Cisco posted $17.25 billion in revenue for its fiscal fourth quarter ended June, up 17.6 percent year-on-year and the highest in the company's history. Net income rose 36.7 percent to $3.86 billion as operating leverage ran ahead of the top line.

The driver is generative-AI infrastructure. Cisco booked $4 billion in AI orders from hyperscalers and cloud builders in Q4 alone, up 4.9x versus the prior-year quarter. Full-year fiscal 2026 hyperscale and cloud AI orders reached $9.3 billion, the company told Wall Street.

Cash on hand stood at $15.92 billion at quarter-end, and product backlog reached $46.7 billion. Both balances have stayed roughly flat for two years and now look set to expand with the 2026–2027 GenAI buildout.

What sits inside Cisco's AI order book?

Of that $4 billion in Q4 hyperscale orders, more than $1 billion covered optical transceivers. The remaining $3 billion bought switch ASICs, switches, servers, and some routers, with software attached where the customer did not bring its own.

Neoclouds and sovereign buyers added $475 million in AI component and system orders, up 4.9x year-on-year. Total AI orders for the quarter reached $4.48 billion, a 4x increase.

Cisco's combined datacenter server, switch, and router business — software included — hit $8.81 billion in the quarter, up 28.3 percent. Operating income for that aggregate is estimated at $2.16 billion, up 25 percent.

Takeaway: AI orders are now large enough to move Cisco's entire datacenter segment by single-digit billions each quarter.

Why is the optical transceiver business so important?

Cisco disclosed cumulative shipments of 850,000 transceivers at 400 Gb/sec and 75,000 at 800 Gb/sec. In Q4 alone, the company shipped 100,000 400 Gb/sec units and 35,000 800 Gb/sec units, well above the 750,000 and 40,000 cumulative totals it reported the prior quarter.

Working backward from typical pricing, Cisco has sold slightly more than $6 billion in 400 Gb/sec transceivers and around $915 million in 800 Gb/sec transceivers since it closed the $4.5 billion Acacia acquisition in March 2021. The 800 Gb/sec ramp is just starting, and 1.6 Tb/sec parts are slated to ramp in late 2026 and early 2027.

The supply backdrop is tight. A shortage of 400 Gb/sec and 800 Gb/sec transceivers has accompanied the GenAI buildout, and Cisco has positioned itself at the center of that scarcity through the in-house Silicon One ASIC family and the Acacia optical line — selling networks, components, and upgraded campus links that feed the AI datacenter.

Takeaway: Optics is the densest cash engine inside Cisco's AI stack and the segment most exposed to industry-wide supply constraints.

How did Cisco reach the AI system market?

Cisco entered converged infrastructure in 2009 with the "California" UCS platform, layering x86 compute onto its networking franchise. That product line drove Cisco's product business up 50 percent to roughly $9 billion per quarter at its post-launch peak.

Silicon One, the merchant switch-and-router ASIC family, extended that playbook into the AI era. The family converges switching and routing on a single architecture, with variants tuned for scale-up networking that shares memory across GPUs and XPUs in an AI cluster. Optical connectivity rides on Acacia silicon photonics, brought in through the 2021 deal.

Analyst estimates put Cisco's UCS installed base above 95,000 customers worldwide. That base, historically a conservative upgrader cohort, is now refreshing systems to accommodate GenAI workloads, mirroring the pattern across IBM, HPE, Dell, and Lenovo installed bases. Several Middle Eastern sovereign wealth funds building domestic AI capability have selected Cisco as platform provider, adding net-new business alongside the existing footprint.

Takeaway: Two decades of UCS trust plus captive silicon is what lets Cisco sell AI clusters that hyperscalers and sovereigns will actually order.

What does fiscal 2027 look like?

Cisco guided FY27 revenue to a range of $72.2 billion to $73.4 billion, roughly 15 percent above fiscal 2026 at the midpoint of $72.8 billion. AI hyperscale and cloud revenue alone reached $4 billion in FY26, up 4x year-on-year.

The margin question is open. Cisco cannot earn large gross margins reselling Nvidia components and reference designs, but it can earn healthy margins on its own switch silicon, systems, and software. If Nvidia reselling grows faster than Silicon-One-attached systems in FY27, operating leverage will erode.

Takeaway: FY27 top-line growth is largely locked in; how much of it drops to the bottom line depends on the mix between merchant Nvidia resales and Cisco's own silicon.

Outlook

Cisco has converted a generative-AI infrastructure shortage into its sharpest revenue acceleration in years. Optical transceivers are the fastest-moving product; AI rackscale systems are the fastest-growing order category. Whether the FY27 margin profile holds depends on how much of the revenue Cisco captures with its own silicon rather than third-party accelerators — a mix that becomes clearer as 800 Gb/sec and 1.6 Tb/sec optics ramp through 2027.