Reassessing Cisco's AI Data Centers: Can the Former Networking Giant Make a Comeback?

Over the past two decades, Cisco has gradually shifted from being the undisputed king of internet infrastructure to a mature dividend stock in investors’ eyes. With the rise of cloud computing, hyperscale data centers increasingly adopted white-box hardware, commercial chips, and self-developed networking architectures, while newer rivals like Arista emerged in the high-speed switching market. Although Cisco still maintains a vast enterprise customer base, distribution channels, and service revenue, it has long lacked a compelling new narrative capable of driving valuation expansion. Now, the rapid build-out of AI data centers is finally giving this former networking giant another chance to re-enter the game.

AI clusters differ significantly from traditional cloud workloads. When tens of thousands of GPUs train models simultaneously, any network latency, packet loss, or congestion can leave expensive compute capacity idle. As a result, switches, optical modules, networking chips, and management software are no longer just supporting components—they have become core infrastructure directly impacting computational efficiency. Market discussions around AI infrastructure have overly focused on GPUs, often overlooking the fact that every new batch of accelerators also requires proportional upgrades in network bandwidth, optical connectivity, power management, and cybersecurity.

Cisco’s latest figures suggest this turnaround isn’t just marketing hype. As of Q3 of fiscal year 2026, the company has accumulated $5.3 billion in AI infrastructure orders from hyperscale customers and has sharply raised its full-year order forecast from $5 billion to $9 billion, with related revenue guidance also increased to $4 billion. Network product orders grew by more than 50% year-over-year in Q3, while data center switch orders rose over 40%, reflecting that AI-related capital spending is beginning to translate into tangible business.

The key to revival lies not in switches, but in platform control

On the product front, Cisco has launched its Silicon One G300 chip with a throughput of 102.4 Tbps, paired with Nexus 9000, Cisco 8000, 1.6T optical components, and liquid cooling systems, directly competing for the AI Ethernet market. This approach must contend not only with Arista and the Broadcom ecosystem but also with NVIDIA’s InfiniBand and Spectrum-X solutions. Cisco doesn’t need to fully displace NVIDIA to succeed; its more pragmatic strategy is to let customers choose between Cisco’s in-house chips and NVIDIA’s switching silicon, while retaining Cisco’s operating system, management tools, and security stack.

Cisco’s real strength lies in its global base of enterprise customers. Many banks, governments, telecom providers, and multinational corporations may not design their own AI networks and are reluctant to manage a dozen or more vendors simultaneously. Cisco can bundle switches, routers, observability tools, zero-trust security, and technical support into a comprehensive solution, lowering the barrier for enterprises deploying private AI, edge AI, and sovereign clouds. Its Secure AI Factory partnership with Nvidia leverages existing channels to extend AI infrastructure beyond a handful of tech giants to mainstream enterprises.

Its approximately $28 billion acquisition of Splunk in 2024 should also be understood within this context. AI data centers generate massive volumes of logs and telemetry data—just one network failure can bring an entire GPU cluster to a halt. If Cisco can tie hardware sales to Splunk’s observability, network management, and security subscriptions, it can transition from a one-time equipment vendor to a recurring-revenue operational platform. That shift—not merely selling a few more switches—is the key to valuation re-rating.

However, investors must distinguish between structural revival and cyclical rebound. Total product orders grew 35% in the third quarter; excluding hyperscalers, growth was 19%. Campus networking orders also rose by more than 25%. In other words, the current strong performance reflects a mix of enterprise equipment upgrades, improved supply cycles, and a limited number of large AI-related orders. Hyperscalers wield significant pricing power, place concentrated orders, and exhibit volatile delivery timelines—factors that may not translate into the high margins the market imagines.

Therefore, Cisco’s ability to reinvent itself hinges on whether AI-related revenue can be sustained, whether its customer base can expand from a few cloud giants to the broader enterprise market, and whether each dollar of hardware revenue can be augmented with additional software, security, and observability subscriptions. Cisco likely won’t become another Nvidia, but it has the potential to reemerge as an indispensable control layer in AI data centers. AI has opened the door for this legacy giant; whether it achieves a full re-rating ultimately depends on whether it sells just hardware—or a platform that’s hard to replace.

(Chips & Compute Series #84)

$Cisco (CSCO.US)$ $Amazon (AMZN.US)$ $Oracle (ORCL.US)$ $Fastly (FSLY.US)$ $DigitalOcean (DOCN.US)$ $Hewlett Packard Enterprise (HPE.US)$

91K Views