Arista Networks Stock: Huge Potential, Attractive Valuation (ANET)

Investment Thesis

Arista Networks (NYSE:ANET) has been building a leading position in the data centre equipment space for over a decade. It has become a key supplier of ethernet switches and routing equipment to large internet companies, data centres, and corporate campuses and has been successfully riding the wave as software has transitioned from on-premise licences to SaaS licences. This shift (among other drivers) has fueled sales growth and it is becoming increasingly clear that artificial intelligence will be the next growth driver. Through large investments in R&D, Arista is perfectly positioned to remain a key supplier to enable the shift as the internet industry once again transitions into a new paradigm: one enabled by AI.

The ANET stock price is close (within 11%) to my estimate of fair value; however, given the potential of AI, there may be upside to the current expectations, simply because the future of AI is so uncertain. In reference to such matters, Bill Gates said it best:

Most people overestimate what they can achieve in a year and underestimate what they can achieve in ten years.

Company Overview

Arista Networks operates in the data centre hardware infrastructure business. The company designs and manufactures ethernet switches and routers that are used in large-scale data centres and enterprise campuses. Arista primarily competes in the switching market for 10 Gigabit Ethernet (GbE) and above. 10 Gb refers to the speed at which computers can transfer data. Ethernet is the fastest periphery for transferring data between computers and servers and is standard in data centres. Arista competes at the high end of this market providing high speed Ethernet switches to enable high speed transfers in data centre fitouts.

The industries that Arista services include large internet companies, service providers, financial services organisations, government agencies, media and entertainment companies, telecommunication service providers and other cloud service providers. Arista has over 9,000 customers; however, the biggest and most important are Meta (META) and Microsoft (MSFT), both significant contributors to Arista’s revenue.

Industry Analysis

Over the last decade or so, software applications have progressively been moving from on-premise installations to “cloud” installations, allowing users/subscribers to access the software from any internet-connected device. These applications are distributed across thousands of servers which are connected by high-speed networking switches that allow for rapid deployment and updating.

This has been a major driver for data centre space and subsequently data centre equipment, such as the switches and routers that Arista provides. However, in recent years, the development of artificial intelligence products has been and will continue to be major drivers of data centre requirements, as the data and compute power required to power these applications are significantly higher than other applications.

Arista estimates the total addressable market for its products will grow 13% p.a. from $31bn in 2023 to $51bn in 2027.

A growing market combined with a demonstrated ability to take market share is likely to result in solid revenue growth. According to the 1Q23 Deck, Arista continues to take market share from Cisco (CSCO), having grown its share of the 10GbE ethernet switch market from 3.5% in 2012 to 25.3% in dollar terms.

There have been several industry characteristics that have been driving demand for data centre capacity including the shift to mobile, the Internet of Things and, most notably, the shift toward cloud-based applications. The next driver will be AI; specifically, cloud-based generative AI. There has been an explosion in announcements of generative AI applications being released for public use including ChatGPT; the new Bing Search, which is powered by ChatGPT; Google’s Bard; and new features in Adobe Photoshop such as generative fill. The mere mention of AI on conference calls has exploded too. One such example was Microsoft. The phrase “AI” was mentioned 46 times on the recent 3Q23 conference call, up from 7 times on the 3Q22 call a year ago.

There are many more generative AI programs that have been announced and there will be many more that are yet to be announced. And it is the latter part of that sentence that my thesis rests on. At the current share price, I estimate that ANET stock is slightly undervalued. This might suggest a hold rating while you wait for a pullback. This might be a sound strategy, but at the same time, it is likely that we are only scratching the surface of what is possible with AI.

It is tempting to compare estimates of what a mockery the computer industry made of forecasts made in the 1970s, or of forecasts made of the internet industry in the 1990s. I don’t think AI will be as revolutionary as what, in hindsight, were tectonic shifts in how the world operates, but there is every possibility it will be more transformational than the investing public (including Wall Street) is inherently forecasting over the next decade or two.

The beneficiaries of the AI boom may or may not be those early in developing AI applications (all you have to do to understand this is look at the number of PC companies in the 1970s and internet companies from the 1990s that are still around today). But those that will almost certainly benefit are those selling the picks and shovels that enable the industry, today’s equivalent being data centres and data centre componentry such as switches and routers.

Competitive Advantages

The reason Arista is likely to benefit strongly is not just the company selling what the industry needs when it needs it. It hasn’t taken 25% market share in high speed ethernet switches just by selling average parts. It has done so because it sells a leading product. There are two key competitive advantages I believe Arista possesses.

Highly Scalable Platform

Arista doesn’t just design and distribute hardware; core to its selling proposition is its extensible operating system (EOS). Arista’s EOS is a purpose-built cloud networking platform that is built on Linux, highly scalable, works on open standards, and is programmable. This changed the way data centres could be designed, allowing them to scale much larger, improve speed and reliability performance, while also doing so cost effectively.

It was the first of its kind and Arista continues to be at the forefront of this innovation.

Outsourced Contract Manufacturing

Arista possesses the intellectual property to design and distribute its market-leading products; however, the physical manufacturing of the products is outsourced to highly trusted contract partners. This allows Arista to focus its R&D on the design of the products without the capex needed to build facilities and manufacturing capabilities. As a result of this (and Arista’s decision to expense most all R&D), Arista’s capex requirements are very light, amounting to only approximately 1% of revenue. R&D is around 20% of revenue.

Risks

Arista might be highly attractive but it is by no means risk-free. There are two specific risks I wish to discuss, as I believe these to be the most pertinent for investors to understand the Arista business model.

Customer Concentration

Microsoft and Meta are Arista’s 2 largest customers, which together contributed 42% of Arista’s 2022 revenue. In 2021 and 2020, the number was somewhere between 25% to 30%, as inferred by the ANET 2022 10-K (page 21). This is incredibly high and can be volatile depending on the timing of the release of new products by these companies. If Microsoft and Meta reduce their spend, which has been flagged as a possibility in 2023 after the bumper 2022 year, this will have an adverse effect on Arista’s revenue.

Long inventory lead times

Arista is required to make its own purchasing decisions (from its contract manufacturers) based on an expectation of future demand. This can be difficult to predict and could (and has recently) lead to mismatches in orders from Arista customers and the inventory purchased by Arista. This risk seems heightened compared to most companies that require investment in inventory for their business. CEO Jayshree Ullal said on the recent 1Q23 conference call that Arista’s products can have a 72-week lead time, meaning the company must anticipate what demand for those products is likely to be more than a year in advance. This can impact cash flow and potentially result in missed sales orders if the company gets it wrong.

Valuation

My preferred valuation method for companies with a track record and plausibly forecastable earnings and cash flows is the discounted cash flow (DCF). ANET fits comfortably into that definition and my DCF rests on the following assumptions:

• 13% CAGR revenue and EBIT growth for 10 years, implying minimal operating leverage
• 8.8% WACC (4% RFR, 0.91 beta, 5.3% ERP). Beta is low because I used a levered beta based on a peer set and Arista has immaterial debt compared to the peer set that is much higher leveraged
• 3% LTGR

This produces a valuation of $172, 11% higher than the current share price. This valuation implies a fair value P/E of 34x, justifiable because of the high return on invested capital the company produces each year. ROIC has averaged 29% over the last 10 years and was 34% in 2022, which is excellent.

There are risks involved in an investment in Arista, which could impact future revenues. But between the small discount to valuation and the upside that an untamed AI market could produce, Arista is a very attractive opportunity for the long-term buy-and-hold investor.