Key Highlights

  • Arista Networks passed $3bn of quarterly Revenue for the first time, reporting $3.04bn, up 38% year on year, with a non-GAAP Operating Margin close to 50%.
  • Cisco reported fiscal fourth-quarter revenue of $17.3bn, up 18%, and took $4bn of AI infrastructure orders in the quarter, though its shares fell after the report as investors focused on margins.
  • Marvell reported record fiscal second-quarter revenue of $2.74bn, up 37%, with its data-centre segment reaching $2.17bn, or 79% of the total.

When thousands of AI accelerators work together on a single model, the connections between them matter almost as much as the chips themselves. Networking has become a hidden constraint of the AI build-out, and it is where a distinct group of S&P 500 companies earns its keep. Arista Networks (NYSE:ANET), Cisco Systems (NASDAQ:CSCO) and Marvell Technology (NASDAQ:MRVL) occupy three layers of that story, the switch, the incumbent platform and the silicon underneath, and their most recent results show how unevenly the spoils are shared. Understanding the differences between them helps explain why AI enthusiasm has lifted some networking names far more than others.

Arista Networks (NYSE:ANET): The AI Ethernet Specialist

Arista Networks, the California data-centre networking company listed on the New York Stock Exchange, passed $3bn of quarterly revenue for the first time in the period ended 30 June 2026, reporting $3.04bn, up 38% year on year, with a non-GAAP operating margin close to 50%. Arista sells high-speed Ethernet switches to the cloud giants building AI clusters, and chief executive Jayshree Ullal called networking the central nervous system for infrastructure stretching to data and AI centres.

The company introduced 1.6-terabit switching with liquid-cooling Options and lifted its full-year outlook, guiding to roughly $3.3bn of revenue and non-GAAP Earnings of $1.06 to $1.08 per share for the third quarter. Its dependence on a small number of very large cloud customers, however, remains a defining risk, since a single deployment decision can move quarterly results materially.

Cisco Systems (NASDAQ:CSCO): The Incumbent Monetises AI

Cisco Systems, the long-established networking group listed on the Nasdaq, is far larger but growing more slowly. For its fiscal fourth quarter ended 25 July 2026 it reported revenue of $17.3bn, up 18%, and full-year revenue of $63.3bn. The headline for AI investors was orders: Cisco said it took $4bn of AI infrastructure orders in the quarter and $9.3bn across the fiscal year, mostly from hyperscale customers, and guided to fiscal 2027 revenue of roughly $72bn to $73bn.

Networking orders grew about 40% year on year, an eighth consecutive quarter of double-digit growth. Even so, its shares fell after the report as investors focused on margins, a reminder that winning high-volume AI orders can dilute the profitability of a mature, software-rich business.

Marvell Technology (NASDAQ:MRVL): The Silicon Beneath the Network

Marvell Technology, the semiconductor company listed on the Nasdaq, supplies components that make fast networking possible, including custom AI accelerators, optical interconnect and data-centre silicon. For the fiscal second quarter ended 1 August 2026 it reported record revenue of $2.74bn, up 37%, of which the data-centre segment was $2.17bn, or 79% of the total. Non-GAAP earnings were $0.94 per share.

Its non-GAAP gross margin was about 59%, and it guided to revenue of roughly $3.15bn for the third quarter. Chief executive Matt Murphy said AI-related bookings were exceptionally robust and pointed to an acceleration in the custom-silicon Business during the second half of fiscal 2027.

Not One Market, But Several

Together the three show that AI networking is not one market but several: switches, integrated platforms and the underlying chips, each with different growth rates and margins. The shared vulnerability is customer concentration, because most Demand traces back to a few hyperscalers' Capital budgets, alongside rising competition that includes Nvidia's own networking products and customers' in-house designs.

Readers should watch order trends, margin trajectories, and whether the guidance for late 2026 and 2027 survives any cooling in AI investment. The revenue and earnings above are reported results, but as Cisco's figures underline, booked orders are commitments that have not yet converted into revenue, and outlooks reflect management expectations rather than certainties. In a market this concentrated, sentiment can turn quickly.