AI growth is outpacing infrastructure – the ceiling will soon be reached

Network traffic is expected to triple in three years due to continued AI development. To handle this increase, a large-scale and rapid modernization of infrastructure is required, according to a new study from Cisco.

Agentic AI presents enormous opportunities for business, but also major challenges for today's digital infrastructure. Over the past year, business network traffic grew by an average of 34 percent, and the growth is accelerating. Traffic is expected to double in the coming year and more than triple over the next three years.

This is evident from a global survey from Cisco with more than 3,400 IT decision makers. A large majority agree that AI will continue to grow rapidly and that infrastructure will sooner or later reach its capacity ceiling. A full 73 percent of respondents say that the point at which the network can no longer handle capacity needs is no more than two years away, and some are already experiencing capacity shortages today.

Even among companies considered leaders in the first wave of AI, there is growing pressure: Only 30 percent of these organizations feel their infrastructure is ready to support the expected AI growth.

The challenges are not just about the increasing amount of traffic, but also about the paths that traffic takes. Traditional network infrastructure is built to prioritize traffic in and out of the data center. But the areas where the load increases most rapidly as a result of agentic AI are the so-called lateral communications between internal devices, servers, and applications.

These are communication paths that, when automated and occurring at machine speed, can quickly overload the infrastructure.

Additionally, agentic AI workloads are more sensitive to network disruptions, capacity issues, and latency than traditional applications.

Security is also highlighted as a major challenge. 78 percent of respondents expect increased security risks as a result of the increased use of AI. At the same time, 71 percent feel that threats are evolving faster than existing protection measures can adapt, and 69 percent believe that it will be more difficult to monitor and gain an overview of the infrastructure, which creates more so-called blind spots.

A third of the companies in the survey have already started applying agentic AI and 85 percent expect to expand their use significantly or greatly over the next two years.

At the same time, the report authors note that there is a growing gap between AI ambitions and how well prepared networks actually are. Development is moving much faster than many businesses previously anticipated, making infrastructure modernization a strategic imperative.

“There is a serious imbalance between campus and branch networks and the scale, speed and variety of AI-driven traffic now moving through networks. As businesses increase the pace of AI usage Networks are under pressure from three directions: rapidly increasing traffic, rapidly approaching capacity ceilings, and increasing security complexity.