An AI-driven switching leader with unusually strong cash generation is expanding into campus/WAN and automation software. The upside case depends on sustaining AI-fabric share while defending against bundled GPU-network stacks.

Arista can compound through 2030 by staying the default “production Ethernet” for AI clusters (800G→1.6T) while expanding enterprise campus/WAN and turning operations data + agentic automation into higher-multiple software revenue, with upside from standards leadership (UEC/ESUN) and selective vertical integration where supply-chain risk is highest.

AI makes network reliability/automation a first-order constraint; Arista sits at the AI throughput bottleneck (the fabric) and can monetize trust + telemetry + automation as cognition becomes cheap.

Arista’s edge is turning merchant-silicon speed into a trusted, operable AI Ethernet fabric (EOS + telemetry + automation) and then re-selling hardened designs into enterprises. Over 5 years, the non-linear upside is software/automation attach (closed-loop ops, AI job-centric observability, security evidence) and WAN expansion from the VeloCloud portfolio, which can reduce cyclicality versus pure hyperscale switching. The stock can plausibly deliver ~2x EV if revenue triples while the multiple normalizes only modestly versus legacy networking peers due to superior margins, cash generation, and AI fabric relevance.

The main downside is a correlated double-hit: hyperscaler capex lumpiness plus competitive bundling (GPU+NIC+switch) that commoditizes Ethernet switching economics. Secondary risks are concentration (customer timing power), geopolitics/tariffs in the supply chain, and whether software ambitions (multi-vendor ops, security) can scale without becoming a services-heavy, lower-margin motion.

$163.37