New research from Cisco indicates that enterprise networks are increasingly struggling to accommodate the rapid expansion of artificial intelligence, as current infrastructure faces significant capacity and security bottlenecks. A survey of 3,472 IT and networking leaders across 15 countries reveals that AI-driven traffic in campus and branch environments grew by 34% over the past year, with projections suggesting this demand will triple as organizations transition from generative AI pilots to continuous agentic workflows. According to Cisco’s 2025 AI Readiness Index, 73% of organizations expect to hit network capacity limits within the next two years, forcing a sudden shift in how companies prioritize their digital infrastructure.
The findings suggest that the industry’s long-standing focus on data center and cloud-based GPU performance has left a critical gap in the networks that connect employees and end-user devices. As AI agents begin to operate autonomously, they generate unpredictable “east-west” traffic—the internal communication between applications—that traditional network designs were not built to handle. With 85% of surveyed leaders anticipating a significant rise in AI agent deployments, the pressure on campus connectivity is becoming a primary constraint for enterprise digital transformation.
Why AI Traffic Patterns Are Overwhelming Enterprise Networks
Enterprise network teams are reporting that the shift toward AI is fundamentally changing the nature of data flow within the workplace. Unlike traditional software-as-a-service (SaaS) traffic, which is generally predictable and consistent, AI workloads involve high-frequency, complex interactions between various internal systems and agents. According to the Cisco survey, 67% of respondents identified a marked increase in east-west traffic, which complicates congestion management and creates bottlenecks in branch offices that were previously optimized for simple web and cloud access. Projections from the study indicate that overall network traffic is expected to climb 209% over the next three years, a surge that threatens to outpace current hardware capabilities.
This surge is pushing IT departments to accelerate their modernization efforts, with 93% of decision-makers now prioritizing network upgrades to maintain operational stability. The challenge is compounded by the fact that many organizations are still relying on legacy architecture that lacks the agility to scale dynamically as AI demand fluctuates. For many, the result is a forced delay in new AI deployments; 61% of those surveyed stated they are holding back on scaling their AI initiatives until they can ensure their network and security foundations are sufficiently robust to handle the increased load.
The Security and Visibility Hurdle in AI Adoption
Beyond physical capacity, the rapid proliferation of AI tools is creating significant security risks and visibility blind spots for enterprise IT teams. The survey found that 80% of respondents believe AI has expanded their attack surface, as employees increasingly utilize a wide variety of third-party AI tools that may not be fully integrated into corporate governance frameworks. This “shadow AI” presents a major observability challenge, as IT teams often lack the tools to identify or monitor the specific AI services running across their distributed networks. As noted in the report, if IT organizations cannot identify the traffic, they cannot effectively apply the necessary security guardrails to protect sensitive corporate data.
The difficulty of managing these risks is a primary barrier to entry for many firms. According to the research, IT leaders are finding it increasingly difficult to establish consistent policies for an ever-growing list of AI-driven applications and agents. This lack of control is leading to a cautious approach, where security posture has become the primary metric for determining the speed of future AI rollouts. Without enhanced observability—the ability to monitor and manage internal network processes in real-time—many firms remain in a state of uncertainty regarding their own AI-driven network demand.
Expanding AI Readiness Beyond the Data Center
Historically, the conversation surrounding AI infrastructure has been dominated by the needs of the data center, particularly regarding power consumption, cooling, and high-performance computing clusters for model training. However, the latest data suggests that this narrow focus is insufficient for enterprises that aim to integrate AI into their daily business operations. Because AI applications are increasingly embedded in the tools used by employees at the edge of the network, the campus and branch environments have become just as vital to AI success as the backend infrastructure.
Cisco’s analysis highlights that companies categorized as “aggressive AI adopters”—those with broad, enterprise-wide deployments—are significantly more likely to recognize the need for a holistic network strategy. Despite this, only 30% of these high-adopters reported that they are fully prepared to support the projected growth across their entire network footprint. This indicates a systemic need for a more integrated approach that bridges the gap between the data center and the end-user environment. As Jeetu Patel, Cisco’s president and chief product officer, noted, the network is now central to the entire AI infrastructure, effectively forcing a “networking supercycle” that requires immediate attention from leadership.
What Happens Next for IT Departments
For organizations looking to scale their AI initiatives, the path forward involves a shift toward AI-native networking solutions that offer automated observability and enhanced security features. Industry analysts and vendors are increasingly pointing toward the need for platforms that can dynamically allocate bandwidth and detect anomalies caused by agentic traffic. IT leaders are expected to focus their upcoming budget cycles on network infrastructure that can provide the visibility required to govern AI tools effectively.
The next major milestone for the industry will be the continued evolution of agentic AI, which is expected to further increase the volume of automated machine-to-machine communication. As companies move to resolve these capacity and security gaps, tracking the performance of these modernized networks will be essential. Readers interested in the latest developments in enterprise networking and AI integration can find official updates and further analysis through the Cisco Newsroom. We invite you to share your experiences with AI-driven network strain in the comments section below.