AI supercycle is transforming global network traffic not only in volume but in shape and behaviour. While data volumes have surged for years due to video streaming and general internet use, AI-driven traffic — including large model training, inference workloads and real-time interaction — creates new patterns that traditional networks weren’t designed to handle. Networks must therefore evolve alongside AI, rather than just scale up in capacity, to support emerging applications and services.
According to Nokia’s latest Global Network Traffic Report, AI traffic is expected to grow significantly faster than non-AI traffic over the next decade, even though a majority of traffic will still come from conventional sources. The report highlights that AI-related workloads — including data movement for training, inference, and distributed AI coordination — are reshaping the nature of how data flows across wide area networks (WANs). This shift demands more symmetric capacity, lower latency and increased responsiveness in network design.
The article also points out that immersive and interactive digital experiences — such as virtual collaboration, real-time 3D modelling, and advanced gaming — are further changing traffic patterns by making two-way, low-latency connections more common. Unlike traditional streaming that primarily sends content one way to a user, modern AI and interactive workloads require continuous and balanced data exchanges between edge devices, data centres, and cloud AI platforms, pushing networks beyond their historical “downlink-heavy” models.
To adapt, network operators will increasingly rely on AI-driven automation and optimization within the network itself. AI tools can help communication systems predict traffic surges, allocate resources dynamically and maintain performance as demands evolve. Meanwhile, advances such as AI-native radio access networks (AI-RAN) and early work toward 6G connectivity aim to support these fundamentally different traffic patterns and accelerate the transition to future network architectures.
