Cisco Sees AI Fueling Network Upgrade Supercycle

Network Firewalls, Network Access Control , Network Performance Monitoring & Diagnostics , Security Operations Cisco Sees AI Fueling Network Upgrade Supercycle Customers Are Reprioritizing Budgets for AI, Security and Quantum Readiness Michael Novinson (MichaelNovinson) , Tiffany Wang • August 13, 2026     Credit Eligible Get Permission Hyperscalers and enterprises are ramping up network upgrades to support agentic artificial intelligence workloads, manage token-related costs and address emerging security risks, Cisco said. See Also: Embedded Security Tactics in a Modern Network Fabric The "multi-year, multi-billion-dollar networking supercycle," dubbed by CEO Chuck Robbins, drove customers to purchase routers, switches, optics, wireless gear and industrial IoT hardware across data center, campus and service provider networks. Companies need efficient inter-data center communication as they expand AI capabilities, and they are bringing agentic AI inferencing on-premises to curb token costs, all the while modernizing network infrastructure to defend against cyber risks that come with scaling deployments, Robbins said. "What we're hearing from our customers is they're currently reprioritizing within their existing budgets," Robbins told investors Wednesday. "I would also say that you're beginning to see a trend where our customers are looking at AI readiness, Mythos readiness, quantum readiness in a similar vein to how they've looked at cybersecurity spend over the last three to four to five years. It's just not optional." Addressing a three-fold concern over cost, security and data sovereignty, companies are looking for smarter ways to deploy their AI, whether that means changing the type of model they use or where they run it, Robbins said. "It all started with this whole discussion around tokenomics, and then it's expanding quickly into open-weight models, foundational or frontier models," Robbins said. "What we believe is going to happen is you're going to have customers that are going to make intelligent decisions about which models they use based on use cases and which ones are most appropriate for whatever agentic applications they're running." Two broad options are either cloud-based models, where companies send prompts to an AI provider and pay for usage, or on-premises models that companies obtain the model weights, run on their own GPUs and govern data locally - an increasingly popular way to maintain AI deployment while cutting costs, Robbins said. "If they continue to use cloud-based models, that's good for us, just like it has been for the last two years. If they move to open-weight models or models that they're running on-prem, that's great for us because it means they will invest in more enterprise private data center networking, which we've seen the last two quarters," he said. Choosing to go local means spending on GPU clusters on-premises and at the edge with low-latency, high-bandwidth networking and built-in security, observability and automation, Robbins said. But regardless of the architecture, securing thousands of agents in the infrastructure is top of mind. "How are enterprises going to navigate the future?" Robbins asked. "There's agentic security that's coming into it. There's security of my data. There's sovereignty of my data that comes into it." Building new AI models and workflows also plays into the need for upgraded networking. "As AI models grow in complexity and size, hyperscalers need to connect or scale across multiple data centers due to physical and power limitations in a single data center," Robbins said. "Power efficiency, reliability and scale are critical in ensuring distributed GPUs function as if they were in the same location." Service providers and cloud companies nearly doubled their Cisco orders from a year earlier, while four major hyperscalers increased AI infrastructure purchases by triple-digit percentages, Robbins said. Cisco racked up $9.3 billion in the past year from hyperscalers' AI infrastructure orders. More than half of Cisco's customers also modernized their workplaces alongside data center upgrades, investing in new switching, routing and wireless products. For example, Robbins said a frontier AI company bought Cisco's Wi-Fi, smart switches and end-to-end segmentation solution for speed and security across different office locations. "We see the momentum in campus networking being driven by infrastructure modernization to both scale AI initiatives and to strengthen defenses against a rapidly evolving cyber landscape," Robbins said. Although Cisco hasn't seen too much impact from Mythos, Robbins said the agentic attack scenario has companies looking at a refresh for equipment past its last day of support. "They start from the premise of 'we don't have a choice,' and then they figure out how to fund it," he said. "And I think that's just a much different place than we have been." "Now we have this impending Mythos effect that they're all looking forward to," Robbins said. "Downstream, we have quantum that they need to prepare for. Those are all drivers of what we've been talking about relative to just an overarching campus refresh, where for the first time ever we have our campus networking - both switching, routing, as well as wireless and everything - going through a refresh at the same time." Stock Drops Despite Sales, Earnings Exceeding Expectations Category Quarter Ended July 25, 2026 Quarter Ended July 26, 2025 % Change Total Revenue $17.25B $14.67B 17.6% Security Revenue $2.23B $1.95B 14% Net Income $3.86B $2.55B 51.3% Earnings Per Diluted Share $0.97 $0.64 51.6% Non-GAAP Net Income $4.87B $3.95B 23.2% Non-GAAP Earnings Per Share $1.22 $0.99 23.2% Source: Cisco Cisco's revenue of $17.25 billion in the quarter ended July 25 crushed Seeking Alpha's sales estimate of $16.83 billion. Meanwhile, the company's non-GAAP earnings of $1.22 per share beat Seeking Alpha's non-GAAP estimate of $1.17 per share. The company's stock dropped $5.04 - or 4.07% - to $118.84 per share in after-hours trading Wednesday, which is the lowest Cisco's stock has traded since Aug. 3, 2026. For the fiscal quarter ending Oct. 24, Cisco expects non-GAAP net income of $1.32 to $1.34 per share on revenue of between $18 billion to $18.2 billion. That compares to analyst expectations of earnings of $1.14 per share on revenue on $16.66 billion, according to Seeking Alpha.