June 16: New research reveals that, globally three years from now AI will triple network traffic. In two years, most enterprises believe they will hit campus and branch network capacity limits. And attack surfaces are already expanding beyond what defenses can manage. This research released by Cisco in partnership with Foundry confirms that the rapid rise of large language models (LLMs) and the emerging wave of agentic AI is placing unprecedented strain on enterprise campus and branch networks. Alongside compute, the network is now a major factor in whether enterprise AI deployments succeed or fail.
3 years from now, network traffic triples as Agentic AI expands
A third of global organizations surveyed already have broad enterprise-wide agentic AI deployments, and 99% of organizations in India overall expect an expansion in agentic AI use within 24 months. Mature AI adopters globally also expect AI’s traffic impact to more than triple in the next three years; a staggering 235% increase.
This is because, unlike human users, AI agents operate at machine speed, triggering dozens of API calls, database lookups, and model inferences in seconds. They generate dense east-west traffic—lateral device-to-device or server-to-server communication required for AI agents to exchange data—that legacy workplace networks were never designed to handle.
“On the generative AI side, the traffic is a lot more north-south. On the agentic AI side, it’s going to be a lot of east-west… Usually, networks are designed for consistent traffic… suddenly, three agents are trying to talk to each other and solve a problem. How do we support increased east-west traffic?” said one Head of AI Strategy from a US tech firm interviewed for this research.
These same agentic AI workloads, that have the potential to transform enterprises, are also uniquely fragile. Networking leaders in India report that AI workloads are acutely vulnerable to networking issues; more sensitive to reliability and uptime (84%), bandwidth (80%), latency (75%), and packet loss (63%) than traditional applications.
In 2 years, network capacity reaches its limit
Fewer than one-third of mature AI adopters globally say their networks are fully prepared for projected AI growth. Overall, 71% of respondents in India admit they need upgrades, and 75% say they have hit, or will hit, campus and branch capacity limits within 24 months. Crucially, Wi-Fi is emerging as a major bottleneck for AI, with more than half of the global respondents listing it as the area driving the greatest increase in capacity requirements.
And there is still a disconnect between ambition and reality, with three quarters of global IT leaders agreeing that they are more confident in their organization’s AI strategy than in the network’s ability to deliver it. But, while 92% of Indian respondents cite budget constraints as a barrier, almost all enterprises are planning to modernize their workplace networks.
Attack surfaces are already expanding while observability is a challenge
AI has also created a challenging security environment, with the vast majority of Indian organisations saying that they are struggling to keep up (97%) and that AI has already caused some damage (93%). Over two thirds of global respondents also believe that AI-related threats are evolving faster than their ability to adapt, and that failing to adapt networks over the next two years will only increase security risks. Meanwhile, an observability gap is widening as traditional monitoring tools struggle with bursty, east-west agentic flows.
According to a VP of IT and Digital Infrastructure in education in the UK: “We’re just playing catch-up at the moment. It’s a worrying time, and I think it’ll stay like this for another 18 months or 2 years.”
Modernization is no longer optional
These findings make clear that network resilience, observability, and adaptive security are not supporting acts in the AI era, they are essential. The network has survived decades of transformation, from dot-com to the cloud, by adapting and evolving to meet the moment. Organizations that treat network modernization as a prerequisite to their AI strategy, rather than a parallel workstream, will define the next decade of enterprise AI.
Methodology
Foundry conducted a quantitative survey, co-designed and sponsored by Cisco, of 3,472 CIOs as well as networking, end user computing, and technology leaders in Asia-Pacific, Europe, the Middle East, Latin America, and North America. The respondents work at organizations with 500+ employees that have an average of 3,292 campus/branch locations. In addition, Foundry conducted six in-depth interviews with executives in Asia-Pacific, Europe, and the United States. All research was conducted between March and April 2026.