2023-08-16 09:00:00 ET
Summary
- Arista Networks offers a reasonably priced investment opportunity in the AI market with strong AI exposure and growth potential.
- The company's software stack and hardware capabilities set it apart from competitors, providing "second to none" deployment capabilities.
- Arista is positioning itself at the forefront of future AI-focused networking technologies, aiming to address bottlenecks and improve throughput in data centers.
- While other AI names fall short of checking all the boxes, ANET passes the test for AI exposure at a reasonable valuation.
You want to invest in AI because you believe it's the real deal, but you look at all of the popular names like Nvidia ( NVDA ), Microsoft ( MSFT ), Google ( GOOG , GOOGL ), and many others and wonder if your thoughts are a day late and a dollar short. Perhaps it's because valuations have gone to a different planet, or the names don't quite carry enough AI exposure or vision to satisfy your objectives. You're left wanting, thinking, "Is there anything reasonable left in this crazed market for the AI boom?" Thankfully, the answer is a resounding yes, and it's found in the reasonably priced and very AI-exposed Arista Networks ( ANET ).
Arista isn't a new name for me; I've been covering it since late-2019 (pre-pandemic, would you believe it?) and have covered it both publicly on Seeking Alpha and with members of my investing group since then. I've observed the evolution of the company, witnessing it push into some of the biggest fortresses of network and switching owned by Oracle ( ORCL ) with resounding success. I mean, what other names in networking can you find growing well over 30% on the top line with a 2023-ending forward P/E valuation of under 30? Oracle's revenue growth rate is less than half of Arista's, with a mid-2024 forward P/E of more than 20.
Even if an AI investment wasn't your objective, you'd still have found a pretty interesting stock to dig into.
But these surface-level readings don't depict the entire story over the years of the company's agility from its core market to an adjacent market to now, a revolutionary one. Not only is Arista a solid business, but its AI potential puts it over the top over the next several years.
Isn't Today's Networking Good Enough For AI?
Your first thought might be networking is networking is networking. And, for the most part, you'd be right. We all take the internet in our homes for granted as you 'simply' log on to the Wi-Fi or plug in the Ethernet cable to your NIC card, and away you go. But eventually, your internet surfing is led to an Ethernet cable (if not right away) and then to a coax or fiber cable. Then eventually, your data is at a major ISP hub where massive switches and networking racks move your data like a futuristic phone switchboard from the early 1900s. And the data more or less is guaranteed to have ventured to a server in a data center somewhere to receive your query.
And then it's all back through the entire pipeline with a hop (or two or ten), skip, and a jump to return your query with an answer (of data needing parsing).
This long but miraculously quick journey required a lot of networking equipment and cables.
Now, for the most part, Arista's top clients are the ones at the 'hop' in the data center in the above journey. They include the likes of Meta Platforms ( META ) and Microsoft, two major players in the current world of AI. And, as these two cloud titans, as Arista's management calls them, need to convert and expand their data center footprints into cutting-edge, highly optimized AI home bases, it requires the fastest switches and the best network equipment.
Even in the non-AI world, faster switches are needed to route the ever-increasing amount of data coming in and out of a data center, not to mention the communication within the data center as certain racks of servers interact with other racks of servers for shared computation while also delivering data to and from hot and warm media (storage).
This has meant cloud companies going from 200-gigabit to 400-gigabit switches to further facilitate the increase in traffic and data in the modern data center.
But then you add the intensely quick onset of AI and the need to host AI in the cloud, and these cloud titans are left with needing the fastest networking known to man. This is on par with 800-gigabit switches - something coming out of the pipeline and readying for deployments over the next year.
Now, why wouldn't a cloud titan or a customer who needs fast networking just buy any networking product? After all, a 400 gig switch is a 400 gig switch; what more do I need than the promise of the speed written on the box?
Something I've said repeatedly, especially in the area of AI accelerators from Nvidia and AMD ( AMD ), is software and the underappreciation of a full software stack to the hardware counterpart. Hardware is great - it takes a lot of technological know-how, expertise, and IP - but hardware is impossible to use, let alone optimize, without software. But the software needed for AI must be highly optimized and contain all the necessary software layers to run the hardware from the registers to the users.
And it's no different with networking, believe it or not.
Arista's secret weapon has been its software stack operating the switches and spines forming the network backbone of a data center, especially one with thousands of ports and endpoints.
This is why I find it interesting to be asked about hardware and how competitive it is. What should be asked is, "What is the software stack, and how well does it perform?"
It's nothing new to be asked about any 'ol network switch being used in place of an Arista product. They're called "white boxes," representing the generic switch or networking product capable of the same speed. They do work - they push traffic and data through the pipe - so why not use those instead of Arista's products which are probably a bit more expensive?
This is where Arista's management impresses me. A lot of CEOs and management teams are good at what they do. Some are really good at business but lack the technical understanding of a product, but they can lead the company. Some are really technical but have less than desirable communication skills to convey why the technology they possess is competitive and can outperform in the market.
Then there's Arista's management.
Arista's management not only knows business and its products, but it provides convincing, concise arguments on its earnings calls, the likes of which I have rarely seen elsewhere. And on its Q2 earnings call just two weeks ago , Jayshree Ullal, the CEO, gave a rock-solid answer to the question about white boxes and competitive alternatives:
...the white box question has remained with the Arista as one of the most popular questions asked right from the time of our IPO, whether it's a 10 gig, 40 gig, 100 gig, 400 gig, or now you ask it at 800 gig, I think there will always be an element of white box if somebody is just looking to build something and throw in some quick traffic. But for some of these most mission-critical networks, it's less about the box and more about the software stack and how much performance availability, power you really get out of this.
So, the cost of putting in the box, if you save something , if you even save something, is far dwarfed by the total OpEx you need to make that box work. So, we continue to believe that we will coexist with white box in some of our cloud titan customers. We will continue to run both SONiC and FBOSS in the case of Microsoft and Meta along with our EOS. But at the end of the day, whether it's a white box or a blue box, it's the software stack that really wins.
Arista gets it - it's about software. And Arista's software provides a level of networking overview clients can't get from other vendors. Forrester says it's "second to none" when it comes to deployment flexibility, and its software is to thank.
Arista's acquisition of Awake in 2020 expanded its networking capabilities into the NAV market and created an easily deployed solution in a space known for complex hybrid networks. Previously known for large-scale deployments of on-prem infrastructure hardware, the addition of Awake has expanded Arista's footprint within very large enterprises. Arista's vision is unique in that the solution, and as an internal organizational focus, is aligned with Zero Trust principles. While many vendors offer internet of things/operating technology (IoT/OT) capabilities, Arista embraces the unknown or nonstandard elements to provide greater visibility into an otherwise murky quagmire of disparate IoT/OT networks.
- The Forrester Wave: Network Analysis And Visibility , Q2 2023
Arista's management has worked its products together with an intentional focus on software to be successful. How does it differentiate itself from white box equipment? Software. Its software lessens the operating expenses of its clients compared to just plugging in a switch and pushing traffic through because it provides network administrators the needed visibility, flexibility, and architecture understanding to operate efficiently. And it clearly shows with one of the leading research companies calling it out as second to none in terms of deployments and network visibility.
Moreover, its full package is how the company has grown revenue at an average of just over 50% over the last four quarters, with 2023's full-year growth expected to be over 30%.
But today's networking isn't tomorrow's AI-focused networking, as networking needs to keep up with the needs of faster model training and inferencing. So, what is Arista doing to maintain an edge?
Arista At The Center Of Tomorrow's Network
There's not a lot of technology in terms of conduits for packets and data across copper and fiber. There's basically InfiniBand on the back end for interconnect within components of servers and across accelerators, and then the front end with ethernet.
Arista is working to find a new method or technology to fix some of the bottlenecks and square-peg-round-hole solutions from over the years. Where only one solution (the only one) is being put in for several use cases, some of these use cases should have their own solution. This is where Arista is aiming. And, going back to management's ability to be detailed and concise, its answer about where it's going with future technologies deserves all the context.
[The] majority of Arista's participation has been in the front end of the network... And we're getting a chance for the first time ever to play in the back end. So when we think AI, there's clearly some ramifications of bandwidth on the front end of the network, but we're not counting that. So we're truly thinking of something that's incremental, brand new , lot of work to do in testing, proving, pilots, trials before we get into production.
Today, I would say in the back end of the network, there are basically three classes of networks. One is very, very small networks that are within a server where customers use PCIe, CXL. There's proprietary NVIDIA-specific technologies like NVLink that Arista does not participate. Then there's more medium clusters, you can think generative AI mostly and inference, where they may well get built on Ethernet. For the extremely large clusters with large language training models, especially with the advent of ChatGPT-3 and ChatGPT-4, you're not talking about not just billion parameters, but an aggregate of trillion parameters. And this is where Ethernet will shine. But today, the only technology that is available to customers is InfiniBand. So, obviously, InfiniBand with 10, 15 years of familiarity in an HPC environment is often being bundled with the GPUs.
But the right long-term technology is Ethernet , which is why I'm so proud of what the Ultra Ethernet Consortium and a number of vendors are doing to make that happen. So near-term, there's going to be a lot of InfiniBand, and Arista will be watching that outside in. But longer term, Arista will be participating in an Ethernet AI network. And, neither technology, I want to say, were perfectly designed for AI. InfiniBand was more focused on HPC and Ethernet was more focused on general purpose networking. So I think the work we are doing with the UEC to improve Ethernet for AI is very important.
- Jayshree Ullal, CEO, Q2 '23 Earnings Call Q&A (emphasis added)
Arista is skating to where the puck is going, not where it is right now. Since it's tightly working with the UEC, the company will be at the center of the new technology and be in the lead to be one of the first to market.
Some Risks
There's always a debate about whether certain materials are better than others and what use cases they slot into. However, having a new technology emerge based on a familiar one may be just what the industry needs. The risk is this may take longer than expected; Arista did say this is a 2024-2025 timeline. It's possible either another technology gets adopted, or its technology can't make the inroads it expects as the build-out for AI got too far down the road to drive a widespread adoption with an unproven product.
But what I like is Arista is a risk taker, which is better than a company playing it safe and continuing to glide on legacy technology and its large market share. So far, it has proven its risk-taking can produce strong growth.
Within A Still Largely Sane Valuation
One of the biggest problems with finding an AI investment is much of the runway has been covered in terms of share price appreciation - at least for the most prominent players. Microsoft, Nvidia, and AMD have been dealing with other factors in their businesses (because none of them are entirely AI) or have seen the stock take off on next year's expectations.
But Arista has remained one of the unsung heroes in terms of being at the center of what AI requires and stock valuation. And the reasonableness of its valuation is both a comparative look and a standalone look. With expectations for 35% earnings growth for this year, trading under 29 times forward earnings isn't a bad start. This is a PEG of less than 1.
Sure, it appears 2024 earnings growth is set to slow to 10%, but the cautious estimates from the shrinking visibility from cloud customers due to better supply chain timelines should come up materially once the company gets a better commitment from its long-time customers. This will raise analyst expectations as 2024 draws near. And this doesn't account for the company's consistent performance at beating earnings expectations by meaningful amounts every quarter. So even what analysts expect leading into a report is routinely beaten by the company.
This brings me to one of the more comforting things with Arista, GAAP earnings. Many tech companies report both non-GAAP and GAAP earnings, and many times, they differ quite a bit. Substantial stock-based compensation with one-time items impacts the divergence, with the former being the most common. And for Arista, stock-based compensation accounts for 4.6% of its expenses for the six months ending June 30, 2023. This puts its Q1 and Q2 combined GAAP EPS at $2.94 while its non-GAAP is $3.02.
This is compared to Nvidia, which has 10.2% of revenue worth of stock-based compensation in its FQ1 '24, growing from 7.0% of revenue in FQ1 '23. Even a mature company like Microsoft had a 4.3% of revenue rate of stock-based compensation in its latest quarter. So for Arista to manage its compensation to a mature-like level while still growing revenue in the 30 and 40% range is a little impressive.
From a valuation standpoint, there's not much to dislike about where ANET trades on an objective basis and where it trades on a comparative basis. It checks all the boxes of a strong tech company with disciplined management. To top it off, even with all the AI buzzing in 2023, you aren't chasing a share price around, as it trades within relatively historical valuation ranges as it has risen.
Who'd A Thought Networking?
Networking is not a sexy sector within the tech industry to find growth. But this networking company checks all the boxes, including being at the center of the AI build-out of data centers, finding the next technology directly related to moving AI forward with faster and highly maintainable networks, and has a valuation in line with growth to match the value paid.
All of the common names like Nvidia, Taiwan Semiconductor ( TSM ), Meta Platforms, Google, Microsoft, and AMD either have weak exposure, poor valuation, don't have a clear vision for AI, or aren't seeing the financial results expected from it. Arista Networks clears the deck and provides a highly AI-exposed opportunity at a reasonable price. And not forgetting, it's a pioneer for an unsung aspect of AI: networking.
It's almost like being off the grid of AI but still plugging into all the forces moving it.
Editor's Note: This article was submitted as part of Seeking Alpha's Best AI Ideas investment competition , which runs through August 15. With cash prizes, this competition -- open to all contributors -- is one you don't want to miss. If you are interested in becoming a contributor and taking part in the competition, click here to find out more and submit your article today!
For further details see:
Arista Networks: The Unsung Hero Of The AI Revolution