Value Investors


by Alaedine Belhouari


– Positive –


Outlook Positive Upside (2025) 143.8%
Current Price (07/03/2020) $210.51 Current Valuation $15.937B
Target Price (2025) $513.12 Target Valuation (2025) $34.166B

Summary: ANET makes the best shovels in the cloud gold rush

Cloud services are a gold rush. Network switchers and routers are the shovels in this gold rush. Arista Networks builds the best shovels and gold rushers know it.

The Cloud Networking industry is growing, and ANET is gaining market share every year.

ANET is a great business with the best management team I have seen in my (short) investment career.

They are increasingly becoming a service provider rather than a hardware provider, making it a better business with more predictable, sustainable and growing cash flows.

Lower revenue (and earnings) in 2020Q1 and expected lower numbers in 2020Q2 and Q3 because of the virus cause mispricing.

Low trading volume might cause many institutional investors to have missed it in their screening, causing further mispricing.

Stock Price (Last 60 Months)

Source: Bloomberg

Business Overview

Arista Networks designs network switches, routers and associated software for large enterprises. In 2004, Google requested from the general market a hundred thousand switches at 1GB for 100 dollars per switch. Arista was the first to make that. Today, Google is still one of Arista’s clients amongst others in industries spanning from healthcare to finance to entertainment. The most notable being Microsoft, Facebook and Amazon, eBay, Yahoo!, Barclays, Citigroup, Morgan Stanley, AOL, Comcast, Equinix, ESPN, Netflix.

ANET has two main revenue segments today; products and services. The products consist of switchers and routers, and services consist of post-consumer support. Geographically speaking, 75% of its revenue comes from the US, 15% from Europe and 10% from Asia.

It is important to note that arista has shown rapid revenue growth since day one, and it’s been profitable and cash flow positive since 2010.


ANET’s founders and top three executives say a lot about the company. They are great managers, inventors, and investors. Insiders own 44% of the company, which shows a great alignment of interest.

The company’s CEO, Jayshree Ullal, is an electrical engineer with over 30 years of networking expertise from silicon to system companies. She spent 15 years at Cisco, where she grew their Catalyst switching division from its beginning in 1993 to a $5 billion revenue segment in 2000. She also oversaw more than 20 M&A deals for Cisco in the enterprising sector. She is worth about $1.2B today, mostly from ANET’s stock.

Andy Bechtolsheim, Arista’s CDO, is also an electrical engineer and he founded three notable companies before Arista. He first founded Sun Microsystems, which was acquired by Oracle. He then founded Granite Systems, which was acquired by Cisco where he became vice president and general manager of their Gigabit Systems Business Unit for 8 years. He then founded Kealia, which was acquired by Sun Microsystems. Today, he personally owns around 16% of ANET, and an additional 16% through his family office. He was also Google’s first outside investor before it incorporated. He is worth about $7B today.

Last but not least is Kenneth Duda, Arista’s CTO. He holds three simultaneous engineering degrees from MIT and a PhD in computer science from Stanford. He is a pioneer in the industry and he co-authored multiple research papers for Microsoft and Vmware. He is the lead architect of Arista’s Extensible Operating System (“EOS”), the proprietary software at the core of Arista’s products. He is worth about $150mm.

The gold rush

Many companies have switched from inhouse data centers to offsite data centers, mostly because private and public clouds are cheaper and easier to set up than legacy data centers. The largest providers per market share for such a service are respectively Amazon (~30%), Microsoft (~15%), Google (~10%), IBM (~10%) and Alibaba (~5%). In order for data to circulate between companies and data centers, but also within the company’s operating system and within the data center’s infrastructure, a network is needed. Just like legacy networks were used for legacy data centers, cloud networks exist for cloud data centers, with aggregate cloud network bandwidth being able to exceed 1 Petabit/second orders of magnitude higher than legacy data center networks. The network is built on switchers and routers, and it is operated through the network’s provider’s operating system. That network can be more reliable, more secure, faster, easier to set up, and easier to expand, depending on how it is designed. ANET seems to sell the most performing switchers and routers, with the best operating system in the industry.

Today we are in the middle of a transformation in the ways people think of technology. Legacy networking that was built on inhouse software and hardware has evolved into a huge cloud networking opportunity – a real gold rush. Tomorrow’s business, including extensions to current operations as well as opportunities such as online data storage, artificial intelligence, entertainment and game streaming will all depend on cloud computing. That’s why spending on cloud and next-generation data centers increases much faster than traditional legacy IT.

Some of the benefits driving the change are that legacy IT provides inferior bandwidth and it lacks integration which forces customers to rely on a time consuming and error-prone process that may be cost-prohibitive. The cloud’s integration with software providers allow for better firewalls and monitoring which enhance both security and visibility. With cloud networks, you can have a firewall between each terminal in your company’s network infrastructure, while it would have cost a fortune with a legacy network. It is more secure because a hacker would need to break a firewall for each terminal before being able to access it, instead of only breaking one firewall and having access to every terminal in the network.

The advantages of cloud computing also come with challenges for cloud networking. Its scale requires data centers to be built with structural capacity, reliability and accessibility in mind. Today, only the biggest players in cloud computing are doing this right, which can be seen in their market share. 

As the rate of hardware improvements are slowly starting to saturate while still being high, the largest advantage to be had in the future will be in software optimization.

The best shovel

For the cloud to exist and grow, it needs to be built and connected. As mentioned earlier, the network consists of switchers and routers and their associated software. To reiterate the analogy, if cloud data centers are the gold rush, then the switchers and routers needed to build the network are the shovel, and Arista makes the best shovels.

Customers’ Cost Efficiency

Arista’s hardware is power efficient, resulting in operating cost efficiency for clients. Power costs $0.03/KWh (PNW) to $0.30/KWh (EU). Arista’s 10GB Ethernet 7050, 7100, and 7500 family can reduce total power consumption compared to competition (according to their website), saving millions of dollar in energy expenses for customers.

Leading R&D

R&D spending is growing almost as fast as sales, which has created an “Arista Advantage” leading the industry’s evolution. While operating expenses margins have decreased overall, R&D margins have decreased the least, reaching ~19.2% in 2019, which is above average in the industry. While it is hard to measure the return on R&D expenses, the profitability ratios suggest that those returns are high (see “Capital allocation” below).

Significant Technology Lead

Arista’s extensible operating system (EOS), integrated into all their hardware, might be their biggest advantage. It allows for more integration, security, speed, through reduced latency and down time, reliability and all while being able to scale.

ANET is recognized in many industries as being the best provider in the industry, and it was just recognized by STAC (Securities Technology Analysis Center) last month as having developed the most performant and reliable devices for high frequency trading in the industry.

Easy and Cheap to Switch to Arista:

ANET’s products are fully programmable, modular and resilient. This is a main driver in acquiring customers that recognize Arista’s advantage. For example, if Alibaba is using products from Juniper Networks (one of Arista’s competitors), and Alibaba wanted to scale up their data center, they could keep their existing Juniper products and add Arista’s switchers and routers, which can function within the existing Juniper network, using Arista’s EOS to manage the whole network. This makes migration easy and cheap for customers. The other way around is not possible, because Juniper’s software (and this is the case for all of Arista’s competitors) was initially programmed not to be adaptable to products from other network providers. Arista’s competitors would need to build an entirely new operating system if they wanted to make it as programmable and adaptable as Arista’s, which would cost them hundreds of millions of dollars and a number of years. Arista’s EOS allows rapid delivery of new features while preserving structural integrity and quality.

Top Customer Service

Customer service is a very important aspect of cloud networking providers, since modifications to a customer’s network need to be made with the help of the provider. I called ANET’s customer service and I was connected with a representative within a minute, I looked into a number of forums and the wait time was always around a minute. Customers reviews said that they were always connected in under 10 minutes with an Arista engineer.

Catering to customer demands

Arista’s team knows their customers very well. In a lot of cases, they identify areas to improve within large companies’ network, and then they build the corresponding products that they directly sell to them. For example, they created a specific spine and leaf architecture for Microsoft, leading to a long term partnership.

Growth drivers

At the industry level, cloud computing is commoditized so data centers need to have the best infrastructure among their competitors if they want to retain their customers. As a result, they will have to keep buying the best technology every time there is a better router or switcher on the market, to have the fastest and most reliable data stream. If Arista keeps being the best innovator, there will always be demand for their cutting edge products. For example, Microsoft spent 79mm at Arista in 2013, and spent more every year, with 554mm spent in 2019. While I could not find numbers for all data centers, Facebook spent $410mm in 2019.

Ethernet switches and routers sales have been growing by 4 to 5% per year on average since 2012, with no slowing down trend. And now, within this growing market, Arista has grown its market share from ~1.5% in 2013 to ~7% in 2019, with only Huawei growing its market share similarly and Cisco’s declining from 65% to 55%.

Source: Statista

Note that the above numbers refer to switches for any type of customers, not only data centers. When looking at the data center switching market specifically, Arista beats Cisco even more clearly. Cisco’s largest market is for data centers, and Arista has been beating them consistently. Other players have eaten market share away from Cisco, but Arista and Huawei are the ones that hurt Cisco the most.

Source: Crehan Research

In addition to the data center market, Arista is expanding its TAM by entering two new markets: campus and enterprise networks. They started focusing on spreading to those two markets in 2019 and they have increased their sales and marketing expenses. It will take until 2021 or 2022 to see a large effect on their revenue because of how the industry and these two market specifically operate. Usually, the customer would purchase a small amount of products and test them for a year or two, before scaling up and building a full network.

Arista’s service segment has compounding revenues, since both new and existing customers keep paying for the maintenance and support service every year after buying their products. Customers have no contractual obligation to do so, but they need to if they want to keep using their Arista network at its full potential.


Arista has been the fastest growing company in the last 4 years in terms of revenue with a 28.8% CAGR, despite not being the smallest company. They have the highest operating and net margins, and the second highest gross margin. VMware has higher gross margins for two reasons. First, they sell network software which has higher gross margins, while Arista sells both software and hardware, averaging its gross margin down. Second, Arista outsources the manufacturing of their products, which costs them more than if they had their own manufacturing facilities and their own workers.

ANET has a high R&D margin and they are the only one with no debt. As mentioned earlier, it is hard to assess how good they are at spending in R&D, but their profitability ratios suggest that they are doing a good job.

Arista’s current PE ratio is lower than its comps’ average, and it is much lower than its historical average. While it is not a very cheap stock, its current price is a good price for a great company.

ANET – The Value Investors 16
ANET – The Value Investors 17

Capital allocation

Arista’s management have proven to be excellent investors and capital allocators.

First, they do not pay dividends, and they have no intention to do so for now. Dividends are very tax inefficient for shareholders, and I would rather see their management investing 100% of that money than sending us back 60 or 70% of it to invest on our own. They have the highest ROA (21.65%), ROIC (47.51%) and ROCE (30.35%) in the industry, and a high ROE (31.24%) but below its comps’ average, mostly because Arista has no debt to make its equity smaller.

They have invested in three cutting edge startups that focus on the software part of their system, which is their greatest asset. Focusing on their strengths is what is going to let them stay at the top of the industry:

  • Big Switch Networks was acquired in February 2020: they develop software driven by SDN technology, in which groups of networking switches and access devices can be virtualized and turned from proprietary hardware to software. This is how they will compete even more with VMware.
  • Metamako was acquired in September 2018: they provide low-latency, FPGA-enabled network solutions.
  • Mojo was acquired in August 2018: they are the leader in cloud-managed wireless networking.

ANET purchased back shares in the last 4 quarters, when multiples were the lowest and they want to keep taking advantage of price volatility in the future.

Source: Marketwatch

ANET’s stock has outperformed both the S&P500, and its comps on a 5-year basis. While it does not mean much, it reinforces the thesis for good capital allocation.

screen shot 2020 07 06 at 7.19.05 am


The first reason ANET could be undervalued is probably its industry. It takes some time to understand the cloud networking industry and its role within the new order of the internet. If an investor doesn’t see the “shovel in a gold rush” situation, and does not appreciate Arista’s technological edge, it might not seem like a great opportunity.

Another reason might be COVID-19 and its effects on Arista’s supply chain. Arista outsources the manufacturing of a large part of its products to plants in Asia, which were not able to operate at full capacity in 2020Q1. It caused a significant part of what could have been revenue to become deferred revenue. As the virus spread from China, transportation became trickier and even fewer orders were fulfilled. Q1 recorded lower sales, and Q2 and Q3 are also expected to see revenue decreasing for the same reasons, but also because customers are expected to reduce or delay their network expenditures to prioritize other expenses.

Finally, ANET has a very low trading volume which might cause institutional investors to miss it in their quantitative screening. As mentioned earlier, 44% of Arista is owned by investors, and an additional 15% is owned by reputable long term investors, which might explain the low volume. ANET’s 90-day average daily volume is $867,957, while its comps’ mean is $8,663,199 or 10 times higher. Extreme Networks (which is 30 times smaller in terms of market cap) has the second smallest daily volume with $1,756,891 or double the one of Arista.


I segmented revenue into Products and Services:

  • Sales mostly depend on customer capex cycles so product revenue is projected using industry data, market share penetration, immersion into new markets.
  • Service revenue compounds because all active customers use it so service revenue is projected using an estimated retention rate, past service revenue, past product revenue, and current product revenue.
ANET – The Value Investors 18

As revenue from services compounds and revenue growth from products slows down, a larger part of the total revenue comes from services. Services also have historically seen their profit margin increase as they scale, and it should still be the case in the future. As a result an even larger share of profit will come from services compared to the total income. This will make a larger part of the cash flow sustainable and predictable.

ANET – The Value Investors 19
ANET – The Value Investors 20


ANET – The Value Investors 21

I calculated target prices for three timelines: 3 years, 5 years and 10 years. I did a sum of the parts valuation of the company between the product segment and the service segment. The service segment has a higher multiple because cash flows are predictable and can scale faster than the product segment. I used three sets of multiples depending on the market cycle, and I took the average for the final target price. Note that the overall multiples I use for the valuation are much lower than Arista’s historical average of over 30, and on par with the industry.

DCF Sanity Check


I did a DCF to check that the results weren’t too different, and while it projects lower upside, the returns would still be satisfying. I think that the sum of the parts is a more accurate way of valuing the company.


Like for any great investment, there are risks.

Supply Chain: Arista buys components from Intel, Broadcom and outsources the manufacturing of their products to other companies which gives them little control over their COGS and delivery schedule.

Revenue Concentration: Large parts of their revenue come from a few clients causing a lack of pricing power, and therefore lower margins as large customers can negotiate the price down. For example, in 2019, Microsoft represented 23% of Arista’s sales and Facebook 17%, and they were able to get better prices.

Revenue Unpredictability: Sales are unpredictable but Arista stays very liquid and generates returns on its cash by purchasing high quality marketable securities.

Large Cash Balance: As of 2020Q1, Arista had about $1.1B in cash & equivalents, and $1.7B in marketable securities (mostly high grade bonds). While I don’t like dividends because they are tax inefficient, holding a large cash balance generating at most 4% per year might be even less efficient for Arista. On the other hand, having this much cash makes ANET a very safe company in the current situation.

Competition: Huawei has also made great innovations in the cloud networking industry, but Arista generates 75% of its revenue from the US, where Huawei cannot compete for the time being.

Copyright infringement: In 2018, Arista paid $400mm to Cisco in a legal settlement. Cisco sued Arista for using some technology that was developed by Arista’s founders when they were working at Cisco. It was the first and most likely the last similar legal issue that they will encounter.


Source link