AI adoption strains enterprise networks, Cisco finds

Enterprise networks are struggling to keep pace with the demands of artificial intelligence, according to new research from Cisco and Foundry. The findings underscore a growing tension between AI adoption and existing infrastructure, as organizations grapple with capacity constraints, security risks, and operational blind spots.

The survey, which included over 3,400 IT and networking leaders, reveals that 73% of organizations are either already experiencing or anticipating network capacity limitations within the next two years. The strain stems from the rapid integration of agentic, generative, and physical AI into enterprise operations, which has led to surging data traffic and widened attack surfaces. While Cisco’s role as a networking vendor may color its framing, the data presents a clear challenge: AI workloads are outpacing the ability of many networks to support them.

What the data shows

The research highlights three primary pain points. First, network capacity is becoming a bottleneck, with AI-driven traffic spikes overwhelming legacy infrastructure. Second, security exposure is expanding as AI systems introduce new vectors for threats, particularly in environments where visibility tools lag behind adoption. Third, operational gaps are emerging, as IT teams struggle to monitor and manage AI workloads effectively across hybrid and multi-cloud environments.

These issues are not confined to early adopters. The survey suggests that even organizations in the planning stages of AI deployment are bracing for disruptions, with many lacking the tools to measure or mitigate the impact on their networks. The findings align with broader industry trends, such as the shift in data center spending toward AI-optimized hardware, as reported by IDC in the same period.

Why infrastructure is the new frontier

The strain on enterprise networks reflects a broader recalibration of IT priorities. For years, procurement cycles revolved around predictable x86 refreshes, virtualization density, and standardized cloud services. AI, however, is reshaping these assumptions. IDC’s first-quarter 2026 data shows worldwide server revenue surging 30.4% year-over-year to $122.6 billion, driven by demand for accelerators, non-x86 systems, and high-performance memory. Yet even this growth is constrained by supply chain bottlenecks, particularly for components critical to AI workloads.

The infrastructure challenge extends beyond hardware. Equinix’s recent expansion of its AI Factory initiative, in partnership with Cisco and NVIDIA, signals a push toward production-grade AI environments. The collaboration includes Presidio’s P.A.T.H. Lab, a testing ground for enterprises to validate AI infrastructure before scaling. This move underscores a shift from experimental AI projects to operational deployments, where network reliability, cooling capacity, and data proximity become critical factors.

What to watch

The next two years will likely see a bifurcation in enterprise AI strategies. Organizations with robust, scalable networks and observability tools may accelerate their deployments, while those with outdated infrastructure could face delays or higher costs. The rise of agentic AI—systems that autonomously execute tasks—will further test network resilience, as these workloads demand low-latency, high-bandwidth connections.

Security will also remain a flashpoint. As AI systems proliferate, so too will the attack surface, particularly in environments where governance and visibility are lacking. Netskope’s recent push to integrate AI security with its managed services partner program reflects this growing concern, as enterprises increasingly outsource both cybersecurity and AI operations to specialized providers.

For now, the message is clear: AI adoption is no longer just about models or compute. It is about the underlying infrastructure’s ability to support, secure, and scale these workloads—and many organizations are still playing catch-up.