✦ VeloCloud & the Enterprise EdgeNarrow moat

The quietest bet reverses a founding principle. Arista spent two decades refusing to chase the enterprise edge — no branch routers, no WAN — and then, in mid-2025, bought VeloCloud, the SD-WAN business, from Broadcom for a price reported well under $1 billion, taking the intellectual property and roughly half of the unit's thousand people, mostly engineers¹. Weeks later it launched an expanded AI-driven campus and branch portfolio around it². The logic: the campaign for the enterprise — Arista's second act beyond the titans — was incomplete without the branch, and an AI-era WAN (agents everywhere, inference at the edge) is a different product than the one VMware let stagnate.

The skeptic's case writes itself: SD-WAN is a crowded, maturing market led by Cisco and Fortinet, VeloCloud changed hands three times in eight years, and integration — VeloCloud onto EOS-adjacent management — is the kind of work that quietly eats a small acquisition. Arista's own history is the counter-argument: it has never bought big, never bought carelessly, and its enterprise campus business grew from nothing into a multi-billion-dollar line against the same incumbents.

Watch whether WAN revenue ever earns a disclosed line or a named contribution to the enterprise segment, and whether VeloCloud's manageability merges into CloudVision within a couple of years. A working branch portfolio makes Arista a full-stack enterprise networking company for the first time; a stalled one would be a footnote bought cheap.