Arista raises 2026 guidance by $1.1B as AI network demand surges

Arista Networks reported second-quarter revenues of $3.04 billion for the period ended in June, up 37.7 percent year-over-year and representing a 12.1 percent sequential increase from Q1 2026. Crucially, this marked the first time the company surpassed $3 billion in quarterly sales. Product revenues rose to $2.61 billion, up 38.8 percent, with growth spread across product lines and segments. Services revenues climbed to $430.5 million, up 31.3 percent, driven by adoption of Arista's EOS network operating system, CloudVision data tracking, LANZ latency analyzer, Path Tracer link analyzer, DANZ Monitoring Fabric packet broker, and Connectivity Monitor jitter and packet loss tools. The body estimates that software subscriptions alone reached approximately $137.2 million, up 4.3X year-over-year, with software and subscriptions combined hitting $567.7 million, up 58 percent.

On the bottom line, operating income grew 39.7 percent to $1.38 billion, while net income rose 36.5 percent to $1.21 billion, representing a robust 40 percent of revenues. The company's balance sheet strengthened considerably: cash and equivalents increased 50.9 percent to $13.34 billion year-over-year, deferred revenues jumped 69 percent to $6.87 billion, and customer purchase commitments surged 2.7X to $9.7 billion. This combination of large forward commitments and substantial cash reserves gives Arista negotiating leverage with component suppliers—a critical advantage in the tight supply-chain environment the body describes as endemic to IT equipment suppliers serving hyperscalers.

Looking ahead, Arista raised its full-year 2026 guidance by $1.1 billion to $12.6 billion, implying 44 percent annual growth versus 2025. Within that, the company is forecasting AI back-end and front-end network revenues of $3.6 billion for 2026 (up $100 million from prior guidance), representing around 2.1X growth year-over-year. Campus-focused networking is expected to reach approximately $1.25 billion, up 53.4 percent. CEO Jayshree Ullal broke down the AI network opportunity into three categories. Scale-across networks—those linking datacenters and regions together—are expected to represent about 30 percent of the AI back-end/front-end revenue category in 2026, or $1.08 billion. Scale-up networks, which link GPU and XPU memory within a single node or rackscale system, and scale-out networks, which loosely couple multiple nodes to share work, are expected to combine for the remaining 70 percent, or $2.52 billion.

Ullal characterized the broader scale-across opportunity as roughly a $3 billion market for switch gear and $1 billion for optics in 2026. Arista's implied market share in this segment is approximately a little more than a third. By 2030, scale-across switching is projected to grow to around a $15 billion opportunity, with another $5 billion for related optics. A major constraint on scale-up expansion is Nvidia's proprietary NVLink and NVSwitch ecosystem. The body quotes Ullal as saying: "Where it's non-Nvidia, Arista will be excited and will be working more closely both in the scale up and scale out – in some cases to build custom racks with their custom processors so that we can better tune our network with the behavior for their inference or training engine. In the Nvidia cases, it's going to take a bit longer. I feel a little bit like two or three years ago we were talking about InfiniBand and now we don't mention InfiniBand. But it took two or three years to move to Ethernet. So it will take time to go from a proprietary scale-up that's been around a long time with NVLink to these other alternatives, even if Ethernet is really good." This suggests that displacement of NVLink by Ethernet-based alternatives may take two to three years, similar to the historical shift from InfiniBand.