One of the worst performers in my portfolio is Arista Networks $ANET. I bought back in mid-2018, and they are down about 19% since then. A trader would have sold this a while ago. But I think my original thesis is still intact:

Arista does switches and routing solutions that are ideal for data centers that struggle to handle the massive amounts of data that AI and machine learning applications require. It is working on AI-ready infrastructure to bring AI capabilities to companies that otherwise would not be able to handle it.

Data centers need suppliers that can provide affordable and reliable switches and routers, which is what Arista Networks does. Arista Networks is mostly a play on big data rather than AI, but it will directly profit as AI begins to play a larger part in the world's economy.

While the hardware of switches and routers are commoditized, Arista also has a cloud software to run its cloud networking solutions called the Extensible Operating System (EOS). EOS allows Arista's products to be integrated with third-party products and customized for applications. EOS coordinates all the participants within the network so they are optimized. This makes it a more affordable and scalable choice than competitors.

Arista is also a player in AI-ready infrastructure,(AIRI), with Nvidia and Pure Storage. Nvidia's GPUs, Pure Storage's FlashBlade storage and Arista's switches combine to create a AI solution that is ready to provide machine-learning answers "in hours rather than days or weeks".

So far its one of the worst performers in my portfolio (bought in 2018 and is down about 18%), but I think the overall thesis is still intact. And I think the market sees Arista as just another Cisco. There are similarities, but there are also differences that may be harder to spot. That might be the opportunity to be right when others are wrong.