When Intelligence Overloads Infrastructure: A Forecast Model for AI-Driven Bottlenecks (2511.07265v1)

Abstract: The exponential growth of AI agents and connected devices fundamentally transforms the structure and capacity demands of global digital infrastructure. This paper introduces a unified forecasting model that projects AI agent populations to increase by more than 100 times between 2026 and 2036+, reaching trillions of instances globally. In parallel, bandwidth demand is expected to surge from 1 EB/day in 2026 to over 8,000 EB/day by 2036, which is an increase of 8000 times in a single decade. Through this growth model, we identify critical bottleneck domains across access networks, edge gateways, interconnection exchanges, and cloud infrastructures. Simulations reveal that edge and peering systems will experience saturation as early as 2030, with more than 70% utilization of projected maximum capacity by 2033. To address these constraints, we propose a coevolutionary shift in compute-network design, emphasizing distributed inference, AI-native traffic engineering, and intent-aware orchestration. Security, scalability, and coordination challenges are examined with a focus on sustaining intelligent connectivity throughout the next digital decade.