CompetitorsNarrow moat

Arista's competitive position changed twice in three years, in opposite directions. It overtook Cisco to become the largest data-centre Ethernet switch vendor in 2024 — the culmination of a decade of taking share from a much larger incumbent. Then Nvidia entered the same market and took its share from under 4% to about 21.5% in a single year¹.

That second event is what these pages are mostly about, and the general competitive argument has its own root threat. What is worth adding here is the mechanism. Arista won against Cisco on product: better software, one operating system, modern architecture, sold to buyers sophisticated enough to evaluate it. Nvidia is not winning on product. It is winning because it sells the accelerators, and a customer buying a rack of GPUs can buy the switching with them. There is no engineering answer to that.

The two other relationships are less discussed and matter. Broadcom supplies the silicon inside Arista's switches and inside almost everyone else's, which makes it a supplier rather than a rival and a chokepoint either way. And the enterprise market, where Arista is expanding, is a fight against incumbents with thirty-year relationships and a very different sales motion from selling to hyperscalers.

Watch gross margin, which eased from 65.6% to 63.4% as larger customers took bigger discounts. Arista's whole argument is that it sells software margins in a hardware body; that number is where the argument is settled.