Nokia Predicts AI to Drive Nearly Half of Global Network Traffic by 2035, Shifting Towards Edge Infrastructure

Nokia's Global Network Traffic report, released today, forecasts a dramatic transformation in network infrastructure over the next decade, with AI traffic projected to constitute 44% of global network traffic by 2035. This represents a nearly quadrupling of total network traffic from 2026 levels, reaching almost 3,789 exabytes (EB) per month. The report emphasizes that the increasing adoption of autonomous AI agents and connected physical AI systems will fundamentally reshape network requirements, particularly by increasing demand for uplink capacity and real-time responses. This shift matters significantly to cloud and DevOps practitioners because it directly impacts infrastructure planning, network design, and operational strategies. The rise of agentic and physical AI means that traditional, download-heavy network models will no longer suffice. Instead, networks will need to support continuous, real-time connections and substantial uplink traffic from devices and sensors. This necessitates a move towards more distributed infrastructure, often referred to as the 'AI grid,' where intelligence and inferencing occur closer to the users and devices at the edge and metro core. For organizations building and deploying cloud-native applications, this implies a need to re-evaluate their current architectures to ensure they can handle these new traffic patterns and latency requirements. This trend aligns with the broader movement towards edge computing and distributed cloud architectures, which have been gaining momentum for several years. As AI models become more efficient and the need for low-latency processing grows, pushing computation closer to the data source becomes critical. Developments in Kubernetes for edge deployments, such as those discussed at events like Kubernetes on Edge Day, underscore this ongoing evolution. The report's projections reinforce the idea that the centralized cloud model, while still dominant for much AI traffic, will be complemented by a robust edge infrastructure, especially for use cases involving physical AI like robotics and autonomous vehicles, and agentic AI that requires immediate decision-making. In practice, this means cloud and DevOps teams should begin strategizing for a more hybrid and distributed operational model. This includes investing in edge infrastructure capabilities, optimizing applications for lower latency, and re-evaluating network designs to accommodate increased uplink traffic. Practitioners should also consider how their observability and security strategies will adapt to a more distributed environment. The report suggests that by 2035, 36% of all AI-native traffic will be served by edge and metro core infrastructure, highlighting a tangible shift that requires proactive planning. Organizations that fail to adapt their infrastructure to these evolving AI-driven network demands risk performance bottlenecks, increased operational costs, and an inability to fully leverage the potential of agentic and physical AI applications.