Cisco Books $9.3 Billion in AI Orders as Enterprise Network Demand Grows

Cisco booked $4 billion in AI infrastructure orders from hyperscalers in its fiscal fourth quarter and $9.3 billion for the full year, the company reported on August 12. Quarterly revenue rose 18% to $17.3 billion. The bigger signal for enterprise technology leaders sits outside the cloud giants: excluding hyperscalers, Cisco's product orders still grew 25%, which suggests that corporate networks, not just AI data centers, are being rebuilt for AI workloads.

What did Cisco report?

For the fiscal year ended in July, Cisco said revenue reached $63.3 billion, up 12%. Total product orders rose 35% in the fourth quarter, and networking orders rose 40%, the eighth straight quarter of double-digit growth. Security revenue was $2.2 billion in the quarter, up 14%, though for the full year it grew only 2%.

Cisco guided to fiscal 2027 revenue of $72.2 billion to $73.4 billion. According to Unite.AI's summary of the results, the company expects about $7.5 billion in hyperscaler AI infrastructure revenue in the new year, up from roughly $4 billion in fiscal 2026. Orders have been climbing all year, from $1.3 billion in the first quarter to $4 billion in the fourth.

CEO Chuck Robbins framed the moment in big terms on the earnings call. "We believe we're in the early stages of a networking super cycle," he said, as quoted by The Stack.

How much of the AI demand is coming from enterprises?

Most of the AI-labeled dollars still come from hyperscalers. But the details point to a second wave. Research firm Futurum Group wrote that Cisco took more than $400 million in AI orders from neoclouds, sovereign AI projects and enterprises in the quarter, and more than $1 billion for the year. Enterprise orders for Cisco's Nexus data center switches tagged for AI rose more than 85% from the previous quarter, Futurum said, while campus orders rose 20% and Wi-Fi 7 made up more than half of wireless orders.

Our read: model training is concentrated in a few giant facilities, but inference, retrieval and agents tend to run near enterprise data, in private data centers, branches and the edge. That traffic crosses networks that many companies last upgraded before generative AI. In a AI Connect webinar on what an AI factory is, Kaushik Shirhatti, VP of Global AI Networking at NVIDIA, said users don't care how many engineers it took to stand up a GPU cluster. What matters is "time to first token," and ultimately "time to wow and time to value."

Cisco is also building toward AI agents. Futurum noted that Cisco's purchases of Galileo, an AI observability company, and Astrix Security, which secures non-human identities, give it coverage for monitoring and securing agents. Firewall orders rose more than 30% for the second quarter in a row, according to Futurum.

Why did Cisco shares fall after a record year?

Investors focused on cost. The Stack reported that Cisco's non-GAAP gross margin fell from 68.7% in fiscal 2025 to 66.9% in fiscal 2026, and that the company guided to about 65% to 66% for the first quarter of fiscal 2027. Shares were down about 6% in premarket trading on August 13, The Stack said. Futurum flagged a higher hardware mix and memory costs as pressures on product margins, and said the fiscal 2027 outlook depends on Cisco converting its AI order pipeline into revenue.

Cisco's release also recorded $511 million in restructuring and other charges in the quarter, tied to a plan announced in May that it expects to complete by the end of fiscal 2027.

What should enterprise leaders take from Cisco's quarter?

Three practical points stand out.

Network capacity is now part of the AI budget. If agents and retrieval systems will pull data across campus and data center networks all day, bandwidth, latency and segmentation become AI questions, not just IT refresh questions. Teams planning inference near their data should involve network architects early, a theme AI Connect panelists raised in discussing AI infrastructure at scale and inference costs.

Hardware costs are rising. Cisco's margin guidance and Futurum's note on memory costs suggest pricing pressure on networking gear. Buyers with refresh cycles in the next year should lock in quotes and compare total cost, including optics and support.

Security and observability are converging around agents. Cisco, like other infrastructure vendors, is assembling tools to watch what agents do on the network. That is useful, but it adds another console. Enterprises should decide where agent telemetry will live before each vendor decides for them, and design for the isolation boundaries that keep one team's problem from spreading.

The next test comes with Cisco's first-quarter results, when investors will see whether AI orders keep turning into revenue and whether enterprise demand holds up alongside hyperscaler spending.

Frequently asked questions

How much AI infrastructure did Cisco sell to hyperscalers in fiscal 2026?

Cisco reported $9.3 billion in AI infrastructure orders from hyperscalers for fiscal 2026, including $4 billion in the fourth quarter.

What is Cisco's fiscal 2027 revenue guidance?

Cisco guided to revenue of $72.2 billion to $73.4 billion for fiscal 2027, including about $7.5 billion in hyperscaler AI infrastructure revenue.

Why does Cisco's quarter matter to enterprises?

Orders outside hyperscalers grew 25% and enterprise AI-tagged data center switch orders jumped, suggesting companies are upgrading campus and data center networks to run AI inference and agents.

Filed under: Data & AI Infrastructure, AI Strategy & ROI, Security & Fraud